Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Telomere length measurements using fluorescence microscopy 1997

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_1998-272273.pdf
ubc_1998-272273.pdf [ 27.89MB ]
Metadata
JSON: 1.0064855.json
JSON-LD: 1.0064855+ld.json
RDF/XML (Pretty): 1.0064855.xml
RDF/JSON: 1.0064855+rdf.json
Turtle: 1.0064855+rdf-turtle.txt
N-Triples: 1.0064855+rdf-ntriples.txt
Citation
1.0064855.ris

Full Text

Telomere Length Measurements Using Fluorescence Microscopy S T E V E N SUI S A N G P O O N , P . E n g . M . A . S c , The Univers i ty of B r i t i s c h C o l u m b i a , 1989 B . A . S c , The Univers i ty of B r i t i s c h C o l u m b i a , 1985 A T H E S I S S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R O F P H I L O S O P H Y in T H E F A C U L T Y O F G R A D U A T E S T U D I E S Depar tment of Elec t r ica l Engineer ing We accept th is thesis as conforming to the required s tandard T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A December 1997 © Steven S u i Sang Poon, 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of <5^2Q^72«tX <5v The University of British Columbia Vancouver, Canada Date DE-6 (2/88) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY i i Abstract We describe a system to estimate the lengths of telomere repeat D N A sequences i n i n d i v i d u a l chromosomes of cells. Conven t iona l systems based on gel electrophoresis of digested D N A can only determine the telomere length d i s t r ibu t ion of a popula t ion of cells bu t cannot provide es t imat ion of telomere lengths of i n d i v i d u a l chromosomes u s i n g a l imi ted n u m b e r of cells . O u r me thod is based on ana lyz ing microscopy images of metaphase chromosomes prepared u s i n g fluorescence in-situ hybr id iza t ion technology. In order to obta in rel iable a n d reproducible images for measurements , we have developed a n d opt imized a n image acquis i t ion system specifically for this purpose u s i n g commerc ia l ly available components . In th is s tudy, two types of images are used . The first image h ighl ights the chromosome regions. The second highl ights only the telomere regions a n d consis ts of a set of mul t i - focus plane images. We first perform segmentat ion on the acqui red mul t i - focus plane images i n order to determine the region that each telomere occupies. For each telomere region determined, the total integrated fluorescence intensi ty (IFI) is obtained. T h i s ca lcula ted IFI va lue is no rma l i zed for spat ia l unevenness i n i l l u m i n a t i o n i n the field of view as we l l as the day to day var ia t ion i n the i l l u m i n a t i o n intensi ty . The acqui red chromosome image is then segmented u s i n g novel a n d s imple a lgor i thms developed for s u c h purpose. The locat ion of the segmented telomere a n d chromosome boundar ies are then overlaid onto the chromosome image. Different colour boundar ies are u s e d to h ighl ight a n d identify different TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY i i i telomeres w i t h i n a chromosome as we l l as the n u m b e r of telomeres detected w i t h i n each chromosome. The resu l t ing image a n d the ca lcu la ted telomere IFI va lues for each chromosome are then presented on the computer screen for use r verif icat ion a n d edi t ing as required. The a lgor i thms descr ibed i n th is thesis are present ly be ing u s e d on a dai ly bas is to collect da ta on telomeres and to s tudy the role telomeres p lay i n the biology of cells . O u r method of analys is has made i t possible to generate rel iable estimates of the length of telomeres i n i n d i v i d u a l chromosome a rms u s i n g a l imi t ed n u m b e r of cells. In addi t ion , our au tomat ion of the ana lys i s has s ignif icant ly reduced the user in teract ion t ime for edi t ing a n d verif icat ion of the resul ts . To date, at least two different biological s tudies of telomeres have been completed on the system we developed. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY iv Table of Contents Abstract ii Table of Contents iv List of Tables ix List of Figures x Acknowledgments xiii Chapter 1.Introduction 1 1.1. D o c u m e n t In t roduct ion a n d Overview 1 1.2. Telomeres a n d Thei r F u n c t i o n i n The C e l l 5 1.3. Telomere Quant i f ica t ion Methods 9 1.3.1. Conven t iona l Technique: Sou thern A n a l y s i s 9 1.3.2. New Technique: Quant i ta t ive F I S H a n d Image A n a l y s i s 10 1.4. Objectives 12 Chapter 2.Background 15 2 .1 . Imaging Systems 15 2 .1 .1 . Overview 15 2.1.2. Wide-f ield Microscopes 20 2.1 .3 . Confocal Microscopes 24 2.2. Image A n a l y s i s 28 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY v 2.2 .1 . Overview 28 2.2.2. Image Pre-Processing 30 2.2 .2 .1 . B a c k g r o u n d Subt rac t ion 30 2.2.2.2. Flat-f ield Compensa t ion 31 2.2.2.3. Wavelength Compensa t ion 32 2.2.2.4. Photobleaching Compensa t ion 32 2.2.3 . F o c u s and Three D imens iona l Recons t ruc t ion 33 2.2.4. Image Segmentat ion 35 2 .2 .4 .1 . Overview 35 2.2.4.2. Thresho ld ing or C lus t e r ing 36 2.2.4.3. Edge Detect ion 38 2.2.4.4. Region Ex t rac t ion 39 Chapter 3.Imaging System 41 3.1. Overview 41 3.2. Image Acqu i s i t i on Hardware 42 3.2 .1 . Overview 42 3.2.2. Fluorescence Microscope 43 3.2.2 .1 . I l lumina t ion Source 45 3.2.2.2. Exc i t a t ion and E m i s s i o n Fi l ters 46 3.2.2.3. Objective Lens 47 3.2.3. F o c u s s i n g M e c h a n i s m 49 3.2.4. H i g h Reso lu t ion C a m e r a 50 3.2.5 . C o m p u t i n g Sys tem 52 3.3. Sys tem Tempora l Stabi l i ty and Aberra t ions 53 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY v i 3.3.1 . Overview 53 3.3.2. Tempora l F luc tua t ions i n I l lumina t ion 53 3.3.3. Photobleaching Effects 56 3.3.4. Uneven I l luminated F ie ld of V iew 58 3.3.5. F l a t -F i e ld Compensa t ion Resul ts 61 3.4. Image A c q u i s i t i o n Software 62 3 .4 .1 . Overview 62 3.4.2. Image Exposure and Photometric Range Select ion 63 3.4.3. Image Pre-Processing: Sensor Defects 64 3.5. Image A n a l y s i s Sys tem 65 3.5 .1 . Image Ana lys i s Hardware 65 3.5.2. Image Ana lys i s Software 65 Chapter 4.Acquisition System Characteristics 70 4 .1 . B a c k g r o u n d 70 4.2. O u r Der iva t ion of the Sys tem O T F 75 4 .3 . Theoret ica l O T F and P S F Resul t s 80 4.4. In i t ia l O T F / P S F Comparat ive S tudy 85 4 .5 . C o m p a r i s o n W i t h Exper imen ta l P S F 89 Chapter 5.Telomere Segmentation and Integrated Fluorescence Intensity Measurements 94 5.1. Overview 94 5.2. Theory i n IFI Quant i f ica t ion 96 5.3. Segmentat ion a n d IFI Quant i f ica t ion A lgo r i t hm 100 5.4. N u m b e r of Focus Planes Required 108 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY v i i 5.5. A l g o r i t h m Eva lua t i on A n d Va l ida t ion 112 5.5.1. Overview 112 5.5.2. Spa t ia l Reso lu t ion of IFI Segmentat ion 114 5.5.3. S imu la t ed Objects 115 5.5.4. F luorescent Beads 121 5.5.5. P lasmids 122 5.5.6. S u m m a r y of A lgo r i t hm Va l ida t ion Resul t s 124 5.6. H u m a n Telomeres Resul t s 126 5.6.1. Number of Focus Planes Requi red 126 5.6.2. Telomere D i s t r i bu t ion i n Cel ls 130 5.7. Chapter S u m m a r y 131 Chapter 6.Segmentation of Chromosomes 133 6.1. Overview 133 6.2. R a n k Difference Fi l ter 136 6.3. C o m p a r i s o n of Edge Detectors 144 6.4. F i r s t Approx ima t ion to Edges: Thresho ld ing 154 6.5. Second Approx ima t ion to Edges: Texture Detect ion 157 6.6. T h i r d Approx ima t ion to Edges: Region Refinement a n d Labe l ing 163 6.7. Feature Ex t rac t ion and Artifact Removal 166 6.8. Associa te Telomere w i t h Chromosome 166 6.9. Segmentat ion Performance 168 Chapter T.Conclusion and Future Suggestions 172 7.1. Overview 172 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY v i i i 7.2. Sys tem Performance 173 7 .2 .1 . Imaging system 176 7.2.2. Sys tem Character is t ics 177 7.2.3. Telomere IFI V a l u e 177 7.2.4. Chromosome Segmentat ion 178 7.3. C u r r e n t Bio log ica l Studies 179 7.4. Fu tu re Suggest ions a n d Appl ica t ions 182 Chapter 8.Bibliography 184 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY i x List of Tables 5.1. IFI values of typ ica l h u m a n telomeres at different focus levels 119 5.2. IFI va lues of s imula ted test objects ca lcula ted at different focus level s amp l ing spacings 120 5.3. IFI values of different size beads 121 5.4. IFI values of different size p lasmids 123 5.5. Normal ized IFI values of objects at different focus posi t ions 125 5.6. IFI values of typ ica l h u m a n telomeres at different focus levels 126 5.7. IFI values of h u m a n telomeres ca lcula ted at different focus level s amp l ing spacings 129 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY x List of Figures 1.1. D i a g r a m of telomere locat ion i n a metaphase chromosome 6 1.2. Telomere lengths d u r i n g chromosome dup l i ca t ion pr ior to cel l d i v i s i o n 8 1.3. Fluorescence image of telomeres and chromosomes 13 2 .1 . B l o c k d iagram of a typ ica l imaging system 17 2 .2 . M o d e l of a microscope system 22 2 .3 . P r inc ip le of confocal microscopy 2 5 2.4. Process for cel l ana lys i s 29 3 .1 . B l o c k d iagram of the imaging system 43 3.2. B l o c k d iagram of the exci tat ion and emiss ion filter system 48 3.3. I l lumina t ion var ia t ions over time 55 3.4. Photobleaching effects on telomere fluorescence 56 3.5. Photobleaching effects on bead fluorescence 57 3.6. I l lumina t ion var ia t ion over the field of view 59 3.7. Intensities of the central ros of pixels of a homogenous sample 62 3.8. F la t field compensa t ion for spa t ia l i l l umina t i on var ia t ions 62 3.9. Telomere ana lys i s program 67 4 . 1 . Rela t ionship of the in-focus and defocus dis tances i n the object a n d image side of the objective lens 73 4 .2 . Theoret ical OTFs of the system i n the xy-plane 81 4 .3 . Theoret ical P S F s of the system i n the xy-plane 82 4.4. Theoret ical P S F of the system i n the xz-plane 82 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY x i 4 .5 . Theoret ical P S F d i s t r ibu t ion of the system at var ious z-spacings 83 4.6. In-focus system response for var ious objectives 88 4.7 . Typ ica l images of four 0. l m beads acquired at different focus spacings 90 4.8 . C o m p a r i s o n of exper imental and theoretical P S F d is t r ibu t ions as a funct ion of z-focus pos i t ion 91 5.1. E r ro r s i n ca lcu la t ing the object IFI 102 5.2. A p p l i c a t i o n of the average difference filter 105 5.3. Process of segmenting a typ ica l telomere image 106 5.4. S imu la t ed test objects for spa t ia l resolut ion estimates 115 5.5. S imu la t ed test object va lues and shapes 118 5.6. Normal ized IFI va lues at va ry ing focus of different objects 119 5.7. IFI d i s t r ibu t ion of different size beads 122 5.8. IFI D i s t r i bu t i on of Different Size P l a smids 124 5.9. IFI of typ ica l h u m a n telomeres at different focus levels 127 5.10. Telomere IFI d i s t r ibu t ion i n a cel l 131 6 .1 . Cen t ra l p ixe l of r a n k difference filter 139 6.2. Shape of r a n k difference filter region 140 6.3. Size of r a n k difference filter region 141 6.4. Effect of va ry ing upper a n d lower r a n k number s 143 6.5. Effect of addit ive noise and vary ing upper a n d lower r a n k numbers . . . . 144 6.6. Performance of edge filters on test object 145 6.7. Performance of edge detectors i n presence of noise added to the test object 146 6.8. Performance of edge detectors on the peppers image 148 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY x i i 6.9. Edge filter performance of peppers image w i t h addit ive G a u s s i a n noise 150 6.10. Edge filter performance of peppers image w i th addit ive un i fo rm noise 151 6 .11 . Edge filter performance on chromosome image 153 6 .12. Chromosome image at var ious thresholds 155 6 .13. Second approx imat ion to edges: texture detection 159 6.14. R a n k difference of the average difference image 162 6 .15 . T h i r d approx imat ion to edges: region refinement a n d labe l ing 165 6.16. Chromosome a n d telomere segmentation resul ts 168 6 .17. C o m p a r i s o n of our segmentation a lgor i thm to the C a n n y filter 169 7 .1 . Telomere lengths of i n d i v i d u a l chromosomes i n a cel l 181 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY Xlll Acknowledgments I a m most grateful to m y supervisory committee for their guidance, advice, support , a n d patience throughout this work. In par t icu lar , I a m i n debt to D r . Rabab W a r d , m y supervisor, for her t ime a n d s o u n d advice d u r i n g our d i scuss ions , pa r t i cu la r ly w i th respect to the t echnica l i ssues . She was also very i n s t rumen ta l i n provid ing alternative methods to solve var ious problems encountered th roughout the project. I a m also very grateful to Dr . B r a n k o Pa lc ic , who i s not only a member of m y supervisory committee bu t also an excellent mentor a n d vis ionary . Wi thou t h i m , I w o u l d not be able to apply m y imag ing methods for use i n a n new a n d interest ing b io logica l research appl ica t ions . Las t ly , I a m most appreciative to D r . Peter Lansdorp , another member of m y supervisory committee, for h i s support , facili t ies, a n d resources for the biological aspects of th is project. H i s eagerness i n the telomere project, h i s v i s i o n into the direct ions of h o w the sys tem w o u l d be used , h i s unde r s t and ing of the impact as we l l as the l imi ta t ions of au tomat ing some of the processes i n the biological s tudies, a n d the regular feedback from h i m a n d h i s group d u r i n g the development of the project were key i n m a k i n g th i s project successful . In addi t ion , I w o u l d l ike to t hank a l l the staff a n d s tudent members of the B . C . Cance r Research Centre who have ass is ted me i n th is thesis . Par t icu la r ly , I w o u l d l ike to t hank the group involve w i t h telomere research i n the Terry F o x Laboratory for Hemato logy/Oncology (Dr. Peter M . Lansdorp , D r . Uwe, M . Mar tens , D r . J . M a r k J . M . Zi j lmans , a n d L i z Chavez). I w o u l d l ike to TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY xiv give spec ia l t hanks to Dr . Uwe M . Mar tens for h i s help i n p repar ing the microscope sl ides, acquis i t ion of some of the images u s e d i n th i s thesis , a n d mos t impor tant ly , offering construct ive c r i t i c i sm a n d feedback on the use of the telomere image analys is software du r ing its development. I w o u l d also l ike to t h a n k M r . Hector H u a n g for modifying some of the code i n the acqu is i t ion p rogram to incorporate the z-drive control a n d automat ic exposure t ime sett ing. In addi t ion . I w o u l d l ike to t hank M r . B o n g C h o u n M i n g for process ing the test images u s i n g the C a n n y a n d Difference of G a u s s i a n edge filters. T h i s work is suppor ted by scholarsh ips from the Science C o u n c i l of B r i t i s h C o l u m b i a a n d grants from the Nat iona l Insti tutes of Hea l th , the M e d i c a l Research C o u n c i l of Canada , the Nat iona l Cancer Institute of C a n a d a , the G e r m a n Science Founda t i on (U.M.M.) a n d the D u t c h Cancer Society ( J . M . J . M . Z . ) , a n d equipment from X i l l i x Technologies Corp . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 1 Chapter 1. Introduction 1.1. Document Introduction and Overview T h i s thesis describes a new method for de te rmining the length of telomeres i n a cel l . Telomeres are nucleo-prote in complexes con ta in ing specific D N A repeat sequences found at the ends of each of the 46 chromosomes i n a h u m a n cel l . These repeat sequences represent approximate ly one ten thousand ths of the total D N A i n the cell . D u e to their minu te size, the task of quant i fy ing the n u m b e r of repeat sequences at i n d i v i d u a l chromosome ends has not been accompl i shed before. The method we developed has made i t possible to estimate the length of i nd iv idua l telomeres a n d to perform new studies i n telomere biology. A l l cel ls , i n c l u d i n g h u m a n , are k n o w n to divide. E a c h t ime cells divide, their n u m b e r doubles . In most somatic cells, the m a x i m u m n u m b e r of t imes the ce l l -d iv i s ion cycle can repeat i tself i s est imated to be 70 to 100 t imes. After that, cells die or become senescent: they consume less food a n d their membranes deteriorate; a process that is very s imi la r to aging. To s tudy the bas is of ce l lu lar aging a n d mortal i ty , current s tudies focus on a s m a l l area at each tip of each chromosome w h i c h is cal led the telomere (from telos = end a n d meros = part). A telomere consis ts of repetitive sequences of base pa i rs w h i c h do not appear to code for any traits and associated proteins. The repetitive D N A sequences appear to protect the ends of chromosomes a n d provide a cap. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 2 E a c h t ime a cel l divides, some of the repeats are lost a n d the telomeres i n the daughter cells become shorter. F ina l ly , w h e n the telomeres reach a c r i t i ca l length , the cel l stops d iv id ing (typically after 50 to 100 repl ica t ion rounds) . Some researchers theorize that w h e n the telomere reaches its c r i t i ca l length , cer ta in genes become active a n d produce proteins that trigger t i ssue deter iorat ion associated w i t h aging. Whi le almost every cel l i n the h u m a n body exhibi ts telomere loss , a few s u c h as sperm, egg, a n d cancer cells do not. S u c h cells are character ized by their abi l i ty to divide not j u s t 100 t imes b u t thousands . Researchers are u s i n g what they k n o w so far about telomeres a n d other ce l lu lar mechan i sms to at tack the diseases that keep the very o ld from becoming even older. Other researchers are s tudy ing h o w to b lock the telomerase R N A enzyme (which helps to lengthen telomeres) i n cancer cel ls , l ead ing to wi ther ing of telomeres and the death of the no longer so prolif ic cells . In h i s recent new novel , "Holyfire", science f ict ion author B r u c e Ster l ing popu la r i zed telomeres. He describes a procedure by w h i c h h i s heroine, a very r i c h 95 year o ld w o m a n , gets a complete cel lular makeover i n w h i c h the telomeres w i t h i n every cel l of her body were lengthened. A s a result , she was t ransformed into a heal thy 20 year o ld y o u n g w o m a n . A l t h o u g h th i s p h e n o m e n a is not l ike ly to occur i n the near future, th is thesis w i l l give researchers a tool w h i c h can help better s tudy the role of telomeres i n biology a n d perhaps , get t hem closer to this goal. Alternat ively, new facts about telomeres uncovered by s u c h studies may rejuvenate en thus i a sm i n b io logica l sciences. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 3 O u r method for de termining telomere length is based on app ly ing image ana lys i s techniques to microscope images of chromosomes prepared u s i n g fluorescence in-situ hybr id iza t ion (FISH). O u r method provides more detai led informat ion about telomere length and is considerably more sensit ive t h a n the convent iona l me thod based on gel electrophoresis. In addi t ion , our technique enables s tudies on telomeres of single chromosomes a n d on telomere length d i s t r ibu t ions w i t h i n cells. B a s e d on this thesis work , telomere length var ia t ions w i t h i n i n d i v i d u a l chromosomes and amongst different chromosomes i n cel ls , c a n now, for the first t ime, be convenient ly determined. Information about telomere lengths i s u s e d i n var ious s tudies to investigate the role of telomeres i n the biology (aging and mal ignant transformation) of cells from h u m a n a n d other species. Bes ides descr ib ing the importance of this research, the present chapter gives a n overview of the thesis. It describes what telomeres are a n d h o w the technology descr ibed i n th is thesis can he lp researchers to further u n d e r s t a n d the role of telomeres i n bas ic cel l biology a n d molecular genetics. It in t roduces our technique for detecting i nd iv idua l telomeres a n d measu r ing their lengths u s i n g fluorescence microscope imaging a n d image analys is systems. The objectives of th is research are also presented i n th is chapter. Chap te r 2 presents the backg round of imaging systems a n d image ana lys i s a lgor i thms for cel l ana lys is . The accuracy of identifying and quant i fying the chromosomes a n d the telomeres is h igh ly dependent on the focus pos i t ion of objects i n the microscope images. Hence, background on three d imens iona l (xy-spatial a n d z-focus plane) image analys is techniques are also covered i n th is chapter . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 4 Chapte r 3 describes the image acquis i t ion system that we developed a n d u s e d i n th is research. A s there are no commercia l ly available systems for th is work , the rat ionale, b e h i n d our selection of components is descr ibed w i t h emphas i s on the key components i n the system. Image ca l ibra t ion of the acqu i red images i s also covered i n this sect ion. Chapter 4 presents a new a n d more accurate me thod for ana lyz ing a n d formula t ing the character is t ics of the hardware system. We derive the character is t ics of the sys tem a n d compare t hem w i t h the experimental resul ts . Chapter 5 describes the theory a n d a lgor i thms u s e d to quantify telomere lengths / f luorescence i n the three- d imens iona l plane. The theoretical analys is leads to a new and s impl i f ied me thod for accurate telomere quantif icat ion. O u r new method direct ly extracts in format ion from the mul t i - focus plane images of the telomeres wi thou t h a v i n g to perform 3D image reconst ruct ion. Chapter 6 out l ines a novel chromosome segmentat ion process w h i c h yields a good approx imat ion to the border of chromosomes ( inc luding those w h i c h are touching). For th is purpose , we developed a new a lgor i thm cal led the R a n k Difference Fi l ter . T h i s filter c a n act as a n edge detection filter as wel l as a selective morphologic d i l a t ing /e ros ion filter. The resul ts from the telomere segmentat ion are then l i n k e d to the chromosome segmentat ion resul ts and presented to the user for ver if icat ion a n d edit ing. F ina l ly , Chapter 7 summar izes the conc lus ions of th is research a n d out l ines possible areas for future invest igat ion. The recent advances i n the fluorescence in-situ hybr id iza t ion (FISH) technology has made it possible to quanti tat ively s ta in specific gene sequences s u c h as those i n telomeres. E m p h a s i s of th is thesis w i l l concentrate on the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 5 engineering aspects of the project. A brief descr ip t ion of the biological aspects is presented for better comprehens ion of the project. Specifical ly, the descr ip t ion of the telomeres a n d the F I S H technology are presented i n th is Chapter . 1.2. Telomeres and Their Function in The Cell Telomeres conta in proteins and specific D N A sequences found at the n a t u r a l ends of chromosomes i n eukaryot ic (nucleated) cells (Figure 1.1). D N A is composed of a un ique sequence of base-pairs w h i c h enables differentiation of i nd iv idua l s w i t h i n a single species. E a c h base-pair can be one of the t hymine (T), adenine (A), cystosine (C), or guanine (G) nuc le ic acids , pa i red w i t h i ts counterpar ts (A w i t h T, a n d C w i t h G). Specifically, telomeric D N A cons is t s of h igh ly repetitive base-pair sequences w h i c h are u n l i k e the rest of the D N A sequences i n the chromosome. (Benbow, 1992; W i l s o n et a l . , 1993) Depend ing on the species, a single telomeric D N A conta ins a r o u n d 20 to 15,000 base-pairs cons is t ing of repetitive 6 to 8 base-pair sequences cal led repeats. W i t h i n a single species, however, the length of the telomeric sequence is c losely control led. In h u m a n s for example, the repetitive sequence is 6 base- pa i rs long a n d consis ts of T T A G G G . The length of the h u m a n telomere ranges from a r o u n d 1000 to 15,000 base-pairs (1 /10000 of the total D N A ) . In total , there are 46 chromosomes i n a h u m a n cel l . S ince each chromosome has a telomere at each end, there are 92 telomeres i n each cel l . A s chromosomes are most often s tud ied after dup l ica t ion bu t before separat ion (metaphase chromosomes) , a total of 184 (twice the number) telomeres can be observed i n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 6 t yp ica l chromosome slide preparat ions of single cells . (Benbow, 1992; W i l s o n et a l . , 1993) F igure 1.1. D iag ram of telomere locat ion i n a metaphase chromosome. Telomeres are located at the ends of chromosomes. In condensed chromosomes , they may appear to only occupy a por t ion of the chromosome t ip . Telomeres are k n o w n to p lay a n d are postula ted to have a n u m b e r of roles i n the funct ion of the cel l (Benbow, 1992; W i l s o n et a l . , 1993). The bas ic func t ion of the telomere is to cap a n d protect the ends of chromosomes . T h i s funct ion is media ted by specific repeat sequences of the telomere w h i c h are not present i n the rest of the D N A . Hence, the cel l i s able to differentiate between the ends of a chromosome a n d a break w i t h i n . B y detecting the break (telomere repeat sequence not present at the chromosome end), the cel l c an appropria te ly ini t ia te i ts repair m e c h a n i s m to repair the damage. Telomeres also appear to p lay a role i n the nuc lear archi tecture of the ce l l . Specific TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 7 proteins interact w i t h the telomere a n d at tach each chromosome to pa r t i cu la r sites on proteins of the nuc lear matr ix . The specific loca l iza t ion of each chromosome w i t h i n the cel l may determine h o w each chromosome funct ions . Fur thermore , d u r i n g meiosis (a special form of cel l d iv i s ion giving rise to spe rm or egg cells), the telomeres at tach themselves to the nuc lea r envelope, r e su l t ing i n a stage that promotes proper chromosome pa i r ing . Telomeres m a y p lay a role i n gene regula t ion as wel l . It has been postula ted that the ' l eng th of the telomere may determine i f par t icular genes i n that chromosome are expressed or depressed. Telomeres also play a n important role i n n o r m a l cel l d iv i s ion . A s s h o w n i n F igure 1.2, a n R N A pr imer first attaches i tself w i t h i n the first 200 base-pairs at the 3' (G-riched) end of a single s t rand of chromosome. After the p r imer has been at tached, the rest of the chromosome is t ranscr ibed or repl icated j o i n i n g each nuc le i c ac id w i t h its complement (A wi th T, a n d C w i t h G) by D N A polymerase enzymes w h i c h can only dupl icate D N A i n the 5' to 3' d i rec t ion of the newly formed s t rand. At the end of the process, the R N A pr imer is cleaved or removed. A s a result , a por t ion of the parent 3' telomere s t r and m a y not be ful ly repl icated i n the daughter cel l . A n enzyme, telomerase, can increase the 3' l ength of telomeres bu t is generally not expressed i n n o r m a l somat ic (non- germ line) cells . It has been recently shown that telomere lengths shor ten w i t h age u n t i l the telomeres reach a cer tain length w h i c h prevents the ce l l f rom d iv id ing (Harley et a l . 1990; Al l sopp et a l . 1992; Levy et a l . 1992; V a z i r i et a l . 1994). The re la t ionship of telomere lengths a n d its re la t ionship w i t h cancer TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 8 a n d w i t h age is a n active research area (Hastie et a l . 1990; de Lange 1995; Lansdorp et a l . 1996; L u n d b l a d and Wright , 1996). ( C C C T A A ) , (a) Parent C h r o m o s o m e Strands ( T T A G G G ) n 1 3 ' R N A Pr imer ( G G G A T T ) n 0 ( A A T C C C ) n (b) Repl icated St rands <̂ 3 5' (c) T w o Sis ter C h r o m o s o m e s <̂  - | - Te lomere Repl icat ion L o s s 5' Figure 1.2. Telomere length d u r i n g chromosome dup l i ca t ion pr ior to cel l d iv i s ion , a) A n R N A pr imer attaches anywhere w i t h i n the first 50 to 200 base- pa i r s of the 3' end. b) Once the pr imer is at tached, the chromosome replicates, c) A t the end of the repl ica t ion process, the R N A pr imer is removed. The resu l t ing sister chromosomes w o u l d then have shorter telomeres t h a n the parent. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 1.3. Telomere Quantification Methods 9 1.3.1. Conventional Technique: Southern Analysis The convent ional technique for measu r ing the length of telomere repeat sequences i s based on a method cal led Southern ana lys i s (Allshire et a l . 1988; de Lange et a l . 1990). In this technique, D N A is first digested by enzymes to ob ta in segments conta in ing the telomeric repeats. The resu l t ing fragments are deposi ted onto a gel and the process of electrophoresis is ini t ia ted. The smal ler D N A fragments move further i n the gel t h a n the larger ones. The dis tance these fragments move i n the gel, under the force of the electric field, has a logar i thmic re la t ionship to the size of the D N A fragments (telomeres). A radio label led probe specific for the T T A G G G telomeric repeat sequence i s then a l lowed to hybr id ize (bind) to the telomeric repeats i n the (blotted) gel. After w a s h i n g off the excess, the presence of radiolabel led probe c a n be v i sua l i zed on a n X - r a y plate to record the f inal dest inat ion of the telomeric repeat segments onto f i lm. The d i s t r ibu t ion of telomere lengths w i t h i n a sample c a n then be obta ined. The size d i s t r ibu t ion of telomere segments c a n be determined from the spot s igna l intensit ies a n d their respective distance from the s tar t ing point . There are a n u m b e r of d rawbacks to th is technique. F i rs t , at least 100 ,000 cells are required to give enough D N A to obta in a good representat ion of the telomeric lengths w i t h i n a sample. Second, the size of the telomere m a y be over-estimated. The digestive enzymes used typica l ly cleaves at the sub- telomeric region resu l t ing i n telomeric segments w i th add i t iona l base-pai rs that do not have the telomeric repeat sequence. Hence the apparent telomere size i s TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 10 increased a n d over-estimated. E v e n i f the same enzyme i s used , the a m o u n t of non-telomere repeats w i t h i n the segments may vary from chromosome to chromosome. Las t ly , the size and n u m b e r of short telomeres i n the sample are under-represented. Short telomeres of s imi la r lengths are more d ispersed i n the gel compared to the longer telomeres because of the logar i thmic size re la t ionship . Hence, a larger n u m b e r of short telomeres are requi red to generate the same or s imi la r s ignal intensit ies as those generated b y long telomeres. D u e to the l imi ted dynamic range of the f i lm, the t ransducers a n d the logar i thmic representat ion of telomere size, longer telomeres are favored as their s ignals are stronger a n d the short telomeres become under-represented. The system descr ibed below overcomes these difficulties. Instead of a r o u n d 100,000 cells , less t han 30 cells are required to measure the telomere length d i s t r ibu t ion . The size of the telomeres obta ined i s not unde r or over- est imated a n d the obtained telomere signals are not b iased i n their lengths. A l t h o u g h our system cannot analyze cells w h i c h do not divide (e.g. senescent cells), m a n y other b iological s tudies on telomeres c a n s t i l l be car r ied out u s i n g the sys tem we developed. 1.3.2. New Technique: Quantitative FISH and Image Analysis A new technique based on the fluorescence in-situ hybr id iza t ion (FISH) protocol has been developed to measure telomere lengths (Lansdorp et a l . 1996). F I S H technology relies on probes (nucleic ac id sequences) that c a n hybr id ize (bind base-pairs) to specific sites i n denatured chromosomes . B y a t taching a fluorescence labe l onto the probe, the loca t ion of specific sites i n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 11 the chromosome c a n be identified under a fluorescence microscope. A l t h o u g h the F I S H technique has been a round for the past 10-15 years, i t i s not u n t i l recently that probes have been developed w h i c h hybr id ize to telomeres w i t h sufficient efficiency s u c h that quantitative analys is of F I S H images c a n be considered. Synthet ic peptide nuc le ic ac id (PNA) oglionucleotide probes are u s e d for th i s purpose. It was found that s u c h probes c a n hybr id ize unde r condi t ions that do not favor D N A to D N A a n d D N A to R N A hybr id i za t ion (Nielsen et a l . 1991; Egho lm et a l . 1993). A s a result , these new probes do not need to compete w i th the su r round ing D N A or R N A to b i n d to the telomeric repeat sites. In our s tudy, two types of images of the cel l are used . The first one is a n image (or images at different focal planes) of the telomeres only. F r o m th is image(s), the fluorescence intensi ty of every telomere is determined. However, to determine w h i c h telomere belongs to w h i c h chromosome, we use a second image w h i c h i s a n image of the chromosomes wi thout the telomeres. To ob ta in the first image, a fluorescent probe (PNA sequence of (CCCTAA)3) is u s e d for the detection of the T T A G G G repeats i n the telomeres. S ince , there are no other sequences present i n the cel l w h i c h compete w i t h the b i n d i n g of th is probe, the resul tant image acquired highl ights only the telomere s ignals . Because i n general, the length of a telomere is correlated very we l l w i t h the n u m b e r of fluorescently labeled probes at tached to it, the measured fluorescence intensi ty of the telomere provide informat ion about the length of the telomere. B y combin ing mul t ip le measurements , r a n d o m var ia t ions i n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 12 f luorescence measurements are filtered out and reliable estimates of telomere length can be obtained. A different coloured fluorescence label can also be u s e d to h igh l igh t a n entire chromosome. A s a result , a second image of only chromosomes (without telomeres) i s then obtained. The telomere image can then be super imposed onto the chromosome image to help determine w h i c h telomere belongs to w h i c h chromosome (Figure 1.3). B y identifying each chromosome i n a cel l , telomere length d i s t r ibu t ion of each chromosome i n a par t icu lar cel l type m a y be obtained. Ini t ial ly, FITC a n d PI probes were used for m a r k i n g the telomere a n d chromosome respectively. Curren t ly , C Y 3 a n d DAPI probes are u s e d ins tead of FITC a n d PI probes respectively. The current ly u s e d probes are better for th is ana lys i s because i) there is less spectral overlap between the probes u sed , ii) the C Y 3 probe photobleaches less t han the FITC probe a n d iii) the D A P I probe h igh l igh ts the bands w i t h i n the chromosome w h i c h facilitates ka ryo typ ing (identification a n d differentiation of chromosome types i n the cell). 1.4. Objectives The hypothes is of this thesis is that the length of i n d i v i d u a l telomeres i n a par t i cu la r cel l type can be determined from fluorescence microscope images of the telomeres i n a l imi ted set of metaphase chromosome spreads of that ce l l type. In order to test th is hypothesis , we h a d to adapt exis t ing microscope techniques a n d develop new image analys is a lgor i thms. Because the hypothes is is based on new D N A detection methods, the cor responding t asks have not been accompl i shed before. A s a result , compar i sons w i t h exis t ing TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY Figure 1.3. Fluorescence image of telomeres and chromosomes. The pseudo colour image is generated by superimposing the red, CY3 labeled telomere image onto the blue DAPI labeled chromosome image. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 14 t echniques are not straight forward. Conven t iona l systems u s i n g Sou the rn ana lys i s can only determine the d i s t r ibu t ion of telomere lengths i n a popu la t i on of cells a n d not the telomere lengths of i n d i v i d u a l chromosomes or cel ls . The goal of th is thesis is to prove our hypothes is by developing a f luorescence microscope imaging a n d analys is system to quantify the telomere f luorescence a n d thereby the telomere lengths i n i n d i v i d u a l chromosomes . We w i l l use the b reak- th rough offered by the P N A - F I S H technology (Lansdorp et a l . , 1996) to quant i ta t ively h ighl ight i n d i v i d u a l telomeres of the ce l l . S ince the telomeres a n d their at tached fluorescence probes occupy a 3 -d imens iona l space, we s h a l l use 3-d imens iona l images i n our analys is . To achieve our goal, the tasks needed are: 1. to b u i l d the fluorescence microscope imaging a n d analys is system; 2. to develop programs for acqu i r ing images of telomeres a n d chromosomes; 3. to cal ibrate the acquired telomere images for i l l u m i n a t i o n unevenness , opt ica l aberrat ions, camera defects, a n d photobleaching effects; 4. to detect a n d segment telomeres from the acqui red 3 -d imens iona l (x,y,z) telomere images; 5. to obta in a rel iable estimate of the length of every telomere from the telomere images by measu r ing the telomere's integrated fluorescence intensi t ies (IFI) i n mul t ip le metaphase cells; 6. to detect a n d segment chromosomes from the chromosomes image; a n d 7. to associate every telomere (and its ca lcula ted telomere IFI value) w i t h the corresponding segmented chromosome i n the images a n d present the resul ts to the user for edi t ing a n d verif icat ion. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 15 Chapter 2. Background 2.1. Imaging Systems 2.1.1. Overview The appl ica t ion of mach ine v i s ion a n d robotics to microscopy began i n the early 1950s w i t h the in t roduc t ion of b lood cel l analyzers (Young a n d Roberts , 1951; Wal ton , 1952). Since then, m a n y appl ica t ions of mach ine v i s i o n to other areas of cel l biology have been developed. These inc lude appl ica t ions i n cervical cel l screening (e.g. Bengtsson et a l . , 1979; Tucke r , 1979; Shoemaker et a l . , 1982, Palc ic et a l . , 1992) and i n the analys is of chromosomes (e.g. Pres ton, 1976; Ph i l ip s a n d Lunds teen , 1985). Cur ren t ly , there are a n u m b e r of commerc ia l ly available imaging systems for general cel l analys is , s u c h as those from Bec ton D i c k i n s o n Incorporated a n d Oncor Ins t rument Systems. Cur r en t imaging systems specifically developed for the ana lys i s of chromosomes inc lude those developed by A p p l i e d Imaging Inc., B io log ica l Detect ion Systems Inc., V y s i s Inc., a n d Perceptive Scientif ic Ins t ruments Inc. A l l of these systems have the capabi l i ty to manipu la te a n d work w i t h m u l t i - spect ra l images. The appl icat ions they are designed for inc lude spot coun t ing , comparat ive genomic hybr id iza t ion and karyotyping , chromosome probe mapp ing , a n d in tens i ty a n d morphometr ic measurements . In general, a l l these systems require h u m a n interact ion to verify and correct the resul ts generated. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 16 A l t h o u g h there are m a n y capabil i t ies bu i l t into these imag ing systems, these sys tems do not suppor t new areas of chromosome research as i n the topic of th i s project. A s these systems are appl ica t ion specific, proprietary a n d are not "open systems", it i s difficult to integrate new algor i thms, modify ex is t ing a lgor i thms a n d incorporate new components into t h e m to perform our research. The source code of these programs is huge a n d difficult to obta in . E v e n w h e n the source code is made available, there i s l i t t le t echn ica l suppor t provided. T h u s , m a n y researchers resort to develop their own in -house systems for address ing new research appl ica t ion areas (e.g. Locket t et a l . 1990; P o u l i n et a l . , 1989, Nederlof et a l . , 1992). Hence , we decided to design and develop our own sys tem for telomere image acqu is i t ion a n d analysis . Th i s w o u l d give u s the ease a n d flexibil i ty to implement a lgor i thms w h i c h need to be tai lored to our specific area of research. In des igning the acquis i t ion system, a good unde r s t and ing of the properties, character is t ics , a n d tradeoffs of the diverse selection of components i s required. U s i n g th is knowledge, the appropriate components can then be selected s u c h that the integrat ion of these components are op t ima l for the in tended telomere appl ica t ion (Chapter 3). A n overview of the different components of the acquis i t ion system is descr ibed below. A l t h o u g h m a n y image cytometry (cell measurement) systems have been developed, they are a l l very s imi la r i n design a n d operat ion. The bas ic sys tem cons is t s of a n i l l u m i n a t i o n source, microscope optics, a m e c h a n i c a l stage, a camera , d ig i t iz ing c i rcu i t ry , image memory, a d i sp lay moni tor , a n d processors (Figure 2.1). The stage w h i c h holds the samples is capable of m o v i n g objects i n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 17 the x a n d y direct ions (in a plane para l le l to the detector's field of view). In some systems, a motorized z direct ion is provided for au tomated focuss ing purposes . L igh t is first t ransmit ted from the i l l u m i n a t i o n source to the sample . A n image of the sample is then acquired by the camera detector. The detected image is t ransformed into a digi ta l image by the d ig i t iz ing c i rcu i t ry . The resu l t ing digi ta l image is then stored i n computer memory from where i t c a n be d i sp layed on the moni tor a n d / o r processed a n d analyzed by the computer . T h i s technology has brought accuracy, uniformity , reproducib i l i ty , a n d a cont ro l level of qual i ty to microscope analys is . Camera Illumination Source Digitizing Circuitry Microscope Optics Microscope Stage Display Monitor Image Memory ~7K Computer/ Processors! Figure 2.1. B l o c k d iagram of a typ ica l imaging system. The major differences amongst these imaging systems are the opt ics a n d t ransducers (detector) employed a n d the method u s e d i n scann ing . The mos t TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 18 impor tan t component i n the opt ical system is the objective lens of the microscope. T h i s lens determines the magnif icat ion of the sample (the object) a n d the spat ia l resolu t ion of the image produced . These lenses are not perfect a n d generally in t roduce distort ions, aberrat ions, a n d shad ing effects. Ano the r impor tan t factor i n the opt ical sys tem i s the arrangements of the opt ics a n d components i n the microscope. The most c o m m o n optics arrangement for l o o k i n g at samples are found i n widefield microscopes . Confoca l microscopes m a y also be used , par t icu lar ly i n s i tuat ions where out-of-focus b l u r is a p rob lem. In imag ing systems s u c h as the ones descr ibed above, the qual i ty of the ou tpu t image is dependent not only on the type of optics u s e d b u t also on the t ransducer a n d electronic (digitizing) c i rcui t ry . The funct ion of the t ransducer is to convert the opt ical image to an electronic form. M o s t systems use a two d imens iona l detector s u c h as those found i n tube cameras or a two d imens iona l array charge coupled device (CCD) cameras (e.g. Tucke r , 1979; J agg i et a l . , 1987, 1988, 1990). Video cameras scan a n d sample the sensor to ob ta in a s igna l i n a n analog (video) format w h i c h then requires s a m p l i n g by a compute r to convert i t to a digi ta l form. Very few systems use l inear detectors s u c h as C C D or diode arrays (e.g. Bengtsson et a l . , 1979; J agg i et a l . 1985, 1986, Pa lc ic et a l . , 1987; Tucke r et a l . , 1987). In these systems, a digi t ized image i s obta ined by moving the sample across the sensor a n d then p iec ing together the image l ines i n the computer . The advantage of l inear scanners i s that they, i n general, have more sensor elements t h a n a single row or a single c o l u m n i n a two d imens iona l detector a n d hence provide a wider field of view. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 19 Another advantage is that the elements of the sensor are digi t ized a n d t ransmi t ted as digi ta l data. The disadvantages of s u c h a system, for the purpose of acqu i r ing h igh resolu t ion 2 -d imens iona l images, are i) the requirements for a precise mechan ica l s cann ing sys tem to move the l inear sensor across the image and ii) the increase i n t ime for acqu is i t ion . Sys tems w h i c h use a combina t ion of the l inear and mat r ix (video) detectors (e.g. G r a h a m a n d Norgen, 1980) also exist. Here, objects detected by the l inear array are moved into the field of view of the two d imens iona l detector where a h igher reso lu t ion image is acquired th rough a higher magnif ica t ion lens . There are also systems w h i c h use a one element detector s u c h as a pho tomul t ip l i e r or a photodiode (e.g. Ingram and Preston, 1970; Shoemaker et a l . , 1982). In these systems, a mi r ro r is u s e d to deflect the laser spot to scan the object a n d the cor responding s igna l from each spot is assembled i n the computer to create the image. M o s t cameras today are bu i l t for the television broadcast c o m m u n i t y where the image detected by the t ransducer is converted to a n analog s ignal . T h i s s ignal , represent ing the image, is later digit ized (at a p ixe l gr id w h i c h is generally different from that of the sensor) for mach ine analys is . A s a resul t , some informat ion is lost due to the indirect image sampl ing a n d related errors. Better noise performance cameras directly sample a n d digitize the sensor image in to d igi ta l da ta for mach ine analysis . They are however more expensive a n d do not conform to a fixed s tandard of data t r ansmiss ion . Regardless of the components u s e d i n the system, distort ions in t roduced by the optics , TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 20 t r ansducer a n d digi t iz ing c i rcu i t ry shou ld be compensated for, to help s impli fy the ana lys i s process a n d obta in more accurate resul ts . 2.1.2. Wide-field Microscopes The wide-field microscope is the bas ic in s t rumen t u s e d to obta in a magnif ied view of the sample. The bas ic components of the wide-f ield microscope are i) the i l l u m i n a t i o n source, ii) the objectives, a n d iii) the v iewing appara tus . The i l l u m i n a t i o n source suppl ies the m e d i u m (light) w h i c h i s modif ied a n d t ranspor ted from one component to the next. The source cons is t s of some l ight generation m e c h a n i s m a n d a focuss ing m e d i a to direct the l ight to the sample as either coherent or incoherent i l l umina t i on . The l ight measu red can arise from l ight w h i c h is i) t ransmit ted th rough the sample (as i n br ight- field microscopy) , ii) reflected off the sample (as i n reflectance microscopy) , or iii) absorbed by the sample a n d re-emitted at a different wavelength (as i n f luorescence microscopy, the mode u s e d i n ou r study). The t ransmi t ted , reflected or re-emitted l ight is focussed by the objectives a n d poss ib ly other lenses to produce a magnified image at the v iewing surface s u c h as the oculars a n d / o r a t ransducer . The t ransducer converts the l ight to a n electronic s igna l w h i c h c a n then be digit ized and stored i n a computer for further analys is . For the wide-field microscope, the l imi t i ng reso lu t ion depends m a i n l y on the optics of the system. The l imi t i ng resolu t ion of a sys tem defines the dis tance by w h i c h two points can be resolved. Let the xy-p lane be perpendicu la r to the direct ion of the focal axis z. It c an be s h o w n from Raleigh 's cr i ter ion that the x and y resolu t ion of the microscope is p ropor t iona l TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 21 to the wavelength of light (k) used and inversely proportional to the numerical aperture (NA) of the system. In the z direction, Francon's criterion states that the resolution is proportional to the wavelength and the index of refraction (n) of the medium and inversely proportional to the square of the numerical aperture. (Inoue, 1986). That is, dv = ^ ± (2-1) NA d z = 1 ^ ( 2 ' 2 ) NA2 where dv is the limiting resolution in the x-y plane, and dz is the limiting resolution in the z direction and is often referred to as the depth of field of microscopes. Sheppard (1988) has also characterized the depth of field of microscopes that have high numerical aperture objectives in an air medium. He has derived a number of different expressions based on diffraction optics. These include i) a divergent beam, ii) a divergent beam in (the more realistic) aplanatic systems where the beam is focussed to a plane wave, and iii) a paraxial approximation to the aplanatic system. These expressions result in depth of field values which are closed to Francon's criterion (stated above) for low numerical aperture (< 0.6) objectives and is 50 to 75% of Francon's criterion for numerical aperture approaching 1.0. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 22 U s i n g equat ions 2-1 a n d 2-2, the l imi t of reso lu t ion i n the x (or y) a n d z plane for a wavelength of 6 0 0 n m and for a n u m e r i c a l aperture of 1.4 w i t h a n o i l m e d i u m index of refraction of 1.5 (conditions w h i c h are u s e d i n our research), are 0.26 a n d 0.46 mic rons respectively. The telomeres a n d chromosomes i n the samples we u s e d generally lie at different focal p lanes s p a n n i n g a depth w h i c h can be th icker t h a n the resolu t ion l imi t i n the z di rec t ion. A s a result , out-of-focus b l u r from planes above and below the focal p lane are present i n the acqui red image. T h u s , more than one image plane (as s h o w n later i n Chapte r 5) are required to obtained a more accurate representat ion of the telomere length. Sample Image M s(x,y,z) Microscope Characteristics h(x,y,z) Observed ^ Image o(x,y,z) Figure 2.2. M o d e l of a microscope system. The image of the sample is modif ied by the microscope to resul t i n the observed image. We need to characterize the system i n order to unde r s t and its behaviour . The general character izat ion is given below, whi le the details specific to our sys tem w i l l be elaborated u p o n i n Chapter 4. The image formed at the t ransducer c a n be characterized u s i n g a s implif ied model of the microscope TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 23 (Figure 2.2). In th is model , the character is t ics of the l ight source, lenses, a n d t r ansducer are combined into a "black box" w i t h the sample image as the sys tem i n p u t a n d the observed image as the system output . The ana lys i s of s u c h sys tem is carr ied i n the three spat ia l d imens ions to cor respond to the three d imens iona l nature of objects. T h u s , i n the ma themat ica l formula t ion , the sample is descr ibed as s(x,y,z). Th i s sample image i s modif ied by the objectives a n d other lenses of the system a n d its character is t ics can be represented by a transfer funct ion i n the form of a three d imens iona l point spread funct ion (PSF), h(x,y,z). The observed image, however, is a two d imens iona l one, o(x,y). Albei t , a three d imens iona l image, o(x,y,z) can be obta ined by acqu i r ing and s tor ing several xy-plane images at va ry ing focal levels (z-direction of the microscope). The resu l t ing three d imens iona l image c a n be descr ibed as the 3-d imens iona l convolu t ion of the sys tem P S F w i t h the sample , i.e.: where ® denotes the convolu t ion operator. In the Four ie r doma in , the observed image spect rum, 0(u,v,w), is the mul t ip l i ca t ion of the microscopes opt ica l (contrast) transfer funct ion (OTF), H(u,v,w), w i t h the image spec t rum of the sample, S(u,v,w). o(x,y,z) = h{x,y,z)®s{x,y,z) (2-3) 0(u,v,w) = U(u,v,w) • S(u,v,u>) (2-4) The O T F of the microscope can be determined theoretically. H o p k i n s (1955) has derived the O T F funct ion for a general opt ica l sys tem w i t h a c i r cu la r TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 24 aperture a n d incoherent i l l umina t ion . Later, S tokse th (1969) has in t roduced a n approx imat ion to the equat ion for large amounts of defocus. In addi t ion , G o o d m a n (1968), Cas t l eman (1979), a n d others have given a general descr ip t ion on th i s topic a n d defined the funct ion for different shape apertures. M a n y a s sumpt ions of ideal i ty are u s e d i n the derivations, bu t these do not h o l d for h i g h n u m e r i c a l aperture objectives (e.g. the approx imat ion of sinB w i t h 9 w h e n 9 i s large). For image reconst ruct ion, some researchers use the exper imenta l ly determined three d imens iona l O T F (Agard et a l . 1989), whi le others use the theoret ical O T F (Erhardt et a l . , 1985). One exper imenta l me thod to determine the O T F is to use sma l l fluorescent beads as point sources of l ight a n d apply Four ier t ransforms to the observed va ry ing focal depth images (Hiraoka et a l . 1990). 2.1.3. Confocal Microscopes The confocal microscope is a n alternate choice of microscope for c ap tu r ing fluorescence images u s e d i n ou r s tudy. It c a n p roduce h igher reso lu t ion images t h a n the wide-field microscopes bu t has other d rawbacks . The confocal microscope is or iginal ly proposed by M i n s k y (1957). The p r inc ip le of i ts operat ion (Figure 2.3) is that the i l l u m i n a t i o n source is focused to a single poin t on the sample (the i l l u m i n a t i o n pinhole) a n d the l ight from th i s po in t i s focussed a n d detected by the t ransducer th rough the detector p inhole . O n l y a por t ion of the informat ion (mostly in-focus) at the i l l umina t ed poin t is a l lowed to pass th rough the p inhole whi le most of the out-of-focus in format ion is b locked . Consequent ly , the resu l t ing image is of h i g h la tera l a n d ax ia l TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 25 reso lu t ion . A three d imens iona l image can be generated by m o v i n g the i l l u m i n a t i o n a n d detection p inholes u s i n g mir rors a n d / o r m o v i n g the sample u s i n g the microscope stage to scan the object. The speed i n w h i c h a n image is acqui red is governed by how fast the scann ing takes place a n d h o w long a spot m u s t be i l l u m i n a t e d i n order to generate enough photons for detection. T h u s a n intense i l l u m i n a t i o n source s u c h as a laser or mercury arc l a m p is often used . illumination condensor object objective detector pinhole lens lens pinhole Figure 2.3. Pr inc ip le of confocal microscopy. A point source of l ight i s u s e d to i l l umina te the sample. The out-of-focus image is b locked a n d only the in- focus image i s observed th rough the detector p inhole . In confocal microscopes, it i s c r i t ica l that the i l l u m i n a t i o n a n d detection l ight pa ths behave s imi la r ly s u c h that a focused image is seen at the detector p inhole . There can be var ia t ions i n lenses as one lens i s l i ke ly to behave differently t h a n another. To overcome these var ia t ions , the objective a n d TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 26 condenser lenses are generally the same lens. In th is ins tance (and also i n wide-f ield fluorescence microscopy), a dichroic mi r ro r i s used to deflect the i l l u m i n a t i o n to the sample and the emitted l ight i s passed th rough to the detector. In biological work, the confocal microscope i s u s u a l l y u s e d for fluorescence imaging where l ight at a par t icu la r wavelength is used to excite the sample . The components w h i c h are excited by the l ight w i l l fluoresce (emit) l ight at a different wavelength. The spectral selection of specific components i s pa r t i cu la r ly useful for extract ing and identifying the objects of interest. A s i n the case of wide-field microscope, the P S F or O T F of the confocal microscope c a n also be characterized. The character is t ics of the confocal microscope has been descr ibed by (Inoue, 1986; P lu ta , 1988; v a n der Voor t et a l . , 1988). In the ideal case where the pinhole i s infini tely s m a l l , the sys tem P S F c a n be descr ibed by the square of the P S F of the wide-field microscope (Wilson a n d Sheppard , 1988) and is given by o{x,y,z) = h2{x,y,z)® s(x,y,z) > (2-5) A s a resul t of the squar ing funct ion, the P S F of the confocal sys tem i s m u c h sharper t h a n that of the wide-field microscope a n d thus is of higher spa t ia l a n d ax ia l resolut ion. In practice, however, the confocal microscope i s rarely u s e d i n i ts idea l state. General ly, the pinhole has a cer ta in diameter a n d i s adjustable to compensate for the faint l ight reflected or emitted from the sample . Hence the P S F or OTF of the system s h o u l d be exper imenta l ly determined for each given setup. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 27 Because of the time i t takes to scan a single point over a n area of the sample , faster confocal microscopes are desired. Th i s has been real ized u s i n g i) a t andem scann ing microscope (Inoue, 1989), ii) the l inear s c a n n i n g microscope (Wilson and Hewlett, 1990), or acousto-optic s cann ing microscope (Goldstein et a l . , 1990). In the tandem scann ing microscope, a large n u m b e r of spa t ia l ly separated pinholes on a d i s k (Nipkow disk) are used s u c h that more t h a n a single point c a n be imaged wi thout interference from ne ighbour ing poin ts at a given t ime. A s the d i sk i s rotated, a different set of po in t s are imaged a n d hence the entire image can be composed at a faster rate. In the l inear s cann ing microscope, a sl i t shaped i l l u m i n a t i o n a n d the cor responding l ine ra ther t h a n a point detector is used . This resul ts i n h igh reso lu t ion i n one di rec t ion (e.g. x) s imi la r to that of confocal and lower reso lu t ion i n the others (e.g. y a n d z) compared to that of confocal. The resolu t ion of th is system i s s t i l l better t h a n that of the wide-field microscope. The reason w h y we chose the wide-field microscope i n th is s tudy i s because confocal microscopes often have a more intense i l l u m i n a t i o n source w h i c h c a n cause higher sample photobleaching. In addi t ion , the acqu i s i t ion t ime i n confocal microscopes is s ignif icantly increased i f h igh spa t ia l reso lu t ion i s to be main ta ined . L o n g exposure t imes, up to 10s, are typ ica l ly requi red to capture the weak telomere s ignals . To reduce the acqu is i t ion t ime i n confocal microscopes , a sub-region of the image i s typical ly used for object focuss ing before the entire image is captured. A s a result , th is sub-region w i l l exhibi t h igher photobleaching effects because it i s exposed to more l ight d u r i n g the focuss ing process than the rest of the image. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 2.2. Image Analysis 28 2.2.1. Overview A typ ica l process for ana lyz ing cells a n d chromosomes involves the fol lowing s ix steps: i) acquis i t ion of images, ii) pre-process ing the acqu i red images, iii) segmentat ion of the cells i n the scene, iv) pos t -process ing the segmented regions, v) extract ion and quant if icat ion of features, a n d vi) c lass i f ica t ion of the segmented objects (Liedtke et a l . , 1987; Poon et a l . , 1989a, 1992c, 1993a) (Figure 2.4). The first two steps, acqu i r ing a n d pre-process ing the images, are c r i t i ca l since h igh qual i ty i npu t images do simplify a n d reduce the amoun t of process ing required i n the later stages of the analys is . Qua l i t y of the sample preparat ion a n d acquis i t ion system is thus very impor tant . The next step segments or defines the regions of interest i n the image, s u c h as the objects from the background . Post-processing (either au tomat ica l ly or interactively) of the defined regions is required to fine-tune the m a s k of each region. Features are then calcula ted based on the defined region boundar ies . The objects i n the scene are then classif ied based on the va lues of the features a n d / o r the segmented resul ts . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 29 Acquisition Pre-processing Segmentation Post-processing Feature Extraction Object Classification Figure 2 .4. Process for cel l analysis . In our analys is , we perform a s imi la r process as descr ibed above. However, the specific algori thms a n d methods u s e d are ta i lored for our telomere a n d chromosome segmentation and the extract ion of the IFI value . These a lgor i thms w i l l be d i scussed i n Chapters 5 a n d 6. A general descr ip t ion of some bas ic pre-processing and segmentation algori thms is given below. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 30 2.2.2. Image Pre-Processing Image acquis i t ion and pre-processing are key steps i n the ana lys i s process for ob ta in ing consistent a n d repeatable resul ts . Images of the same sample m a y be different depending on i) the consis tency i n prepar ing the sample w i t h i n the same ba tch and amongst batches, ii) the loca t ion of the object on the sl ide, iii) the locat ion of the object i n the microscope 's field of v iew (x, y, a n d focus), a n d iv) the t ime the image is acquired ( i l lumina t ion stabil i ty, aging a n d photobleaching effects). Sample preparat ion differences c a n be compensa ted for by ca l ib ra t ing the cells of interest to s imi l a r s ta ined objects /ce l ls w i t h k n o w n character is t ics on the same sl ide (Palcic et a l . , 1992). Pre-process ing techniques have been developed over the years, to correct for differences i n i l l u m i n a t i o n , sensor, a n d optics aberrat ions at different x , y locat ions (Poul in et a l . , 1994). Amongs t these techniques are b a c k g r o u n d sub t rac t ion methods (Cast leman, 1979), flat-field compensa t ion methods (Poul in et a l . , 1994), wavelength compensat ion methods (Cast leman, 1993), a n d photob leach ing methods (Rigaut et a l . 1990, 1991). These methods are briefly descr ibed below. Some of these methods or var ia t ions of t hem are u s e d as appl icable i n our a lgor i thm (Chapter 3). 2.2.2.1. Background Subtraction One method of correct ing for the shad ing a n d aberrat ion effects is to perform b a c k g r o u n d subtrac t ion. In this method, the loca l difference i n the field of view (represented by the bright backg round image, B(x,y)) i s subt rac ted from the acqui red image, I(x,y), resu l t ing i n the corrected image, C(x,y): TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY C(x,y) = l{x,y)-B(x,y) + K 31 (2-6) A cons tant value, K , equal to the average value of the br ight image is added s u c h that the range of grey levels i n the corrected image is s imi la r to that of the acqui red image. A s no mul t ip l i ca t ions or d iv is ions are u s e d i n th is method , t r unca t ion errors are m i n i m a l . 2.2.2.2. Flat-field Compensation A more accurate method t h a n the background sub t rac t ion correct ion is to use flat-field compensat ion . Th i s compensat ion attempts to scale the acqui red image depending on the local br ight backg round image. T h i s me thod is based on the convers ion of the l ight t ransmit tance (which is detected by the sensor) to opt ica l densi ty values by t ak ing the logar i thms of the p ixe l values . The opt ica l densi ty values can then be added a n d subtrac ted to s imula te the p h y s i c a l properties of the system. The convers ion operat ion is determined as follows: T h i s equat ion c a n be converted to the following flat-field compensa t ion equat ion by removing the logar i thms and subt rac t ing a dark b a c k g r o u n d image, D(x,y), from the measured images (i.e. I(x,y) = F(x,y) - D(x,y) a n d B(x ,y) \ogC{x,y) = logl{x,y) - logB(x,z/) + log(k) (2-7) = B'(x,y) + D(x,y) : C(x, i / ) = k V{x,y)-D{x,y) B'{x,y)-D{x,y) (2-8) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 32 A s d iv i s ion is u s e d i n this technique, t runca t ion errors are l ike ly to dis tor t the d i s t r ibu t ion of (discretely binned) grey levels. 2.2.2.3. Wavelength C o m p e n s a t i o n M u l t i p l e probes are often u s e d i n microscopy imaging. Th i s generally requires acqu i r ing images at different wavelengths to h ighl igh t different features of the cel l or t issue. Depending on the probe u s e d a n d their spectra l character is t ics , there can be an overlap i n the spectral properties of the probes used . C a s t l e m a n (1993) has developed a method of compensa t ing for the overlap effects a n d iso la t ing the signals from each probe u s i n g mat r ix algebra. In th is method , the spectral response of each probe at each of the observed wavelengths can be represented by a vector i n mat r ix C. For example, probe X m a y have normal i zed responses of 0.1, 0.7, a n d 0.2 at wavelengths Xlf X2, a n d X3, respectively. The response of each probe (vector R) c a n be ca lcu la ted from the observed spectral images (vector I) as follows: R = I C 1 (2-9) 2.2.2.4. Photobleaching C o m p e n s a t i o n . Photobleaching of the sample, par t icu la r ly i n acqu i r ing fluorescence images, c a n have a considerable effect on the in tens i ty of images t aken over t ime. E a c h t ime the sample is i l lumina ted , the amoun t of f luorophores w h i c h c a n be excited i s reduced resu l t ing i n a less intense emiss ion . Hence , the in tens i ty of the image may vary depending on the t ime required to focus the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 33 image before acquis i t ion takes place a n d / o r the t ime the image is acqui red i n the 3 D (multiple focus plane) image acquis i t ion cycle. An t i -pho tob leach ing agents are generally be ing used i n sample preparat ion to reduce the effects of photobleaching . However, there is u s u a l l y some reminance photob leach ing effect (of approximate ly a few percent) over minu tes of exposure. Rigaut et a l . (1990, 1991) have proposed methods to apply recons t ruc t ion techniques to correct images i n confocal microscopy for effects of optics a n d photobleaching . These methods typical ly use an exponent ia l t ime decay funct ion to s imula te the photobleaching effects. 2.2.3. Focus and Three Dimensional Reconstruction General ly , it i s a s sumed that the image taken conta ins objects w h i c h are a l l in-focus or near focus. Th i s a s sumpt ion is not necessary true. A s we have s h o w n earlier, images of sufficient detail and clar i ty are required i n quant i ta t ive mic roscopy to obta in consis tency i n object c lassif icat ion a n d d i s c r imina t i on (Poon et a l , 1987, 1989b, Spadinger et a l . , 1989, 1990, Poon a n d Palc ic , 1991). To see the details, these images are often taken w i t h h igher magnif ica t ion a n d n u m e r i c a l aperture objective lenses. These lenses have a lower depth of focus a n d as a result , the details of the entire object can not be captured i n one focus plane. In addi t ion , objects i n a scene a n d even the details i n the objects themselves are not a l l at the same focal plane. Therefore, we have s h o w n that these objects m u s t be ind iv idua l ly focussed before they are analyzed (Poon et a l . , 1989b, 1991, 1992a,b,c). For correct a n d consis tent object segmentat ion a n d feature ca lcu la t ion , an objective method i n focuss ing a l l objects of interest TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 34 i n the sample is necessary. In some s i tuat ions , 3 D (x,y a n d z-focus) image recons t ruc t ion i s in i t i a l ly performed. Several methods have been developed to solve the three d imens iona l p rob lem of recons t ruc t ing the in-focus image from the observed images w h i c h conta ins bo th in-focus and out-of-focus informat ion. Cas t l eman (1979) have proposed methods of u s i n g i) inverse filtering, ii) s imul taneous equat ions, or iii) nearest ne ighbour approximat ion . Agard 's group (1984, 1989, 1990) later modif ied some of Cast leman 's work a n d in t roduced the so lu t ion of iterative cons t ra ined deconvolut ion w i t h a non-negativi ty constraint . Carr ington ' s group (1987, 1989, 1995) analyzed the inverse a n d iterative cons t ra ined deconvolu t ion techniques a n d proposed the cons t ra ined least squares technique as a so lu t ion . Holmes ' group (1989a, 1989b, 1991) m a x i m i z e d the log- l ike l ihood funct ion of the system wi th respect to the object's opt ica l densi ty for object recons t ruc t ion . D u e to l imi ta t ions i n the detector a n d optics , h igher reso lu t ion images are desired for detailed image analysis . Bertero et a l . (1987, 1989, 1990) a n d Sheppa rd (1988) proposed methods for obta in ing "super" reso lu t ion from images t aken from a confocal microscope. In th is system, a two d imens iona l C C D detector is u s e d ins tead of a single element photomul t ip l ie r to detect the image created by a single point of i l l umina t i on . A s the point i l l u m i n a t i o n i s s canned over the image, a 2 -d imens iona l image response rather t h a n a single va lue i s obta ined for each point i l l umina t i on . T h i s extra informat ion at each p ixe l is u s e d to help reconstruct a n image w i t h higher resolu t ion . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 35 2.2.4. Image Segmentation 2.2.4.1. Overview The most difficult step i n au tomat ing the analys is process is to define the regions w h i c h belong to the object. U n l i k e feature extract ion where only ma themat i ca l computa t ion over the defined region w i l l suffice, segmentat ion also requires pr ior knowledge of the geometrical, morphologica l , a n d topological propert ies of the objects i n the scene as we l l as a heur i s t i c approach for ana lyz ing the problem. Since object classif icat ion is based on feature va lues w h i c h are derived from the segmented regions, segmentat ion is c r u c i a l for the correct interpretat ion of the objects i n the scene. M a n y segmentat ion techniques have been developed over the past several decades (e.g. Davis , 1975; F u a n d M u i , 1981, Man taz , 1987, M a c A u l a y et a l . , 1988). These methods can be categorized into three different classes: i) character is t ic feature threshold ing or c lus ter ing, ii) edge detection, a n d iii) region extract ion. A single a lgor i thm generally cannot segment a pa r t i cu la r scene a n d hence a combina t ion of segmentat ion processes is often used . Genera l ly these processes perform wel l i n some appl ica t ions bu t m a y fai l i n others. Deta i led descript ions of the different classes of k n o w n segmentat ion a lgor i thms are d i scussed i n the following sections. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 36 2.2.4.2. Thresholding or Clustering Thresho ld ing i s a c o m m o n technique u s e d i n segmenting regions i n a scene. The process assigns dis t inct labels to areas based on some propert ies of the image. A property may be a character is t ic feature s u c h as the image grey levels or may be of loca l nature s u c h as the gradient or L a p l a c i a n of the grey levels. In a l l cases, a specified range of values of a given property is u s e d to define the p ixe ls i n the image w h i c h belong to the same region. Often, a h i s togram of a n image property is u s e d to determine the thresholds for each region. These h is tograms are generally smoothed to remove noise. Care m u s t be taken , however, to avoid smooth ing out sma l l bu t va l id m i n i m a or m a x i m a . A th resho ld ing technique w h i c h can be appl ied to grey level h i s tograms is the mode method. Th i s type of h i s togram gives a n ind ica t ion of the n u m b e r of p ixels w h i c h have the same grey level i n the image. E a c h peak (mode) of the h i s togram represents areas where large n u m b e r of pixels have a s imi l a r in tens i ty level. A bounda ry is then placed at the val ley between peaks to separate the regions. The rat ionale for choos ing s u c h points is to min imize the probabi l i ty of misc lass i fy ing each region. S ince the n u m b e r of pixels at the val ley compared to the peaks is relatively smal l , misp lacement of the th resho ld from the exact locat ion has relatively little noticeable effect on the resu l t ing image. For example, Poon et a l . (1992c, 1993a) u s e d th is technique to segment the b lood cells from the background of the image. A different technique is u s e d for th reshold ing gradient h is tograms. S ince these his tograms represent the s u m of the magni tude of gradients at a TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 37 given grey level, the boundary is p laced at the highest point i n the h i s togram. T h i s point signifies the locat ion of the largest in tens i ty differences (the edges) of the image. Th i s me thod works wel l w i t h some images bu t fails i n others. For images where there are m a n y s imi la r in tensi ty pixels w i t h a s m a l l gradient, their s u m may overmask the s u m generated at the edge of rare objects a n d hence a wrong th reshold level is generated. C l u s t e r i n g extends the technique of th reshold ing to the m u l t i - d imens iona l space. T h i s technique is u s e d w h e n a single feature y ie lds poor d i s c r imina t i on regions whi le dis t inct regions can be seen i n h is tograms of two or more character is t ic features (e.g. c luster plots). A n y feature w h i c h is usefu l for segmenting a region, s u c h as the grey levels of images seen t h rough different spectra l filters, gradients, texture features, etc., c an be used . Poon et a l . (1992c, 1993a) u s e d the green and b lue image components of the image to separate the nuclea ted cells from the red blood cells. A lgor i thms for c lus ter ana lys i s have been available for locat ing the dec is ion bounda ry between regions i n a mu l t i -d imens iona l space (Amadasun a n d K i n g , 1988; U m e s h , 1988). To reduce the amount of computa t ions required i n the ana lys is , the smal les t n u m b e r of features w h i c h can d iscr iminate the regions is employed. Thresho ld ing a n d c lus te r ing techniques are global operators w h i c h use some aggregate properties of different features. These features are very dependent on the type of regions w h i c h are segmented i n the image. A l t h o u g h the segmented regions are closed, some images may require smooth ing to e l iminate the no isy boundar ies . S ince no spat ia l informat ion i s u s e d i n the select ion of the threshold , the resu l t ing regions may not be cont iguous . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 38 2.2.4.3. Edge Detection Edge detection algori thms use the informat ion of edge poin ts to determine the bounda ry between objects. The edge po in t s are located where there is a n abrupt change i n grey levels i n the image. In th i s technique , the elements w h i c h are candidates to belong to a n edge are first extracted a n d then combined to form the boundary . The extract ion of edge pixels requires a measure w h i c h corresponds to the change i n grey value of the p ixe l w i t h its su r round ing . V a r i o u s methods , s u c h as the gradient, Sobel , K i r s c h , a n d Prewitt operators (Rosenfeld a n d K a k , 1982; Y o u n g a n d F u , 1986), have been developed for th i s purpose . These operators c a n be implemented as a series of image convolu t ions where the weights i n the convolu t ion kerne l are different for each filter. The resu l t ing va lue of the convolu t ion at a p ixe l gives an ind ica t ion of the s t rength of the changes a r o u n d the p ixe l . The edge points are then extracted by th resho ld ing the processed image. M a r r (1982) has developed a Lap l ac i an of a G a u s s i a n edge filter. In th is method, the zero-crossings of the filter cor respond to the edges of the s t ructures w h i c h have a space constant greater t h a n (or a lower spa t ia l frequency than) a selected value used i n the G a u s s i a n b l u r r i n g process. C a n n y (1983, 1986) developed a good edge detector w h i c h convolves the no isy image w i t h a spat ia l funct ion (representative of the result) a n d then f inds the m a x i m a values i n the resu l t ing convolu t ion . There are several problems w i t h edge detection techniques . F i r s t , the t rans i t ion from one region of the image to the other sometimes occurs over TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 39 several p ixels a n d is then not abrupt enough. Second, the contours p roduced from th resho ld ing edge informat ion are generally more t h a n one p ixe l wide a n d not necessar i ly closed. Hence, some post-processing u s i n g t h i n n i n g a n d contour -c los ing algori thms are required. Another p rob lem is that the texture of some regions is significant enough to be thresholded a n d interpreted as edge points , r e su l t ing i n erroneous image segmentation. Nevertheless, the resul ts from the edge detection techniques can be u s e d i n conjunc t ion w i t h other methods i n de termining par t icu lar regions. 2.2.4.4. Region Extraction Another segmentat ion approach is to group pixels w i t h s imi la r propert ies, s u c h as grey levels, texture, color informat ion, etc., into regions. These region extract ion techniques can be separated into three categories: region merging, region spl i t t ing, and a combina t ion of region merg ing a n d sp l i t t ing (Ohlander et a l . 1978; Garbay et a l . , 1986). In region merging or growing techniques, the image is in i t i a l ly d iv ided in to m a n y s m a l l regions s u c h as a p ixe l or a s m a l l ne ighbourhood of pixels . V a r i o u s properties that reflect the character is t ics of the object are computed for each region. The character is t ics of each region are compared w i t h i ts ne ighbour ing regions. If the properties of the adjacent regions are s imi la r , these regions are combined or merged into one. T h i s process is i terated by r ecomput ing the object membersh ip properties for each enlarged region a n d merg ing the regions w h i c h have s imi la r character is t ics . The segmentat ion is TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 40 completed w h e n a l l adjacent regions have significantly different propert ies s u c h that no merge c a n further be made. The region spl i t t ing or d iv id ing techniques begin w i t h the entire image ins t ead of m a n y s m a l l regions. A predicate descr ib ing the va r ious propert ies of the object i s evaluated from the entire region. A n example i s to determine i f a l l p ixels i n the region have grey levels w h i c h do not differ by a cer ta in amount . If the predicate is not satisfied, the region is d ivided into smal ler regions a n d the predicate for each of the sub-regions is recomputed. The process cont inues u n t i l the predicates for a l l regions are satisfied. The spl i t a n d merge technique uses a combina t ion of region merg ing a n d sp l i t t ing to obta in regions of s imi la r properties. Regions are merged w h e n adjacent regions have s imi la r properties a n d are spli t w h e n the predicate desc r ib ing the property is not satisfied. Liedtke et a l . (1987) u s e d th is technique on microscope images of b lood cells to extract the pr imi t ives u s e d i n h i s segmentat ion method. Region extract ion techniques ut i l ize the loca l properties of the image directly. A l t h o u g h they produce closed and cont iguous regions, the d rawback is that these a lgor i thms are computa t ional ly intensive. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 41 Chapter 3. Imaging System 3.1. Overview T h i s chapter describes i n detai l the imaging system that we developed for th is project. A s ment ioned earlier i n Chapter 2, there is no commerc ia l ly avai lable sys tem that is capable of performing the required task, nor i s there any sys tem where modif icat ions can easily be made. Hence, we bu i l t the imag ing system for th is project by selecting a n d integrat ing bas ic commerc ia l components a n d developing the a lgori thms and software (as descr ibed later i n th i s chapter). Th i s imaging system performs two bas ic funct ions. F i rs t , it acquires images a n d stores them into files. Second, it analyzes the acqu i red images a n d generates telomere length informat ion for each detected chromosome. The first funct ion requires developing image acqu i s i t ion hardware a n d software whi le the second funct ion p r imar i ly involves developing the ana lys i s software. Rather t h a n combin ing both funct ions into one p rogram a n d one system (as i s done i n most commerc ia l systems), we developed separate software programs for the acquis i t ion a n d for the ana lys is . The acqu i s i t ion software c a n only operate w i th the hardware of the acqu i s i t i on system. The ana lys i s software, on the other h a n d , do not depend on the dedicated acqu is i t ion hardware a n d a darkened room to operate. S ince , the ana lys i s systems are less costly to b u i l d and operate t h a n the acqu i s i t ion system, the separat ion into two software programs resul ts i n a more TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 42 economica l a n d efficient use of our hardware resources. G i v e n the h igh d e m a n d for the use of the systems for biological s tudies , we bu i l t one acqu i s i t i on system and mul t ip le analys is systems a n d operated them independent ly . After the images are acquired, the images are transferred from the acqu i s i t ion system to the analys is system v i a a computer network. The te lomere /chromosome image analys is c an then be performed i n the ana lys i s sys tem. T h i s chapter first describes the components of our acqu is i t ion system. It then out l ines the algori thms we developed for acqu i r ing the mul t i - focus p lane images. Las t ly , th is chapter describes our analys is system a n d the a lgor i thms we developed for pre-processing the acquired images. 3.2. Image Acquisition Hardware 3.2.1. Overview A b l o c k d iagram of our image acquis i t ion system i s s h o w n i n Figure 3 .1 . The major components of this system are i) the microscope, ii) the focuss ing m e c h a n i s m , iii) the camera, and iv) the comput ing system. The motor ized focuss ing m e c h a n i s m varies the focus of the objects p laced on the microscope . T h i s a l lows the acquis i t ion of a series of 2 -d imens iona l images of the object where the focus pos i t ion of each image wi th respect to that of another i n the series i s computer control led and k n o w n . A 3-d imens iona l image representat ion of the object i s t hus obtained. The camera t ransforms the image i n the microscope into digi ta l form and this image c a n be d i sp layed i n the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 43 compu t ing system. The comput ing system controls the entire process a n d stores the mul t i - focus plane images for later processing. Camera CCD Sensor] Digitizer luminatioiii j Microscope! Source \~j Optics ! j : ! X | Microscope! Microscope ! Stage j Image ! ! Display ! Memory \~ j 1 Monitor j Computer / Processor Computing System Focussing Mechanism Figure 3 .1 . B l o c k d iagram of the imaging system. The major components are the microscope, focussing mechan i sm, camera , a n d the compu t ing system. 3.2.2. Fluorescence Microscope The fluorescence microscope is the key component of the imag ing system. It t ransforms and magnifies the telomeres a n d chromosomes for v i sua l i za t ion . We chose the widefield microscope over the confocal microscope for th is app l ica t ion to avoid photobleaching (fading of the fluorescence probe TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 44 caused by the i l l u m i n a t i o n source and the long acquis i t ion time) as we l l as cost considera t ions . Fluorescence microscopes from brand-name manufac turers (e.g. Lei tz , N i k o n , O l y m p u s , and Zeiss) a l l have comparable performance. We u s e d the Zeiss A x i o p l a n fluorescence microscope for th is project because of i ts ava i lab i l i ty i n the laboratory. The major components of the fluorescence microscope are i) the i l l umina t i on source, ii) the exci tat ion a n d emiss ion filters, iii) the objectives, and iv) the microscope stage. A n impor tan t res t r ic t ion i n the select ion process of each of these components is that the component m u s t be compat ib le for use w i t h our chosen Zeiss microscope. Before d i s cus s ing the selection of components , the bas ic operat ion of the fluorescence microscope is descr ibed. A fluorescence microscope general ly has a n u m b e r of slots for filters and d ichroic mi r ro r b lock assemblies . F i rs t , a selected wavelength of l ight from the i l l umina t i on source is passed th rough a n exci ta t ion filter. The selected l ight is reflected off the d ichro ic mi r ro r to the sample v i a the objective lens. The sites w i t h i n the sample w h i c h are excited by the selected wavelength of l ight w i l l emit l ight at a higher wavelength (lower energy). The emitted l ight s ignals are focussed and magnif ied by the objective lens of the microscope. The emiss ion filter then a l lows only a selected wavelength of the emitted l ight from the sample to pass th rough a n d b l o c k s other wavelengths of the l ight i n c l u d i n g the reflected l ight at the exci ta t ion wavelength. The resu l t ing image of the selected sites are then v i sua l i zed th rough the d ichro ic mi r ro r a n d onto the oculars of the microscope or onto a camera for d i sp lay on a monitor . Thus , the choice of exci tat ion, d ichro ic , a n d TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 45 emiss ion filters are impor tant for determining w h i c h fluorescence labeled probe i s desired to be seen i n the image of the object. The objective lens a n d i l l u m i n a t i o n source are also impor tant i n quanti tat ive imag ing because they govern the image magnif icat ion, distort ions, s tabil i ty, a n d intensi ty. 3.2.2.1. Illumination Source The choice of the i l l umina t i on source i s dependent on the fluorescence spect ra l character is t ics of the sample. A n ideal i l l u m i n a t i o n source i s one w h i c h i) gives even (uniformly distributed) i l l u m i n a t i o n , ii) has sufficient in tensi ty i n the desired excitat ion wavelength, and iii) does not f luctuate over t ime. B y careful adjustment and al ignment of the i l l u m i n a t i o n source for Koehler i l l u m i n a t i o n , a fairly even i l l umina t i on a round the center field of v iew (5% variation) c a n be obtained. Typica l ly , fluorescence mic roscopy ft i l l u m i n a t i o n sources are based on either mercury or xenon . M e r c u r y (200W, Zeiss) tends to have a n u m b e r of intensi ty peaks i n c l u d i n g the wavelengths at 4 0 5 n m a n d 5 4 6 n m . These two intensi ty peaks can be used to excite the D A P I a n d C Y 3 probes used i n th is project, respectively. X e n o n (150W, Zeiss) tends to be less intense at these wavelengths bu t has better tempora l s tabi l i ty t h a n that of mercury . In addi t ion , the lifetime of xenon b u l b s is longer t h a n that of mercu ry (500 hour s compared to 200 hours) . However, the x e n o n l a m p i s not sui table for th is project because it does not give sufficient in tensi ty for C Y 3 exci ta t ion. Fortunately, a h y b r i d m e r c u r y / x e n o n l amp (200W, O p t i Q u i p d i s t r ibu ted by Zeiss) is commerc ia l ly available. Th i s h y b r i d l amp fluctuates less t h a n the mercury l amp i n time and is more intense t h a n the xenon l amp at TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 46 the wavelengths of interest. Its spectral character is t ics i s very s imi l a r to that of the mercu ry l amp but has a m u c h longer bu lb lifetime (2000 hours) . Hence, we chose th i s l amp for this project. The resul ts and ana lys i s of the t empora l s tabi l i ty of the h y b r i d l amp used are d i scussed later i n th is Chapter . 3.2.2.2. E x c i t a t i o n a n d E m i s s i o n Fi l ters A s ment ioned earlier, the choice of the exci tat ion a n d emiss ion filters a n d the d ichro ic mi r ro r p lays a n impor tant role i n wha t i s seen i n the resu l t ing image. Fi l ters a n d d ichroic b lock assemblies are then selected to m a t c h the propert ies of the fluorescence probes used i n the experiment. Fo r example, one probe highl ights the entire object whi le other probes, w i t h different fluorescence spectra l characterist ics , h ighl ight specific sites w i t h i n the object. A s the b lock assemblies are interchanged throughout the experiment, the a l ignment of these b locks w i th each other and w i t h its previous pos i t ion i n the imag ing pa th m a y vary. Consequent ly, the p rob lem of image regis t ra t ion of mul t ip le probe images resul ts (up to 10 pixels shift). For tunate ly , new types of filters have been developed i n the las t five years to min imize the image registrat ion problem. We u s e d one type of these filters i n th is project. A s i n convent ional systems, the exci ta t ion filter i n th is filter sys tem selects a n d al lows a specific wavelength of l ight to pass th rough to the object. Instead of a l lowing only one wavelength to pass , . the emis s ion filter we u s e d a l lows mul t ip le bands of wavelengths of l ight to pass . Hence, w i t h mul t ip l e exci ta t ion filters and only one emiss ion filter, objects labeled w i t h mul t ip le fluorescence probes can be independent ly imaged. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 47 The spectra l character is t ics of the filters used i n th is project are s h o w n i n Figure 3.2. In our filter system, the mechan ica l selection of exci ta t ion is performed only i n the i l l umina t i on pa th u s i n g a filter wheel (Pacific Scientif ic Inc.) w h i c h has openings for 8 different exci tat ion filters. A single m u l t i - spec t rum dichro ic mi r ro r and emiss ion filter assembly is then used to image a l l probes i n the experiment. The excitat ion and emiss ion filters are selected i n conjunc t ion w i t h the appropriately matched probes to min imize spect ra l c ross ta lk i n the observed image. Since no opt ical components are moved i n the imag ing pa th , the shift between mul t i - spec t rum images is s m a l l a n d i s less t h a n 2 pixels . There are 2 d rawbacks i n u s i n g this type of filter system w h i c h c a n be compensated for w i th adequate intensi ty i l l u m i n a t i o n a n d appropriate select ion of filters a n d probes. Firs t , the amount of l ight that is a l lowed to pass th rough , at a selected wavelength, is d imin i shed by approximate ly 10 -50% compared to single bandpass emiss ion filters. Second, more noise i s present as undes i red l ight from other wavelengths, a l though m i n i m a l i n mos t cases, is a l lowed to pass through. 3.2.2.3. Objective Lens The objective lens i s perhaps the most impor tant component i n the microscope. It p lays a n impor tant role i n determining i) the spa t ia l resolu t ion , ii) the chromat ic response, iii) the chromat ic and spher ica l aberrat ions, a n d iv) the in tens i ty of l ight i n the system. Spa t ia l resolu t ion i n the objective lens i s dependent on the wavelength as wel l as the n u m e r i c a l aperture of the objective lens (Section 2.1.2). Spher i ca l aberrat ion relates to consis tency i n the size of TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 48 Camera Figure 3.2. B l o c k d iagram of the excitat ion and emiss ion filter sys tem. A n 8- pos i t ion filter wheel is used to select the exci tat ion wavelength i n the i l l u m i n a t i o n pa th . A double b a n d pass d ichro ic a n d emiss ion filter i s u s e d i n the imaging pa th . the object at different points i n the field of view. Chroma t i c aberra t ion relates to the size of a n object when seen under different wavelengths. The object size var ies because the focal point changes w i t h different wavelengths. Fo r h i g h qua l i ty objective lenses, s u c h as the ones we used , spher ica l a n d chromat ic aberrat ions are relatively s m a l l i n the central field of view where the objects l ie TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 49 a n d thus c a n be ignored. Las t ly , the amount of l ight loss th rough the objective lens is of pa r t i cu la r importance i n fluorescence as the s igna l intensi t ies are low. The process of chromosome imaging typica l ly requires h i g h magnif ica t ions a n d h igh numer i ca l aperture objectives. We u s e d the fluorescence 6 3 x magnif icat ion objective lens w i th a n u m e r i c a l aperture of 1.4 (Plan Apochroma t 6 3 x / 1 . 4 , Zeiss) for this project. Th i s objective lens was found to exhibi t the best a l l a round performance over other (Zeiss) 6 3 x a n d lOOx magnif ica t ion objectives for imaging chromosomes a n d telomeres. 3.2.3. Focussing Mechanism We incorporated a n automated focussing m e c h a n i s m into the sys tem for two reasons. Fi rs t , the telomeres and chromosomes do not a l l fal l at the same focus plane. Second, the focus depth of the objective i s smal le r t h a n the size of the objects be ing s tudied (i.e. the entire object is not captured by the lens). The motor (ZSS 43-200-1 .2 , Phypt ron , Germany) a n d control ler ( M A C 4 0 0 0 , Marzhause r , Germany) used can move the focus pos i t ion i n step size increments of 0.1 um. However, the b a c k l a s h or hysteresis effects inherent i n the mechan ics of the microscope (for moving the focus pos i t ion i n a d i rec t ion that is opposite from its previous movement) can range u p to 0.3 um. A s imple technique i s then employed to obta in the mul t ip le focus pos i t ion 2 -d imens iona l images s u c h that the spac ing i n the z-direct ion i s consistent . In our technique, images are acqui red by first moving the objects so they are out-of-focus a n d then acqu i r ing mul t ip le images as the focus pos i t ion is s tepped i n equal TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 50 in tervals i n the opposite direct ion pass ing through the focus point of the object. A s a result , the acqui red mul t i - focus plane images are evenly spaced from each other. The exceptions are i n the first couple of images where the focus spac ing c a n be different because of the back lash /hys te re s i s effects. Hence, these first couple of images i n the series are not used i n the analys is . 3.2.4. High Resolution Camera The qual i ty of the image acquired is h igh ly dependent on the camera used . Hence, we based our selection of the camera on a n u m b e r of key considera t ions w h i c h are desirable for quantitative fluorescence microscopy. These considerat ions resul ted i n selecting a camera w h i c h have the fol lowing requirements: i) h igh spat ia l resolu t ion and large field of view, ii) sufficient photometr ic resolut ion, h igh sensit ivity, and large dynamic range, a n d iii) mul t i - spec t ra l image acquis i t ion capabi l i ty and relatively fast readout rates (Jaggi et a l . , 1993; Pontifex et a l . , 1994, Poon and Hunter , 1994, Vro l i j k et a l . 1994). We chose the Micro lmager MI1400-12 digi ta l camera (Xi l l ix Technologies Corp.) for this project as it meets these requirements . Other cameras w h i c h employ the same C C D sensor and meet these requirements are avai lable from other manufacturers (e.g. Photometries L t d . or P r ince ton Ins t ruments Inc.) bu t are more expensive. For the first requirement stated above, i t i s desirable to have a detector w h i c h c a n sample at a rate w h i c h i s at least twice the resolu t ion of the rest of the sys tem s u c h that a l i as ing effects do not occur . Th i s t ranslates to a sensor w i t h a p ixe l spac ing of less t han 7 microns . Given s u c h a s m a l l p ixe l size, the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 51 detector s h o u l d have approximately 1000x1000 pixels so that it c a n capture the entire cel l (metaphase chromosomes) w i th sufficient deta i l a n d reso lu t ion i n one image. Square and 100% fi l l factor pixels detectors are used i n the X i l l i x camera . Th i s e l iminates the need for geometrical p ixe l compensa t ions a n d increases the p ixe l sensit ivi ty as the entire p ixe l region (rather t h a n a portion) i s sensit ive to l ight. The second requirement dictates the use of h igh dynamic range cameras w i t h sufficient photometr ic resolut ions and sensit ivity. In fluorescence microscopy, s ignals may range over several orders of magni tudes i n intensi t ies (0.0001 - 1 lux) . Typica l ly , var iable exposure time (less t h a n 10s to avoid s ignif icant photobleaching effects) cameras combined w i t h a d y n a m i c range of 10 or more true bi ts (>1024 grey levels or >60dB s ignal to dark-noise ratio) are u s e d i n fluorescence microscopy imaging. General ly, a photometr ic reso lu t ion of only 8 true bi ts (256 grey levels) of informat ion w i l l suffice especial ly w h e n fluorescence images are typical ly fairly noisy. Hence, the eight mos t s ignif icant b i ts of h igh reso lu t ion cameras (e.g. X i l l i x 12 bi ts or 4096 levels camera) are typ ica l ly u sed to represent the image stored i n the computer . Al ternat ively, for less intense images, a sub-region of the sensor's dynamic range m a y be u s e d to ob ta in the same photometr ic resolut ion data (256 levels) wi thout the need to increase the exposure and acquis i t ion t imes. In th is ins tance, a selected por t ion of the 4096 grey levels of the camera i s l inear ly mapped to 256 grey levels. The remainder of the 4096 grey levels are then set to 0 or 2 5 5 depending on whether the respective grey level i s below or above the selected por t ion . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 52 Last ly , the camera m us t be able to capture images over a b road spec t rum of v is ib le l ight and acquired images at sufficient rates. A l t h o u g h the filters i n the fluorescence microscope are used to select the wavelength of l ight for imaging , the sensor mus t also be sensitive to l ight i n th is range. M o s t C C D detectors are made from s i l i con and these exhibi t a higher q u a n t u m efficiency i n the red region than that i n the blue. The favoured red region corresponds to the emiss ion wavelength of the C Y 3 probe used for l abe l ing telomeres i n th is project. Readout rates is a n issue i n megapixel cameras as each new object needs to be re-focused. Too fast a readout rate (30MHz) c a n pose a s t r a in on the sensor a n d degrade the signal-to-noise performance. A s a compromise , we use the b i n n i n g a n d the 8 M H z readout rate features of the X i l l i x camera for focuss ing purposes . In the b i n n i n g mode, several pixels are combined together before they are readout. Th i s increases the sensi t ivi ty a n d reduces the total n u m b e r of p ixels i n the image. A s a result , shorter readout and exposure t imes are required to obta in a s imi la r intensi ty image i n the n o r m a l (not binned) mode. 3.2.5. Computing System O u r comput ing system is s imi la r to the one w h i c h was developed for general imag ing by X i l l i x Technology Corp . (Jaggi et a l . , 1991). Th i s sys tem was chosen because of i ts avai labi l i ty , our famil iar i ty w i th the system, a n d the avai lab i l i ty of the source code to facilitate modif icat ions as required . The compu t ing system consis ts of i) a host 486-based personal computer , ii) 1 gigabyte m a s s storage d i s k space, iii) h igh resolu t ion image acqu i s i t i on a n d TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 53 d i sp lay ca rd (1280x1024x24 bits), and iv) a corresponding h igh reso lu t ion moni tor . 3.3. System Temporal Stability and Aberrations 3.3.1. Overview It i s impor tan t to ensure that the images can be a n d are ca l ibra ted s u c h that the resul ts from w i t h i n a n experiment and from different experiments c a n be compared i n a meaningful way. In order to determine wha t correct ions a n d compensa t ions are required on the acquired data a n d generated resul ts , we performed a n u m b e r of experiments on the system. These experiments a n d their resul ts are descr ibed i n the fol lowing sections. The ana lys i s of the data from these experiments is then used to just i fy what pre-process ing a lgor i thms are required for da ta correct ion a n d compensat ion. 3.3.2. Temporal Fluctuations in Illumination The a i m here is to ensure that accurate measurements of telomeres are made irrespective of w h e n these images are acquired. The i l l u m i n a t i o n source is the major cause for temporal var ia t ions i n the acqui red image. The l ight in tens i ty emitted by the l amp w i l l vary due to the var ia t ions i n the power supp l i ed . In addi t ion , the l ight intensi ty w i l l change over t ime as the l ight b u l b ages. A l t h o u g h we have selected a h y b r i d mercury-xenon lamp w h i c h exhibi ts good tempora l s tabi l i ty properties, we w o u l d s t i l l need to characterize i ts TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 54 s tabi l i ty s u c h that appropriate measures for image norma l i za t ion a n d correct ions c a n be developed and implemented. Fo r character iz ing the f luctuat ions i n the i l l u m i n a t i o n source, we chose a co loured plas t ic (acrylic) sl ide as the test object. Th i s p las t ic exhibi ts s im i l a r f luorescence character is t ics as telomeres by giving off a faint red fluorescence w h e n it i s i l l umina ted w i th green light. S imi l a r filter settings were u s e d to acquire images of the plas t ic and those of telomeres. Images of the p las t ic sample were then acquired at different t ime intervals to determine the s tabi l i ty of the i l l u m i n a t i o n source over t ime. The images were acqui red i n b i n n i n g mode to reduce the n u m b e r of pixels to process a n d store, that i s , every 2x2 p ixe l i n the image was combined into one p ixe l . Hence, the acqui red image was reduced to 6 4 0 x 5 1 2 pixels i n size. For each acqui red image, we ca lcu la ted the average of the measured fluorescence intensi ty i n the centra l 512 x 512 p ixe l region. The averaging helps to smooth out the spa t ia l non-uni formi t ies a n d noise over the region. If we assume that the fluorescence emis s ion of the sample i s direct ly propor t ional to the intensi ty of exci tat ion, the average va lue then gives a n ind ica t ion of the amount of l ight emitted by the i l l u m i n a t i o n source. The d i s t r ibu t ion of the mean scene intensi ty as a funct ion of t ime i s s h o w n i n Figure 3.3. It c an be seen that the l amp intensi ty c a n change dras t ica l ly over t ime. Th i s can be explained by changes that have been made to the i l l u m i n a t i o n source du r ing the course of i ts use. The changes m a y inc lude center ing/ focuss ing the l ight bu lb , rep lac ing filters i n the l ight pa th , etc. It i s also not iced that the l amp does r ema in constant over several hou r s of TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 55 use at a t ime. Thus , the i l l umina t i on can be a s sumed to be constant for each acqu i s i t ion sess ion w h i c h takes 2 to 3 hours to complete as long as no changes are made to the system du r ing the experiment. F igure 3.3. I l lumina t ion var ia t ions over t ime. The i l l u m i n a t i o n remains constant d u r i n g the dura t ion of the (2-3 hour) experiment. The i l l u m i n a t i o n level c a n vary i n between experiments as the optics m a y be moved or changed d u r i n g s u c h t ime (represented by the d i scon t inuous hor izonta l l ines). B a s e d on the above experiment, a ca l ibra t ion method for va r i a t ion i n i l l u m i n a t i o n was developed. Fi rs t , a n average of 10 fluorescence in tens i ty measurements of the acryl ic test sample was made at the beg inn ing of each "telomere" experiment. The averaging helps to compensate for the t empora l noise i n image acquis i t ion . Th i s average intensi ty value signifies the a m o u n t of l u m i n a n c e of the l ight source for each experiment. Hence, the da ta from each exper iment c a n then be scaled appropriately by compar ing the average acry l ic fluorescence va lues and appropriately sca l ing the telomere resul ts amongst experiments . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 56 3.3.3. Photobleaching Effects Photobleaching (fading) is another impor tant aspect to cons ider i n quanti tat ive fluorescence microscopy. Ant i -b leach ing agents are often u s e d to min imize the fading effects of the probes used . If a sample was s ignif icant ly photobleached (i.e. the sample fluorescence emitted was reduced as a funct ion of l ight exposure), a method w i l l need to be developed to determine the amoun t or the stage of photobleaching i n this sample so that the resul ts generated c o u l d be correlated or compared w i th another sample. The d i s t r ibu t ion of the normal ized m a x i m u m intensi ty of telomeres over exposure t ime is shown i n Figure 3.4. A s imi la r d i s t r ibu t ion for 0.1 um fluorescence beads is shown i n Figure 3.5. 0.00 200 4.00 6.00 8.00 10.00 Elapsed Time (minutes) Figure 3.4. Photobleaching effects on telomere fluorescence. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 57 200 §J 140 CU - 120 £, ioo en en c CD 80 60 40 20 4. 0 0.00 1.00 2.00 3.00 4.00 5.00 6.00 E l a p s e d Time (minutes) 7.00 8.00 9.00 Figure 3 .5 . Photobleaching effects on bead fluorescence. It c a n be seen from Figure 3.4 that the fluorescence decay i s approximate ly 1% over minutes of l ight exposure. The fluorescence decay of beads i s h igher at approximately 3 % per minute . In our experiments , we decided not to compensate for photobleaching effects. The reason i s that the va r i a t ion due to photobleaching (approximated by the l ine i n Figure 3.4) i s m u c h less t h a n the var ia t ion i n acqu i r ing a n image (represented b y the difference between consecutive sample points i n Figure 3.4). The latter va r i a t ion c a n be u p to 10% from the expected (line) va lue . The est imated tempora l var iance i n the intensit ies of the acqui red images due to photobleaching is 1% as i t takes 2-3 minutes (with exci tat ion l ight exposure) to TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 58 setup a n d focus the telomere image and approximately another 1-2 minu te to capture the series of mul t ip le plane images. 3.3.4. Uneven Illuminated Field of View We next investigate the unevenness of sample i l l u m i n a t i o n . It c a n be seen from Figure 3.6 that there i s a var ia t ion of i l l u m i n a t i o n over the field of view. The br ight spots i n the image corresponds to the image of the arc of the b u l b . The, r ing- l ike contours of equal intensit ies show that the in tensi ty decrease away from the bright spot. The intensi ty difference between the brightest a n d d immes t spot i n the field of view is approximate ly 10%. To correct for this spa t ia l var ia t ion , we chose the flat-field compensa t ion method (mentioned i n Sect ion 2.2.2.2). A s shown i n Chapter 2, the der ivat ion for the flat-field compensat ion involves t ak ing the logar i thm of the t ransmit tance to obta in the opt ical densi ty values . However, the integrated fluorescence intensi ty (IFI) value i n fluorescence microscopy i s p ropor t iona l to the fluorescence intensi ty and not the logar i thm of the intensi ty. E v e n though no logar i thm convers ion is required, th is compensa t ion is s t i l l v a l i d . O u r reason for th is c a n be explained by the unevenness i n i l l u m i n a t i o n i n the field of v iew as s h o w n below. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 59 Figure 3.6. I l lumina t ion var ia t ion over the field of view. The image shows a contour map of the var ia t ions i n l ight intensit ies. There is a two grey level difference between two adjacent regions. The brightest spot i n the image is represented by the grey intensi ty region near the center. The l ight intensi t ies decrease towards the edges of the image. If the i l l u m i n a t i o n at the center of the field of view generates a n object fluorescence of intensi ty Ic - D (where D is the fixed offset va lue w h i c h is independent of the level of i l luminat ion) , then a different intensi ty i n i l l u m i n a t i o n scaled by a factor of s, w i l l generate a scaled object fluorescence of in tensi ty slc -D= IE -D. If the scaled i l l umina t ion is not i n the center of the field of view, the fluorescence intensi ty at the off-centered pos i t ion is s t i l l a scaled vers ion of the center intensi ty. Hence, each measured fluorescence value I(x,y) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 60 i n the image c a n be scaled by the appropriate sca l ing factor s(x,y) (corresponding to i l l u m i n a t i o n differences i n the field of view) at that p ixe l to generate a compensated fluorescence value C(x,y). Th i s i s expressed as: C ( x , „ ) - D = M ^ ,3-!) s[x,y) The scale factor s(x,y) at each point can be obtained from a n homogeneous fluorescence mater ia l p laced i n the microscope. If the fluorescence response of the background mater ia l at each poin t i s B(x,y) - D a n d the fluorescence at the center of the field of view i s Bc - D (constant k), the scale factor at each point i s given by: »,«,„, = ̂ E £ = « ( £ f z £ , 3 . 2 | B y inser t ing th is scale factor into equat ion (3-1), the fol lowing resul t w h i c h i s s imi l a r to the k n o w n flat-field compensat ion fo rmula (Section 2.2.2.2) is obtained: Clx,y) = k I { X , y ) ° +D (3-3) 1 ,y> B(x,y)-D The above flat-field compensat ion method involves d iv i s ion operat ions w h i c h w i l l generate rea l (rather than integer) numbers . Since images are stored i n integer number s (8 bits), t runca t ion or quant iza t ion errors w i l l arise. Hence, th i s flat-field compensat ion image is not carr ied at th is stage. Instead, we TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 61 performed flat-field compensat ion later a n d only on those pixels w h i c h are u s e d i n ca lcu la t ing the IFI value of the telomeres. The resul ts and d i scuss ions of this compensa t ion method i s presented below. 3 . 3 . 5 . Flat-Field Compensation Results The fluorescence acryl ic sample that i s used earlier for the t empora l f luc tuat ions i n i l l u m i n a t i o n experiment is used for flat-field compensa t ion measurements . In this experiment, images are acqui red at different i l l u m i n a t i o n levels (inserting neut ra l densi ty filters i n the i l l u m i n a t i o n path). The brightest image is u sed as the background reference image a n d the other images are processed u s i n g the flat-field compensa t ion method of Chapte r 3.3.4. A cross-sect ion of the intensi ty d i s t r ibu t ion i n the centra l row of p ixels of the field of view, before a n d after the flat-field compensa t ion a lgor i thm is s h o w n i n Figure 3.7 a n d 3.8 respectively. U s i n g flat-field compensat ion, i t c a n be seen (Figure 3.8) that a l inear response c a n be obtained over the field of view (standard deviat ion of less t h a n 1 grey level) for different levels of intensi ty of i l l umina t i on . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 62 250 _ 200 m I 150 cu S £ 100 cu ^ 50 200 217.9 +.6.1 150.6 +4.3 106.4 +3.4 43.2 .+ 1.9 400 600 800 Pixel Position 1000 1200 1400 Figure 3.7. Intensities of the central row of pixels of a homogenous sample . 250 ~ 200 v> > cu - 150 cu 5 100 '35 e co £ 50 228.0 +.0.0 157.6 JiO.7 111.3 Hi 0.7 45.2 +0.8 +- 0 200 400 600 800 Pixel Position 1000 1200 1400 Figure 3.8. F la t field compensat ion for spa t ia l i l l u m i n a t i o n var ia t ions . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 3.4. Image Acquisition Software 63 3.4.1. Overview The " S S M " program w h i c h we used i n our acqu is i t ion sys tem was developed by X i l l i x Technologies Corp . for acqui r ing , s tor ing, and m a n i p u l a t i n g the images acqui red by the Micro lmager camera . Th i s p rogram r u n s u n d e r the Microsoft D O S operat ing system. It has the capabi l i t ies to control a l l camera funct ions bu t does not have a l l the functions w h i c h we need for th is project. Hence, we modif ied the program to a l low for i) automat ic camera exposure t ime a n d photometr ic range selection, ii) integrat ion of mul t ip le focus p lane acqu i s i t ion , a n d iii) corrections for p ixe l defects. 3.4.2. Image Exposure and Photometric Range Selection We implemented automat ic exposure t ime a n d photometr ic range select ion funct ions into the program to control the camera . These funct ions max imize the dynamic range of the image s u c h that the grey level i n the acqui red image span over most of the 256 levels. The au tomat ion of these funct ions also s implif ies the setup process for image acquis i t ion a n d reduces the t ime required for adjustment. The l ineari ty of the camera as a funct ion of exposure t ime and photometr ic response makes the au tomat ion a lgor i thm s imple a n d easy to implement . In the program, the photometr ic range is implemented by a lookup table w h i c h maps the camera 's 0 -4095 grey levels scale to the computer memory 's 0-255 grey levels. In this a lgor i thm, a l l grey levels below the chosen m i n i m u m value i n the camera i s set to 0 i n the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 64 computer memory. S imi la r ly , a l l grey levels above the chosen m a x i m u m value i n the camera are set to 255 i n the computer . The range in-between the m i n i m u m a n d m a x i m u m values is then l inear ly mapped to the closest integer i n the range of 0-255. In addi t ion , the range between the chosen m i n i m u m a n d m a x i m u m levels is set to a mul t ip le of 256 . Th i s avoids a b ias i n m a p p i n g more levels i n the camera to a given level (between and not i n c l u d i n g 0 to 255) i n the computer . The resul t ing grey level h is togram of the image w o u l d then be smooth a n d w o u l d not have peaks at the b ias locat ions. The image exposure time and the chosen values for the m i n i m u m a n d m a x i m u m levels are based on informat ion i n the image. Ini t ial ly, a n image i s acqui red w i t h a n exposure time of I s and w i t h a l inear m a p p i n g of the 0-4096 range to the 0-255 range. The intensi ty d i s t r ibu t ion of the image i s evaluated a n d the exposure t ime is adjusted (down to a m i n i m u m of the 3 0 m s l im i t a n d u p to a m a x i m u m of the 10s l imi t of the camera). The a i m i s to ob ta in a n image w h i c h i) has the brightest p ixe l w i th a n intensi ty va lue of greater t h a n 230 grey levels, ii) has less t han 10 pixels set to a grey level of 2 5 5 (i.e. to avoid image saturat ion), and iii) has the da rk region set to w i t h i n the grey levels of 0- 50. If the selected exposure time does not give the desired image intensi ty d i s t r ibu t ion , the m i n i m u m and m a x i m u m values for the m a p p i n g funct ion are adjusted accordingly. 3.4.3. Image Pre-Processing: Sensor Defects Megapixe l C C D cameras often have a n u m b e r of inherent ly defective pixels . These p ixe l defects are largely caused by the loca l a c c u m u l a t i o n of d a r k TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 65 current . A s the size of the accumula ted da rk current i s related to the temperature of the sensor, some cameras employ cool ing to reduce the size a n d n u m b e r of s u c h defects. These defects typical ly appear as br ight spots or l ines i n the image. They are not a problem at low integrat ion t imes. However, at long integrat ion t imes (more than 3s a n d longer for cameras where the C C D i s cooled), the defective pixels may become saturated even w h e n no l ight is present. A s the informat ion i n these pixels are lost, the only method i s to f ind a n approximate value for the p ixe l . We used the average of the s u r r o u n d i n g non-defective p ixe l intensit ies as the replacement in tensi ty va lue . A s these defects on ly pose a p rob lem at long integrat ion t imes, the defective p ixe ls i n the image are replaced only i f a n integrat ion time of greater t h a n 2 s i s u s e d to acquire the image. 3.5. Image Analysis System 3.5.1. Image Analysis Hardware The image ana lys i s hardware consis ts of j u s t a compu t ing system. The requirements for th is comput ing system are less stringent t h a n those of the acqu i s i t ion system. We chose bo th Pen t ium and 486-based persona l computers for th is purpose. The Pen t ium based mach ines are str ict ly u s e d for da ta ana lys is . The slower 486-based systems c a n be used for image ana lys i s w h e n i t i s not be ing used for image acquis i t ion . We also chose commerc ia l ly avai lable d i sp lays (e.g. N E C 3D) and d isp lay cards (e.g. ATI V G A Wonder) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 66 capable of h a n d l i n g a m i n i m u m resolut ion of 6 4 0 x 4 0 0 x 8 bi ts . Typica l ly , a reso lu t ion of 1024x768 i s used . 3.5.2. Image Analysis Software The ana lys i s software first reads the image files acqui red from the acquis i t ions system into the computer memory. It then performs the segmentat ion a n d generates the resul ts for interactive ver if icat ion a n d edit ing. We developed th is program to operate i n the Microsoft W i n d o w s V e r s i o n 3.1 operat ing system environment. The resul ts of the telomere a n d chromosome ana lys i s i s s h o w n i n Figure 3.9. The key user- interact ion features w h i c h were required as defined i n consu l ta t ion w i th the users are implemented into the ana lys i s program. These factors are: i) d i sp lay ing a n enhanced view of the b a n d i n g s t ructure i n the chromosome for ease i n karyotyping (chromosome classification), ii) d i sp lay ing the relative telomere posi t ions on the chromosome image, iii) m a r k i n g a n d labe l ing each telomere i n a chromosome, iv) d i sp lay ing the segmentat ion resul ts , v) m a r k i n g a n d labe l ing each chromosome, vi) d i sp lay ing the resul ts of the quant i f ica t ion (chromosome number , n u m b e r assigned, telomere IFI for each chromosome, chromosome IFI a n d area feature values), vii) sor t ing the chromosome l is t based on the selected feature, a n d viii) edi t ing capabi l i t ies . The resu l t ing implementa t ion i s shown i n Figure 3.9. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 67 pie fcdit applications Options! Erocess Window S t e v e n ' s l i n a g e A n a l y s i s N : \ S T E V \ I E L O X . I M U \Aiti x 4 J Z x U H F P | W f/li L a b e l l e d O b j e c t H i s t a g r a r r O p t i o n s F d i t O b j e c t s 1 3 n S i : ? - 6 =s 3 n IFI 06 «in Mas C o u r t M - d r Mm i n =Td D e v 9 : 5 7 6 52 i : o 273 - : ' 14 J7 73 5 3 8 46 :-6 257 55 1 33 27 7 i 5 5 8 39 24 256 46 1 36 3 ; 84 4 5 2 3 ' : i 6 247 51 103 83 1 : 1 1 4 3 18 86 21 47 52 2A i£4 i 3 ! *"•• =| Cliruinusumt: S ta t i b t i u& i h - F i t it S a v r S o r t P r i n t L:.l C I* T I T ; T 3 T4 T- T o t . l =>a : i nT» l . C D 3 j r .u A i a a Cu r r r - iwn ± |.| 8 II 11 Oil ' ' 6 3 9 1 1 " 5 8 2 -- 135 : 7 • 2 : 0 232 0 917 0 79 4 8 1 : 0 67 •4 34 2" 7 0 0 0 217 0 00 9 4 1 9 99 ; • so : m 251 216 0 1232 1 43 4 7 2 = 7 ft ft • 4 4 5 145 • 24 120 129 1.1 5 8 3 0 85 4 •3 3 3ft 7 ! • 4 2 6 _ 2* : 365 115 99 1.1 r 9 6 .' 71 4 31554 •I: 9 2 9 • 341 2 2 5 • :>: 0 983 1 64 4 5 5 1 0 ' i • 0 3 5 • •1 : 4 : 3 ; 4 3 8 1242 0 61 4 10742 8 7 2027 0 0 0 0 0 00 0 5 l 00 22 : 2" 2 231 654 542 0 1645 0 38 4 44-36 ' 5 9 7 9 207 31.1. II 8 0 4 0 Lll.l 2 4 9 ' t l 9 i ' 0 5 8 •: 2 5 : 55 1 : 7 0 0 577 2 4 6 3 2226 67 5 4 2 147 ' 6 . I.I IJ 0 30 : 0 00 2 • 6 4 r . ' . 4 : - • J • : : ; •392 150 155 Ij 1601 4 25 4 1 0421 95 " 9 1 9 ft 3 5 : 2 5 7 :-9 20 Ij 3 7 ; 2 79 4 2324 7 3 65 331 294 =9 132 Ii 9 5 : 2 71 4 : 0 92 " 3 3 5 • 3 3 ; 381 2 ; s 256 0 1252 1 32 4 10329 »• 2 0 2 8 i ft 1 95 1 1 254 0 614 II 72 4 4 2 : 3 1 1 i o 147 2 1 ; 3 :1 0 Ij ft 01 1.09 3 5421 80 0 32 1 5 : 297 6 1 4 0 Ij l l j f t : 0 74 3 6 1 2 9 80 " 129 407 341 V S ; s f i Ii ' 394 1 " 6 4 7,3 70 114 " 5 5 8 _ 1 5 : 1 " : ' 1 : 0 3 : 0 0 458 0 II 1 5 ; 1114- 0 00 ' 1 'ii : 25 12216 1 1 3 21.19 ' 14 243 261 = 6 19- Ij 701 1 82 4 7 4 1 7 79 6 3 2 _ :-: 3 9 : 30 1:8 20 3 35 l " 9 2 1 40 5 3 2 7 6 T2 7 : 7 : (. 0 2 : 317 138 112 1 07 l i f t : 2 02 6 9 1 2 7 104 " 8 2 5 " i 212 2 5 2 0 0 ci 1 39 3 2 9 3 0 99 8 4 7 579 4i • • ' 10" I 1 3 2 ; .' Ill 4 6 2 2 7 ••I'l 1,4113 20 5 5 ; 271 i ; 4 0 j i o i :• 4 51 3 9 9 2 9 68 • 9 1 0 3 0 234 " 6 2 0 0 0 404 0 00 2 " 8 2 9 95 632 : i 164 2 9 : 3 ; 4 268 0 1062 0 76 4 •:: i o 8 3 • 4 1 " - _ 224 0 0 c 2724 0 00 155 1 3 " 226 1 77 • II .•111 . 85 II 86 0 71 4 [•,'•11 II 91) " 4 4 " 2 4 ; 2 2 : 177 229 Ij s s : 1 • ? 4 7 9 : 3 33 i W 1 5 : " U : 2 : 4 . j : IJ 81 : 0 47 4 : : : 8" 9 5 6 4 6 ; 564 1 : 8 295 0 1511 2 " 2 4 6 5 3 3 80 " 2 5 6 - ~ 164 Ij 0 0 ft 4 0 00 11 98 78 ; Ci 264 ' 27 2 5 9 233 I'l 004 0 79 4 8124 98 " 9 7 5 : 9 1 7 : 0 0 Ij 173 I] 00 25 35 : 9 : 5 9 7 4I-.4 2 * 6 253 I 5 3 r 2 22 4 II 4 B" 6 7 6 J 1 2 ; ' 1 ; 0 0 I] 242 0 00 2 • 6 3 4 97 5 3 3 y 1 7 ; 14 -::: 493 0 1312 0 29 4 31 3 3 77 9 3 3 : • 474 in' 2 4 ' Ii 145 = 1 82 4 13550 ? 2 2 5 1 5 u 2 2 7 • 6 : 1=4 I T ; 0 =•53 1 08 4 7 5 7 7 1 OU • 8 3 2 '_ J " 3 D : 294 5 2 8 : 8 9 1 1433 8 72 4 'I i 1 111 734 41 i s : 9 5 3 2 2 ; 6 0 0 967 0 40 4 1 2 4 2 0 73 2 6 5 7 " 4 7 141 15 2 ; : 3 35 27 907 0 39 5 I 3 0 2 7 32 2 5 3 5 _ 48 6 7 : 5 1 : 149 189 0 152-2 3 02 4 G729 • • ' 5 1 7 F i g u r e 3.9 T e l o m e r e ^ A n a l y s i s P r o g r a m . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 68 The resu l t ing image d isp layed shows the borders of the telomeres over la id onto the processed chromosome image. To accom pl i sh this , the image of the chromosome i s first thresholded s u c h that the b a c k g r o u n d pixels are set to a mid- in tens i ty grey level of 128. Th i s grey level a l lows the chromosome borders a n d bo th da rk and bright intensi ty bands i n the chromosomes to be seen on a grey background . The image is then inverted (linear map of 0-255 to 255-0). Cont ras t s t retching is then performed on the chromosome to enhance the detai ls of i ts b a n d i n g structures. The image of the telomere objects are next processed (as descr ibed later i n Chapte r 5). To ensure that the telomeres lie at the ends of the chromosomes i n the super imposed telomere-chromosome image, a pat tern m a t c h i n g a lgor i thm i s then used to determine the placement of the detected telomere borders onto the chromosome image. In this ma tch ing a lgor i thm, the telomere image i s first shifted, p ixe l by p ixe l , from the chromosome image. For each p ixe l shift locat ion , the n u m b e r of detected telomeres that are w i t h i n the borders of the detected chromosomes is determined. The p ixe l shift va lue (x,y) w h i c h generates the highest n u m b e r of te lomere /chromosome object matches becomes the p ixe l shift value for a l igning the two images. The border of the telomeres are then super imposed onto the enhanced chromosome image. A different color i s ass igned to each of the telomeres i n the chromosome for ease of user recogni t ion (purple, red, blue, cyan , orange, a n d green are ass igned to telomere n u m b e r 1, 2 , 3, 4, 5, and 6 or more). A l t h o u g h there are only 4 telomeres i n each chromosome, 2 addi t iona l colors are used to facilitate the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 69 edi t ing of the detected telomere resul ts and to identify other non-telomeric probes (e.g. centromere of chromosome 17). F ina l ly , the border of the chromosomes are overlaid on the enhanced chromosome image w i t h the highl ighted telomere borders. A g a i n , a different color i s u s e d to indicate the n u m b e r of "telomere" objects i n the chromosome (cyan, green, a n d yel low are assigned to 3 or less, 4 a n d 5 or more "telomere" objects, respectively). E a c h chromosome objects are then labeled w i t h a un ique number . A corresponding l is t of chromosome and telomere IFI can be d i sp layed i n a w i n d o w beside the image. A n enlarged sub-region of the image c a n also be d i sp layed to a id i n v i sua l i z ing the details of the image. The user c a n then edit the informat ion d isplayed. The edit ing capabi l i t ies inc lude i) the j o i n i n g of chromosomes (which are improper ly segmented), ii) sp l i t t ing "telomere" objects ( touching telomeres), iii) reass igning the telomere n u m b e r i n the chromosome (pair sister telomeres: 1 w i t h 2 a n d 3 w i t h 4), iv) r a n k i n g the chromosomes based on i ts size or IFI value, v) ass igning a chromosome n u m b e r to each chromosome (for ka ryo typ ing purposes), and vi) add ing a comment to the data. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 70 Chapter 4. Acquisition System Characteristics 4.1. Background A n under s t and ing of the character is t ics of the acqu is i t ion sys tem w o u l d give ins ight into how images are formed a n d how telomere fluorescence measurements cou ld be evaluated. Th i s system character is t ics c a n be represented mathemat ica l ly . A s ment ioned i n Chapter 2 (equation 2-3), the i n p u t object i s modif ied by the transfer funct ion of the system to resul t i n a n output image. In our system, the image i s affected by the components i n bo th the microscope and the camera. If we assume l inear i ty a n d space invar iance i n the system, the overal l opt ical transfer funct ion, O T F s y s , w i l l consis t of the con t inuous transfer funct ion of the microscope, O T F o p t , mu l t i p l i ed by the transfer funct ion of the discret iz ing camera, O T F c a m . OTF s y s (u,y) = OTF o p t(u,y) x OTFcam(u,v) (4-1) The transfer funct ion of the microscope can be derived based on geometric a n d Four ie r optics theory. If l ight is considered to be composed of m a n y point rays, the incoherent i l l umina t i on used i n the fluorescence microscope w o u l d then be composed of point l ight rays where the phase of each poin t var ies independent ly of one another. The incoherent P S F i s then represented by the power spec t rum of the p u p i l funct ion of the sys tem (Goodman 1988). Thus , i f the p u p i l funct ion p(x, y) of the sys tem i s TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 71 represented by the system's aperture, the p u p i l funct ion at the image plane (at d d is tance from the aperture of the objective lens) i s then represented by p(Xdx, Xdy), where X i s the wavelength of light. The P S F o p t of the microscope i s then the Four ie r t ransform of the p u p i l funct ion a n d is given by: P S F o p t ( x , t / ) = F\ p(Xdx, Xdy) } (4-2) The normal ized incoherent O T F o p t c an then be determined from the normal i zed auto-correlat ion of the p u p i l funct ion a n d i s given by: OTFAu,v) o p t K " " U ) 2?p(0,0) J Jp{Xdx, Xdy)p(Xdx - u,Xdy - v)dxdy (4-3) —oo—oo 00 00 J" j p2 (Xdx,Xdy)dxdy H o p k i n s (1955) was the first to theoretically determine the O T F of a n incoherent i l l umina t ed system for both the out-of-focus a n d in-focus cases. In h i s der ivat ion, the out-of-focus p u p i l funct ion for a c i r cu la r aperture sys tem of r a d i u s A (as s h o w n i n Figure 4.1) observed at plane d2 w h e n the object i s at dis tance da a n d the in-focus p u p i l funct ion is at plane d{ (i.e. 1 = — d0 dt f w h e r e / i s the focal length of the lens and rij and n2 are the refractive ind ices i n the object a n d image sides of the lens) is given by: TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 72 p(r) = red 2A e x p jkw (4-4) where rect(x) - 1 if x < 1 and = 0 otherwise tu = - d , - 8 Z cosa + ( d , 2 + 2 d , 8 z + 5 2 2 cosa 2 )^ = defocus error X a = arctan d,. dt = locat ion of the focussed image of object at dis tance d Q = d 2 =location of the focussed image of object at dis tance d1 di + 5 z = d2 Notice that i f 5 Z = 0, then w = 0 and the p u p i l funct ion p(r) becomes the system's aperture funct ion. B y subs t i tu t ing the p u p i l funct ion into equat ion 4-3 a n d the spa t ia l frequencies (u,v) w i t h (s), the resul t ing O T F o p t becomes: OTF o p l (s) = —cos 1 as fMH + z l - i r 1 sin(2np) n = l 2n [ J 2 n - l ( a ) - ^ 2 n + l ( ° 0 sin %a — as v 2 j sin[(2n + l)p] Z(-D n = 0 where a = 2kws (3 = cos - / c = spa t ia l cutoff frequency. o i V a n ( a ) - J 2 n - 2 ( a ) ] 2n + l L J 2 V u 2 (4-5) s = fe TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 73 Refractive Index n n In-focus Object _JL + _ 2 = J L d + d f 1 2 Image / Detector Plane Figure 4 .1 . Rela t ionship of the in-focus and defocus dis tances i n the object a n d image side of the objective lens. Th i s der ivat ion takes into account both the x ,y spa t ia l (s) a n d defocus error (w) effects. The resu l t ing OTF is not a s imple funct ion a n d takes a long t ime to compute . S tokseth (1969) later p roduced a n approx imat ion of th i s O T F w h i c h takes less t ime to compute. Stokseth 's approx imat ion for a defocus error of vu i s given by: OTF o p ( (s, w) « ( l - 0.69s + 0.076s 2 + 0.043s 3 ) j inc 4ktu r i s S 1 - J V 2 ) 2 where TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY J,(x) jinc(x) = 2 x for s < 2 74 (4-6) J1(x) = the first order Besse l funct ion of x. C a s t l e m a n (1979) found a more accurate vers ion of Stokseth 's app rox ima t ion for s m a l l va lues of defocus error to. For Cas t leman 's method , the p o l y n o m i a l i n Stokseth 's approximat ion is replaced by a n in-focus OTF term (—(2(3 - sin2(3)) mul t ip l i ed by a defocus effect (jincfc) term w h i c h is a K funct ion of the spa t ia l frequencies, u a n d v, and the defocus va lue , u>). Hence, at a focus plane (out-of-focus distance of amoun t 8 2 ) from the in-focus image at dis tance di i n Figure 4.1) and wi th a lens system hav ing a n aperture of r ad ius A , the O T F is given by: O T F o p ( (q,w) « - (2p - sin2p) jinc where q2 = u 2 + v2 4kw fa fc 2A fc (4-7) The above O T F o p t represents the transfer funct ion for general defocus optic systems. For the specia l case of concern w h i c h i s the microscope system, E r h a r d t [1985] a n d others used a s imi la r vers ion of the above formula . They expressed the theoretical OTF i n terms of the parameters of the microscope system. For example, Erhardt ' s derivat ion of the system O T F i s based on Stokseth 's method and is given by: TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 75 2TI • h(p) -fc-dz where (4-8) f(p) = 1 - 1.38/? + 0 . 0 3 p 2 + 0 . 3 4 4 p 3 h(p) = NA-p{l-p) 2 U 2 +V 2 P = L2 NA = n u m e r i c a l aperture of the objective. In Erhard t ' s derivat ion, the var iables are based on the parameter dz w h i c h i s on the object side of the lens. The defocus error, here, i s a s s u m e d to be the out-of-focus distance dz. Because J1[dJ/dz or jinc(dj i s symmet r i ca l about dz=0, the OTF values at the positive defocus distance dz i s the same as that at the negative defocus distance -dz (i.e. O T F o p t [p,dz) = O T F o p ( (p,-dz )). Thi s symmetry characteris t ic , however, i s not true i n reality. In what follows, we f ind the OTF w h i c h shows its asymmetr ic nature. 4.2. Our Derivation of the System OTF In our system, as wi th most microscopes i n c l u d i n g Erhard t ' s , the loca t ion of the sensor from the objective lens, dis tance d^, i s fixed a n d the sample i s moved i n relat ionship to the objectives (Figure 4.1). W h e n the sys tem i s in-focus, a n object point source i n the object plane d1 p roduces a poin t source i n the imaging (sensor) plane d^. W h e n the object po in t source is moved by a distance dz to the out-of-focus plane d^, a point source at dis tance d{ resul ts a n d the response observed at the sensor plane d^ (given by equat ion 4-9) i s a b lu r red image. B y moving the object point source to m a n y different TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 76 pos i t ions a r o u n d the distance d1 a n d a long the optic axis , a n d by de termining the O T F va lues at the detector plane for each of these object posi t ions , the 3- d imens iona l O T F of the system can be obtained. It c a n be deduced from dx+d + n 0 1 (refer to Figure 4.1) that the var iable dj, a n d thus 8 Z a n d dt f c o s a va ry as the defocus amount d z varies. These var iables are used to calculate the defocus error, iv (equation 4-4). We now consider Cast leman 's approx imat ion for the O T F of the microscope for s m a l l defocus amounts (equation 4-7): OTF o p ( [u,v,tu) « - ( 2 P - sin2p) jinc 8% X w- ylu2 + V2 fc 4u2 + v2 fc (4-9) where w, the defocus error, is defined i n equat ion 4-4. For posit ive a n d negative defocus amounts , ± d z , the calculated va lues for the defocus error, w, are not the same. Hence, the P S F is asymmetr ic (i.e. has different va lues at equal posit ive a n d negative defocus dis tances d z from the focal point). Thus , we s h a l l use equat ion (4-9) to represent the OTF of the microscope. O u r next step i s to define expressions for the d{, 5 Z, and cos a terms a n d use these terms to evaluate the defocus error, w, and then O T F o p t of equat ion 4-9. To f ind the re la t ionship between the in-focus and defocus dis tances 8 Z , we write the re la t ionship between the object plane distance a n d that of the image plane (as indica ted i n Figure 4.1) as determined from Lens law. The re la t ionship of the in-focus pos i t ion is : TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 77 1 _ Mj n2 f d2 ^ 10^ a n d that of the out-of-focus is 1 f d„ d- dx + dz n2 d 2 - (4-11) where the distance d^ i s generally fixed i n a microscope system. The two equations i n (4-10 and 4-11) are then equated (dropping the focal length term). The terms i n the equat ion are then rearranged s u c h that the change of focus i n the image side is expressed i n terms of i) the change i n the object side (dz), ii) the refractive indexes of the med ia on bo th side of the objectives ( n 1 ( n^), iii) the distance of the sensor from the objectives (d^), a n d iv) the magnif ica t ion (M) of the objectives. Th i s express ion is given by: d0 • M -n. -d, 5Z = 2 L _ E (4-12) Tl2 d0 + (M n. +n0)-d, M 1 2 / 2 where M = ^ dx There r ema in two other var iables i n the general equat ion (4-4), the aperture A a n d angle a, w h i c h need to be expressed i n the more c o m m o n l y k n o w n parameters of the microscope system. Firs t , the aperture, A , of the sys tem c a n be derived from geometry and equated to the n u m e r i c a l aperture, NA, of the objective lens. Th i s is given by: TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 78 r\ • s i n 9 = nx • = NA (4-13) ^d2 + A2 In the der ivat ion of the OTF (equation 4-9), i t was a s sumed that A « dv Hence, equat ion (4-13) can be approximated by: nl- — = NA (4-14) dx Rearranging equat ion (4-14) for A and express ing the dis tance d1 i n terms of the distance of the imaging plane from the lens d^ resul ts i n the fol lowing express ion for A: d9 • NA A = ^ (4-15) Second , the term cos a, c a n also be derived from geometry a n d i s given by: d, cosa = , (4-16) Vd,a + A2 where di = d^ - 5 Z . The values for 8Z a n d A c a n be evaluated from equat ions (4-12) a n d (4-15) respectively. The expressions defined for d ; , 5 Z, a n d COS a , c a n then be used i n equat ion 4-9 to evaluate O T F o p t Now that the O T F of the microscope is determined, the next step i n de termining the O T F of the system is to derive the transfer funct ion of the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 79 camera . C C D cameras are commonly used for quanti tat ive mic roscopy imaging . The photons (light), w h i c h reaches the C C D a n d have sufficient energy, w i l l generate electrons i n the C C D . In the active sens ing area of the C C D , the electron (or the corresponding hole) charges w i l l be t rapped i n the potent ia l wel ls (pixels). In the non-sens ing area, the electrons are absorbed by the sensor a n d does not contr ibute to the charges i n the potent ia l wel l . Hence, the camera samples a n d integrates the l ight w h i c h only falls i n the active sens ing area of each p ixe l . If the area of the sens ing element i s rectangular , a n d i f the response to l ight over the entire area of the p ixe l dose not vary , then the resu l t ing point spread funct ion, P S F , is the convolu t ion of the s a m p l i n g funct ion, sampQ, and the rectQ funct ion. This can be wri t ten as: P S F c a m ( x , y , z ) _1_ Px rec t Px s a m p (*}] 1 T- rec t p . V y_ P« s a m p v A J . PxPy rec t v P x y rec t ry J i j where (4-17) Px > Py = s ^ z e °f p ixe l i n the x and y direct ion respectively A x , A y = sample spac ing i n the x and y di rect ion respectively. For our camera , the p ixe l size a n d the spac ing between any 2 p ixe ls are essent ial ly the same i n bo th the x and y direct ions a n d m a y be replaced by p (i.e. p = p x = p y = A x = A y ) . If we assume that the x a n d y di rect ions are independent of each other, then the OTF of the camera w h i c h is j u s t the Four ie r t ransform of the P S F and is given by: TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 80 (4-18) Previously , only the transfer funct ion of the microscope a n d the s a m p l i n g propert ies of the camera have been used for image recons t ruc t ion purposes (Erhardt et a l . 1985) a n d the integrat ion process at each sample point is ignored. If our addi t ion of the p ixe l sens ing area i s t aken into account i n the der ivat ion (as shown above), the resul t ing theoretical transfer funct ion (given by equat ion 4-1) shou ld better resemble the exper imental ly determined funct ion (i.e. the desired OTF ). The images of our theoretical system OTFs are s h o w n i n Figure 4 .2 . The ca lcu la t ion for the OTFs are based on the derivat ions i n Sect ion 4 .2 . These OTFs are evaluated w i t h a n objective lens hav ing a magnif ica t ion of 6 3 x a n d a n u m e r i c a l aperture of 1.4 a n d located at a dis tance of 1 5 0 m m from the camera detector, a n o i l m e d i u m wi th a refractive index of 1.515 between the objective lens a n d the sample , a n emiss ion wavelength of 5 7 0 n m , z-spacings of 0.1 um, a n d a detector p ixe l size of 6 . 8x6 .8um 2 . A point source object is pos i t ioned at focus p lane z=0. Star t ing at a n out-of-focus pos i t ion of -1 .6um, a 3 2 x 3 2 p ixe l image of the O T F d is t r ibu t ion is generated. The out-of-focus pos i t ion i s subsequent ly increased by 0.1 um and another image of the OTF i s generated. The process of generating OTF images at 0.1 um z-spacing is then repeated u n t i l a total of 32 O T F images are generated ranging from -1 .6um to 1.5um (Figure 4.3. Theoretical OTF and PSF Results TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 81 4.2). The corresponding calculated images of the system P S F s are shown i n Figure 4 .3 . E a c h of the 32 images of the P S F is obtained from the inverse Four ie r t ransform of the corresponding OTF image i n Figure 4.2. The system P S F i n the x (or y), z-plane is d isplayed i n Figure 4.4. Last ly , a plot of the system (radial xy) P S F d is t r ibu t ion at vary ing z-spacings (focus levels) i s shown i n Figure 4 .5 . Figure 4 .2 . Theoret ical OTFs of the system i n the xy-plane. E a c h OTF image is evaluated at 0.1 um i n the z-direct ion from its ne ighbour ing images. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 82 Figure 4 .3 . Theoret ical P S F s of the system i n the xy-plane. E a c h P S F image is evaluated at 0.1 um i n the z-direct ion from its ne ighbour ing images. Figure 4 .4 . Theoret ical P S F of the system i n the xz-plane. Image (b) is a logar i thmic scaled vers ion of image (a). The p ixe l spac ing i n the z-direct ion i s 0. l u m . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 0.00 5.00 10.00 15.00 Pixels (0.1um) J(a) c CO •o CO "3 E ho O Z 6 8 Pixels (O.lum) J(b) Figure 4 .5 . Theoret ical P S F d i s t r ibu t ion of the system at va r ious z-spacings. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 84 It c a n be seen from the images of the system OTFs a n d P S F s that their responses appear c i r cu la r symmetr ic i n the xy-plane; th is i s as expected since the active sens ing element a n d p ixe l areas have the same x a n d y d imens ions . The in tens i ty of the 3 D P S F is highest at the centra l po in t (in the xyz-space). The intensi ty decreases as the distance from the centra l point i s increased i n the x , y direct ions. F r o m Figure 4.4, it c a n be seen that the w i d t h of the spot i n the x (or y) d i rect ion is ha l f that of the z direct ion. Hence, as est imated i n Sect ion 2 .1 .2 , the x (or y) direct ion is twice the reso lu t ion of the z d i rec t ion. W i t h i n approximate ly 0 .6um (the locat ion corresponding to the cutoff frequency) i n the z-direct ion from the central point , the in tensi ty also decreases w i t h inc reas ing distance from the central point . A t above +0.6um i n the z- di rec t ion, however, a l ternat ing da rk a n d bright r ings (intensities) begins to appear i n the OTF (Figure 4.2) and P S F (Figure 4.3) images. Th i s is caused by the osci l la tory effects of the Besse l funct ion i n the equat ion of the theoret ical O T F . It c a n also be seen from the images that the P S F s a n d O T F s at equa l d is tances i n the positive and negative z-directions (about the centra l point) are s imi l a r (part icularly w i t h i n ± 0 . 6 u m range). A t approximate ly ± 0 . 6 u m , s l ight difference i n the P S F s i n the positive and negative z-direct ions begins to become apparent i n the image. It appears that the negative z-direct ion appears to decrease i n magni tude sl ight ly faster t h a n i n the posit ive di rect ion. The asymmetr ies of the P S F i n the z-direct ion i s the resul t of the non- l inear m a p p i n g of the positive a n d negative z-movements as d i scussed i n Sec t ion 4 .2 . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 85 It c a n be seen from Figure 4.5 that the asymmetr ies between posit ive a n d negative focus amounts are present even at s m a l l defocus amoun t s (< 0.6um). A n interest ing point to note from the theoretical P S F image (Figure 4.3) is that the s u m of intensi ty values over the 32x32 p ixe l (x,y) area for each focus p lane i s the same. That i s , the total intensi ty informat ion c a n be obta ined from either the "sharp" in-focus image or the "very b lur red" out of focus image (e.g. at 1.5um) where the point source i s ha rd ly seen. The intensi ty s u m var ia t ions are i n the order of 2%. These var ia t ions resul t from t runca t ion errors (256 grey level images) a n d because the 32x32 p ixe l area i s not large enough to encompass the entire funct ion. A l though the intensi ty informat ion i s present i n the out-of-focus image, it i s very difficult to extract s u c h informat ion s ince the b o u n d a r y of the point source cannot be easily detected. W h e n there are other objects nearby, the segmentation problem at out-of-focus p lanes becomes more difficult . A s seen later i n Chapter 5, we have developed a segmentat ion a n d in tens i ty extract ion method to estimate the total intensi ty informat ion from the spot. 4.4. Initial OTF/PSF Comparative Study A n in i t i a l s tudy was carr ied out to compare the theoret ical P S F s y s w i t h the exper imenta l P S F s y s of the system. Two different exper imental approaches i n ob ta in ing the P S F s y s of the system were taken (Poon et a l . 1993b). O u r first app roach u t i l i zed s m a l l fluorescent beads to s imulate point sources of l ight. The resu l t ing 3 D images acquired w i l l then represent the P S F of the sys tem. In th i s method, 0 .15um fluoresbrite beads (Polyscience Inc. War r ing ton , PA) were TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 86 used . These beads were excited w i th blue l ight a n d emitted l ight i n the green (515nm) spec t rum. These beads were smal ler t h a n the resolving reso lu t ion of the microscope, bu t yet large enough to enable detection by the sensors. Sets of 50 images t aken at O . l u m z-spacing were acquired. Because the fluoresbrite beads photobleaches, the images acquired were corrected (restored), to compensate for the decrease i n s ignal intensi ty as a funct ion of t ime. The correct ion was based on the a s sumpt ion that the fract ional r educ t ion i n in tensi ty i n sequent ia l images is a constant (Lockett et a l . , 1994). After the correct ion, the highest intensi ty p ixe l i n the s tack of images was chosen as the center of the bead. E a c h xy-plane image i n the z-stack was then rad ia l ly t ransformed to produce the r ad ia l d i s t r ibu t ion of in tensi ty levels about the center point . The r ad i a l d i s t r ibu t ion was later u sed for compar i son w i t h other theoret ical a n d exper imental P S F of the system. O u r second approach for determining the P S F uses a step edge to derive the l ine spread funct ion, L S F , of the system (Tatian, 1965; Cas t l eman , 1979). The L S F c a n be obtained by differentiating the image of a step image response (since the derivative of a step funct ion has a n infinite va lue at the poin t of the edge a n d zero elsewhere). The L S F i s then rotated to give the P S F (i.e. P S F ( - N / x 2 + y2 ) = LSF(x)). The test target slide for the step edge cons is t s of chrome evaporated onto glass (Edmund Scientific Company) to produce a n opaque area on the sl ide. L ike the bead images, sets of 50 images were acqui red at 0.1 um z-spacings. F r o m each xy-plane image, the centra l po in t of each l ine perpendicu lar to the edge was determined. The centra l po in t i s the loca t ion where the intensi ty i s ha l f way between the br ight a n d d a r k regions of TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 87 the l ine profile. E a c h l ine i n the xy-plane image was then al igned w i t h others at their cor responding center points . The respective va lues a long the a l igned l ines were s u m m e d and averaged. The resu l t ing step funct ion was then m e d i a n filtered, by a filter of 5 pixels i n w id th , to remove sharp peaks a n d val leys i n the funct ion. The step funct ion was then differentiated u s i n g a difference ke rne l of [-1, 1], to generate the P S F of the system. The in-focus (z=0) system responses for va r ious objective lenses ( 2 0 x / 0 . 7 0 NA, 4 0 x / 0 . 8 5 NA, 6 0 x / 1 . 4 0 NA, and 1 0 0 x / 1 . 4 0 N A , Nikon) are s h o w n i n Figure 4.6. For each objective lens used , a compar i son i s made amongst i) the exper imental P S F from bead images, ii) the exper imenta l P S F from step edge images, iii) the theoretical P S F based on Erhard t , a n d iv) the theoret ical P S F we derived. B o t h the bead and step edge images are acqu i red th rough a 5 1 5 n m b a n d pass filter. It c a n be seen that the measured and theoretical system responses are very s imi l a r i n shape. For a l l objective lenses, the P S F obta ined from the step edge has the widest response, followed by the bead P S F , our theoret ical P S F , a n d then Erhard t ' s P S F . The resul ts for this sequence can be expla ined b y the following. The step edge is not inf ini tes imal ly t h in bu t has some th i ckness w h i c h i s larger t h a n the size of the beads. In tu rn , the beads a l though s m a l l are of finite size, and hence they are only approximate point sources. O u r P S F is wider t h a n Erhard t ' s because we took into cons idera t ion the integrat ing effects of the camera sensor i n our derivat ion. Hence, our P S F i s more representative of the behaviour of the system as it i s closer to the exper imenta l resul ts . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 88 20x/0.70Objective Lens 0.0 0.5 1.0 Distance f rom center (microns) 40x/0.85 Objective Lens 0.0 0.5 1.0 Distance from center (microns) 0.0 0.5 1.0 Distance from center (microns) 0.0 0.5 1.0 Distance from center (microns) O u r T h e o r e t i c a l PSF Erhardt 's PSF PSF from B e a d s PSF from Step Edge Figure 4.6. In-focus system response for var ious objectives. It c a n also be seen that our theoretical P S F is very s imi l a r to the exper imenta l bead resul ts . Th i s i s especial ly true for the 4 0 x a n d lOOx TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 89 magnif ica t ion objective lenses. A s the beads are smal le r t h a n the reso lu t ion of the microscope a n d hence can be considered as point sources, ou r theoret ical P S F gives a very good approx imat ion to the character is t ics of the microscope system. O f the two experimental approached carr ied, the step edge images are easier to acquire . They c a n be performed i n brightfield microscopy. T h u s , they require less exposure time for image acquis i t ion a n d hence less noise i s generated i n the images. Due to the th ickness of the chrome i n the step edge a n d the jagged look ing edge at h igh magnif icat ions, however, the system response generated i s not as good as the response of the beads. Hence, the f luorescent beads at a s imi la r wavelength to that of the C Y 3 telomere probe are u s e d for further compar i son (mainly the values of the P S F i n the z-direction) on the system P S F . 4.5. Comparison With Experimental PSF In the fol lowing experiment, we again use beads as test objects for compar i son w i t h the theoretical P S F . Th i s t ime, however, the beads are acqu i red unde r the s imi la r condi t ions as those used for telomere acquis i t ion . That i s , the beads are acquired under the same magnif ica t ion and spect ra l wavelength used for acqu i r ing telomere images. We used the 6 3 x magnif ica t ion objective w i t h a numer i ca l aperture of 1.4 and the green exci ta t ion a n d the h y b r i d red (570nm) emiss ion filters. To s imulate point sources of l ight , beads of 0.1 a n d 0.2urn i n diameter are used as test objects. The images acqu i red from these beads w i l l then be a n approx imat ion to the P S F of the sys tem. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 90 Smal le r size beads are more representative of point sources bu t are harder to manufacture consis tent ly and have a weak fluorescence s ignal w h i c h cou ld not be re l iably detected by our imaging system. Mul t i - focus plane images of a n u m b e r of 0.1 um beads at z-spacings of O . l u m are shown i n Figure 4.7. It c a n be seen from the bead images that their responses resemble that of the theoretical resul ts (Figure 4.3). They a l l have c i r cu la r symmetr ic responses i n the xy-plane. U n l i k e the theoretical results , the images of the beads occupy over approximately twice the distance i n the z-plane. Th i s difference can be more easily seen i n the intensi ty profile of the bead over different focus levels (Figure 4.8). Figure 4 .7 . Typ ica l images of four O . l u m beads acquired at different focus spacings . E a c h image i s spaced 0.1 um i n the z-direct ion from its ne ighbour ing images. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 91 100 cu en <d 80 w 60 c •a 40 cu N E 20 O Z 0 -0.5 0 0.5 Z-Focus Pos i t ion (microns) Figure 4 .8 . C o m p a r i s o n of experimental a n d theoretical P S F d i s t r ibu t ions as a funct ion of z-focus posi t ion. A typ ica l 0.1 um bead i s u sed to represent the exper imenta l P S F . The intensi ty d i s t r ibu t ion of the bead is broader t h a n that of the theoret ical P S F . Th i s is caused by the addi t iona l b l u r r i n g effect in t roduced w h e n the beads are imaged through mater ia ls w i t h different refractive index. In p repar ing the sample, the beads are first p laced onto the glass sl ide (which ha s a refractive index of 1.5. A moun t ing m e d i u m so lu t ion i s then p laced over the beads. Th i s so lu t ion contains the ant i -photobleaching agent, Vec torsh ie ld , a n d has a refractive index of 1.45. The th ickness of th is so lu t ion c a n va ry from 50 to lOOum. A glass coverslip of refractive index of 1.5 i s then p laced over the so lu t ion . Immers ion o i l of refractive index of 1.5 i s then p laced i n between TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 92 the coversl ip a n d the objective lens. Because of the difference i n the refractive index between the moun t ing m e d i u m and the rest of the mater ia l i n the imag ing pa th , l ight from a point source object is refracted w h e n i t h i t s the refractive surface interface. Thus , from the objective lens, it appears that the l ight rays that passes through originate from a n u m b e r of different po in t sources at different focus planes. Th i s b l u r r i n g effect has been recognized by a n u m b e r of investigators (Vander-Voort and Brakenhof f 1990; Sheppa rd a n d Cogswel l , 1991 , Car l s son , 1991, Visse r et a l . , 1991 , 1992; a n d H e l l et a l . 1993, 1995). We d i d not take into account the moun t ing m e d i u m i n the der ivat ion of the theoret ical P S F because the th ickness of the m o u n t i n g m e d i u m is var iable a n d i s not fixed from sample to sample. Hence, the amoun t of refraction a n d the resu l t ing b l u r r i n g w i l l be different from image to image. In addi t ion , it i s difficult to predict the reflection of the fluorescence s ignals from the glass surface on w h i c h they lie as the distance of the object from the surface m a y vary over 0 .5um. Th i s phenomena i s seen i n Figure 4.8 where another in tens i ty peak i s observed i n the negative focus pos i t ion . Th i s secondary peak i s derived from the addi t iona l l ight w h i c h is par t ia l ly reflected from the bot tom glass surface. Since the intensi ty and d i s t r ibu t ion of the reflection s ignals are governed by how far the object is and what reflective index is at the m o u n t i n g m e d i u m a n d glass interface, a general izat ion for the theoret ical P S F w o u l d be difficult to generate. In summary , i f the moun t ing m e d i u m has the same refractive index as the other mater ia l i n the imaging path , then our theoret ical P S F gives a good TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 93 approx ima t ion to the response funct ion of the system (Figure 4.6). Th i s , however, i s not the case w i th our telomere study. Nevertheless, we u s e d our theoret ical P S F to generate images of s imula ted objects s u c h that we c a n evaluate the robus tness of our IFI a lgor i thm (Section 5.5) to objects of va ry ing intensi t ies a n d shapes. A s shown later i n Sect ion 5.2, the IFI va lue of the object ca lcula ted from the image generated u s i n g our theoret ical P S F w i l l be s imi l a r to that ca lcula ted from the image obtained u s i n g the true P S F of the system. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 94 Chapter 5. Telomere Segmentation and Integrated Fluorescence Intensity Measurements 5.1. Overview Thi s chapter describes the method we developed to calcula te the integrated fluorescence intensi ty (IFI) of telomeres i n fluorescence microscope images. The IFI value is a measure of the total amount of fluorescence emitted from the object and is correlated to the length of the telomere. For our s tudy, we use a n u m b e r of images acquired at different focus planes . The IFI va lues are first determined for each of the mult iple-focus p lane images. These IFI va lues are then combined i n a n u m b e r of different ways a n d the resul ts are compared to determine the best combina t ion scheme to evaluate the IFI value of the object. Tradi t ional ly , most of the research i n the quant i f icat ion of F I S H images involve spot count ing , relative distance measurements , a n d event enumerat ions . W i t h the in t roduc t ion of more accurate a n d efficient quant i f ica t ion probes, and better microscope systems a n d ant i -photobleaching agents, new biological s tudies to determine the fluorescence intensi ty of the probes are be ing carr ied out. Typica l ly , only one image for each p robe / spec t r a l wavelength at the best focus level is analyzed to extract the relevant informat ion . Th i s "best" focus level is often chosen manua l ly . That i s , the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 95 sharpest l ook ing image is v i sua l ly selected as the focus i s m a n u a l l y var ied . A different method of locat ing the "best" focus locat ion is to automate the "best" focus select ion process. In this case, the "best" focus level i s chosen by m i n i m i z i n g or m a x i m i z i n g a par t icu la r feature. However, as we have s h o w n i n a n earlier s tudy (Poon et a l . 1992a,b), the op t imal focus level for one feature m a y not be the o p t i m u m one for other features. Fur thermore , the z-distance between op t imal focus level of different features w i t h i n one cel l differs from cel l to cel l . In that s tudy, we have also found that a tighter feature d i s t r ibu t ion can be obta ined i f more than one image a round the "best" focus pos i t ion i s acqu i red a n d the feature va lues are accumula ted over these planes . We have also s h o w n i n the s tudy that a tighter d i s t r ibu t ion of feature va lues c a n be obta ined i f the feature ca lcu la t ions are performed on processed images where out-of- focus b l u r are first removed. A s ment ioned earlier, most quantitative F I S H image research use only one image, the best focussed image, at each wavelength of l ight. In some cases, more t h a n one image acquired at different focal p lanes are used . For mul t ip le focus plane analys is , the tendency has been to first reconstruct the 3 D image. In th is instance, the b l u r of the mul t i - focus plane images i s removed before extract ing the features. U n l i k e our present research, i n the t rad i t iona l appl ica t ions of F I S H imaging , the values of the fluorescence intensi t ies do not have to be accurately k n o w n . In those appl ica t ions , the reso lu t ion i n the quant i f ica t ion c a n be coarse bu t w i t h i n l imi t s w h i c h enable detection of the loca t ion of the fluorescence spots or objects. A h igh degree of accuracy , however, i s required i n our present work i n de termining the IFI s ince the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 96 amoun t of fluorescence detected i s used to estimate the length of telomeres and the d i s t r ibu t ion of their lengths i n a chromosome. Recently, a s tudy w h i c h requi red m u c h higher degree of accuracy than those of t rad i t iona l app l ica t ions a n d w h i c h determine the IFI has been investigated (Lansdorp et a l . , 1996). Th i s chapter first describes the theoretical ana lys i s for ca l cu la t ing the IFI of a n object. It then d iscusses our method for ca lcu la t ing a n d de te rmin ing the IFI va lues of the objects. O u r telomere segmentat ion a lgor i thm a n d our method for determining the n u m b e r of focal p lanes required for the IFI quant i f ica t ion are also described. F ina l ly , a n u m b e r of methods to val idate our resul ts are d i scussed . 5.2. Theory in IFI Quantification In fluorescence microscopy, the IFI of a n object i s the total fluorescence detected from the object, o(x,y,z). Hence, the IFI of a fluorescence object, i s the total in tensi ty of the object and can be represented by the integral of in tens i ty at each point of the object (in the x,y ,z plane) and is given by: (5-1) A s ment ioned i n Sect ion 2 .2 .3 , the or ig ina l object c a n be reconst ructed from the observed image, i(x,y,z), by convolving the latter w i t h a recons t ruc t ion filter, g(x,y,z), i.e. o{x,y,z) = i[x,y,z) ® g(x,y,z), a n d thus , TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 97 IFI = jjji(x,y,z)® g(x,y,z)dxdydz (5-2) x,y,z where ® represents the convolut ion operator. E q u a t i o n (5-2) c a n be simpli f ied by replac ing the convolu t ion operator w i t h i ts mu l t i p l i ca t i on counterpart as follows: IFI - |JJ jjji(x,y,z) • g(u - x, v - y, u> - z) • dudvdwdxdydz x,y,zu,v,w - jjji(x,y,z) • jjjg[u - x,v- y,w-z)dudvdu> \-dxdydz x,y,z \u,v,w J = jjji(x,y,z) • K • dxdydz x,y,z = K • jjji(x,y,z) • dxdydz (5-3) x,y,z where K i s the integral over the entire range of the recons t ruc t ion filter. Th i s integral va lue , K , i s a constant and does not vary w i th the image. Since we are only interested i n the relative or normal ized IFI value of objects, i t i s not necessary to determine the value of K as i t w i l l be factored out i n the normal i za t ion process. It i s shown i n equat ion (5-3) that the IFI of the object i s propor t iona l to the IFI ca lcula ted from the observed image. Hence to calculate the relative IFI of the object, i t i s not necessary to first reconstruct the observed image i(x,y,z) to obta in o(x,y,z). The IFI of the object c a n also be calcula ted from a t ransformed image s(x,y,z) obta ined by convolving the observed image i(x,y,z) w i t h some filter, p(x,y,z), (whether i t i s reconstructive or not). To show this , we define IFIp as: TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 98 IFIp = K • jjj s(x,y,z)dxdydz x,y,z = K • 111 i(x,y,z) <8> p[x,y,z)dxdydz x,y,z F r o m a s imi l a r ana lys is as above, the resul t of the convolu t ion i n equat ion (5-4) i s a constant factor mul t ip l i ed by the IFI of the object. IFIp = KpK • jjji(x,y,z) • dxdydz = Kp-IFI (5-5) x,y,z Thi s means that i f the IFI is difficult to calculate from the observed image, the observed image can be transformed into another image where the IFI ca lcu la t ions c a n be more readi ly determined. The above ana lys i s shows that the IFI of a n object c an be ca lcu la ted from the observed 3 D image of the object accord ing to equat ion (5-3). In our system, the 3 D image is obtained by acqui r ing a n d c o m b i n i n g a series of different focus level images as seen on the image detector plane. The image at the detector plane, i (x,y,z d ) , contains either a n in-focus or out-of-focus object. W h e n the object i s in-focus, most of the s ignal intensi t ies w i l l be loca l ized i n a s m a l l region. W h e n the object i s out-of-focus, the observed intensi t ies are dispersed over a larger region as the image is altered by the out-of-focus P S F . It c a n be seen from the previous chapter that the magni tudes of the cent ra l frequencies of the OTF of a n image at any z locat ion are the same. That i s , OTF(0,0 ,z) = l V z (5-6) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 99 This then impl ies that the total intensi ty of the P S F i n the xy-plane i s the same for a l l z-focus values: A s s h o w n i n equat ion 5-7, the total intensi ty observed i n the detector p lane (assuming that the image is large enough to capture the in tens i ty variat ions) is the same for in-focus or out-of-focus objects. T h i s c a n also be explained from a geometric perspective. It i s k n o w n that the in tensi ty of l ight observed at a distance from the poin t source var ies inversely w i t h the square of the distance. If we place a n aperture i n the l ight pa th , on ly the l ight that enters the aperture of the objective lens is a l lowed to pass through. If we look at the amount of l ight w h i c h h i t s the image plane w h e n the latter i s at a distance located close to the aperture, the l ight w i l l be d ispersed over a n area s imi la r to the size of the aperture. If the image plane i s moved further away, the l ight observed w i l l be less intense bu t w o u l d cover a larger area w h i c h increases by the square of the dis tance from the point source. The total l ight observed i n both ins tances is the same. In our sys tem, we a s sumed that the changes i n focus is s m a l l compared to the dis tance of the object to the objective. Hence, the amount of l ight w h i c h passes th rough the aperture c a n be a s sumed to be invar iant w i t h respect to s m a l l focus changes. Th i s ana lys i s is v a l i d i f we also a s sumed that the l ight losses i n the optic pa th (e.g. filters, objectives, cover s l ip , o i l , moun t ing m e d i u m , etc.) are invar ian t w i t h focus changes. (5-7) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 100 The ca lcu la t ion of the IFI can then be s impl i f ied to ca lcu la t ions performed on only one image plane (since s imi la r intensi ty s u m s are present i n other focus planes). Equa t i on (5-3) becomes: where Z i s a constant representing the n u m b e r of Z-focus p lanes of the observed object. 5.3. Segmentation and IFI Quantification Algorithm F r o m the above analys is , the IFI of the object c a n be s i m p l y obta ined i f their observed intensi t ies are s u m m e d over the entire region i n w h i c h they occup ied regardless of w h i c h focus plane the image is taken . The difficulty l ies i n de termining the region i n w h i c h to perform the i n t eg ra t i on / summat ion . Telomeres are generally not isolated from one another s u c h that the s igna l intensi t ies of one do not interfere w i th that of another. W h e n they are in-focus, most of the telomere s ignal intensit ies are concentrated i n a spot whi le the rest of the s igna l i s dispersed i n the su r round ing w h i c h m a y con ta in other telomeres. W h e n they are out-of-focus, the telomere s igna l intensi t ies are spread over a larger region and hence is more l ike ly to interfere w i t h the s ignals of other telomeres m a k i n g i t more difficult to determine the IFI of each telomere. (5-8) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 101 Thus , the major problem i n accurately quant i fying the IFI of telomeres l ies i n their segmentation, i.e. determining the exact boundar ies of each telomere. Mos t telomeres are relatively easy to detect since they appear as br ight spots. Approximate locat ions of these spots can be found by th reshold ing or edge detection methods (e.g. Sect ion 2.2.4). However, a p rob lem arises w h e n f inding the exact locat ion of the borders . F igure 5.1 shows the (two possible) borders B1 and B 2 of a cross sect ion of a telomere. If the est imated telomere borders are closer to their centre in tensi ty (Bj), the IFI value w i l l be under-est imated. Conversely, i f too m u c h b a c k g r o u n d in tens i ty i s i n c l u d e d i n the estimated border (B 2 ) , the IFI value w i l l be over-estimated. There i s also the p rob lem of segmenting telomeres w h i c h are close to each other a n d de termining w h i c h pixels belong to w h i c h telomere. In addi t ion , not a l l telomeres lie i n the same focus plane. Hence, segmentat ion i n 3 D space (which is rarely done i n b iological imaging) may be required. The re la t ionship between the true IFI of the object a n d that ca lcu la ted from the segmented object from a given 2 D image a s s u m i n g the level for the b a c k g r o u n d intensi ty is B g n d (Figure 5.1) c a n be represented by the fol lowing equat ion: = ^ R e g . B g n d + e B g n d , R e S + e 0 u t + QE (5"9) where g . B g n d ~ the calcula ted value of the IFI of the segmented region calcula ted at the given backg round (Bgnd) level a n d over the defined region (Reg) 8 B e n d R e g =the error i n selecting the backg round level TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 102 s D u t = the error cons is t ing of IFI ca lcula ted outside the defined vo lume QE = the error i n quant iz ing the s ignal into discrete va lues . Cross section of telomere Figure 5 .1 . E r ro r s i n ca lcu la t ing the object IFI. The inherent noise i n the system (optics and i l l u m i n a t i o n aberrat ions, camera noise , sample prepara t ion noise, etc.) and ne ighbour ing telomeres m a k e i t difficult to define the "true" background level a n d segmented region for the IFI ca lcu la t ion . It c a n be seen from equat ion (5-9) that there is a compromise i n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 103 select ing the segmentation area a n d background level s u c h that the total of a l l IFI errors is sma l l . It c a n also be seen that it i s easier to estimate the true bounda ry of the telomere i f i t i s in-focus rather than out-of-focus. A s a result , the errors of the IFI ca lcu la t ion are smal ler for in-focus objects. Not a l l telomeres are at the same focus level a n d have the same shape. Thus , segmentat ion over different focus p lanes are required. The first step of our segmentation a lgor i thm i s to f ind the loca t ion of each telomere i n the x , y, as wel l as the z-direct ions since telomeres have va ry ing lengths a n d d is t r ibut ions i n a l l direct ions. Th i s is accompl i shed by first search ing a n d recording the locat ions of the different loca l br ight spots (center par ts of telomeres) i n each of the different focus plane (x,y) images. These spot locat ions are then compared to their cor responding spot locat ions i n other images of different ne ighbour ing focus planes. For each spot locat ion , the image at the focus (z) plane w h i c h has the brightest spot in tensi ty is considered to be the z-plane conta in ing the center of the spot. Once the center of a telomere spot i s found, the extent of that spot i n 3 D space is then determined by u s i n g the a lgor i thm descr ibed later i n Sect ion 5.4. The segmentat ion a lgor i thm we used to f ind the spot i n each of the mul t i - focus level x y planes is s imi la r to that of the L a p l a c i a n filter (Russ, 1990). For each p ixe l , the average intensi ty value of i ts s u r r o u n d i n g p ixe ls i s subt rac ted from its intensi ty, I(x,y) to generate a n edge image, E(x,y), that i s : TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 104 1 1 1 E[x,y) = I(x,y)--YYl{x-i,y-j) (5-10) We n a m e d th is operat ion the Average Difference filter a n d i ts operat ion (in one d imens iona l space) i s i l lus t ra ted i n Figure 5.2. If the resu l t ing E(x,y) value i s above a threshold level, the p ixe l i s considered to be a telomere p ixe l . Otherwise, i t i s considered as a b a c k g r o u n d or bounda ry p ixe l . The algor i thm j u s t descr ibed el iminates noise by u s i n g a threshold above the noise level and detects intensi ty peaks . A t the center por t ion of a telomere, the average value of the su r round ing pixels i s generally less intense a n d hence the value E(x,y) i s positive a n d large. A t the edges of the telomere, the average value of the su r round ing pixels i s generally the same as the p ixe l va lue since on average, ha l f of the su r round ing pixels have lower intensi t ies t h a n the central p ixe l and the other ha l f have higher intensi t ies . A s a result , the value E(x,y) i s sma l l and near zero. S imi la r ly , at the b a c k g r o u n d region, the average value of the su r round ing pixels is s im i l a r to that of the p ixe l . Hence, by u s i n g thresholding, the noise p ixels i n the b a c k g r o u n d a n d the edge pixels of the telomeres are removed. Telomeres w h i c h are close to each other c a n also be separated u s i n g th is technique. The reason i s that the val ley i n between two nearby telomeres is lower i n intensi ty t h a n the average s u r r o u n d i n g a n d thus can be removed by thresholding. A n example of the use of our segmentat ion method on a typ ica l telomere image i s s h o w n i n Figure 5.3. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 105 Figure 5.2. App l i ca t i on of the average difference filter. The Average Difference filter is first app l ied to the s imula ted s ignals of the telomere spots. A th reshold above the Average Difference operat ion then defines the centers of spots (shaded i n the diagram) detected by the a lgor i thm. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 106 (0 (d) Figure 5.3. Process of segmenting a typica l telomere image. A typ ica l f luorescence image of the telomeres (a) is processed w i t h the Average-Difference filter to generate image (b). A threshold level is then selected from the h is togram of the processed image and this threshold is appl ied to 5.3b to generate a b inary image (c) of the telomeres. The resu l t ing bright spots i n 5.3c are first labeled and then dilated to generate the f inal m a s k image for the telomeres. The boundar ies of the segmentation resul ts (d) are then generated. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 107 The threshold level for the segmentation i s chosen s u c h that mos t of the b a c k g r o u n d noise are e l iminated by thresholding. If the d i s t r ibu t ion of noise i s a s s u m e d to be G a u s s i a n , the peak value w i l l correspond to the centra l po in t i n the G a u s s i a n d i s t r ibu t ion . Since most of the pixels i n a telomere image belong to b a c k g r o u n d pixels , the m a x i m u m peak of the h is togram corresponds to the b a c k g r o u n d level. The value at 5% of the peak value corresponds to the poin t at 1.7a of the G a u s s i a n d i s t r ibu t ion . Hence, a l l points less t h a n 1.7a va lue from the peak corresponds to approximately 9 5 % of the points i n the G a u s s i a n d i s t r ibu t ion . The search for the point corresponding to 5% of peak value is made at the left of the peak (lower intensi ty values) i n the h is togram. T h i s side of the h is togram peak is u sed (instead of the higher in tensi ty side) because the interference from the telomere s ignals w h i c h are located at higher in tensi ty va lues w o u l d be less. The 1.7a point to the right of the peak c a n then be ca lcu la ted . The distance of this point from the peak is then added to the peak point to ob ta in the threshold level w h i c h el iminates approximate ly 9 5 % of the b a c k g r o u n d pixels . Th i s level seems to be the op t ima l for removing b a c k g r o u n d noise p ixels and also preserving the relevant telomere peaks . Objects w h i c h are too s m a l l and whose intensi t ies are s im i l a r to the b a c k g r o u n d level (those objects pointed out by the arrows i n Figure 5.3c) are classif ied as artifacts a n d rejected from further analys is . The above a lgor i thm removes the edges of telomeres. To recover these edges, we first need to dilate (Figure 5.3d). However, th is w i l l combine or fuse mul t ip l e telomeres into one. To overcome this p rob lem, we first perform label ing . In the labe l ing process, each con t inuous connected object i s given a TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 108 un ique n u m b e r a n d then the size of each telomere m a s k is di la ted by one to recover the lost edges of that telomere. Note that i f d i l a t ion was performed before the label ing , objects w h i c h are closed together m a y be connected a n d considered as one object by the labe l ing process (those objects po in ted out by the ar rows i n Figure 5.3d). 5.4. Number of Focus Planes Required It was shown i n Sect ion 5.2 that only a single image from a single focus p lane is required to represent the IFI value of the object (see Sect ion 5.8). However, from Sect ion 5.3, i t c an be seen that i f the objects are out-of-focus, more errors are in t roduced i n the IFI value due to segmentat ion errors (i.e. def ining the region for the IFI ca lcu la t ion and the intensi ty of the background) . Hence, the best focus image shou ld be used for de termining the IFI value of the object. In a given image, however, not a l l objects are at their cor responding best focus. Telomeres can be located at different focus levels a n d they c a n have va ry ing lengths a n d d is t r ibut ions i n the z-direct ion. In h u m a n s , chromosomes on a microscope slide are typical ly 0 .5um th i ck i n the z-direct ion (the approximate size of the chromosome tip). Telomeres range i n size from 0.1 to 0 .3um i n diameter a n d can lie anywhere w i t h i n the tip of the chromosome. Hence , to accurately quantify the telomeres, images from mul t ip le focus p lanes s h o u l d be examined . To experimental ly determine the n u m b e r of p lanes requi red for ca lcu la t ing the telomere IFI values , approximately 20 images at 0.1 um spac ing TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 109 are acqui red . One of the centra l images i n this series conta ins the "best" focus image. The first a n d last few images i n this 20 image series are b l u r r y b u t are s t i l l d iscernible images of the telomeres. A z-spacing of O . l u m i s u s e d because it i s the approximate l imi t of the step size of the z-drive motor. It also corresponds to the size of the smallest telomere, and corresponds to at most 2 0 % of the s amp l ing resolut ion i n the z-direct ion of the objective lens i.e. we are over sampl ing . Three different methods for ca lcu la t ing the IFI of a telomere are performed. The resul ts of these are then compared to determine w h i c h i s the best method to use. One method uses the image plane w h i c h appears to be at the best focus to calculate the IFI of each telomere i n the image. Not a l l telomeres i n the image are in-focus i n a single image plane. Hence, the ca lcu la ted IFI va lues for out-of-focus telomeres is lower i n value from the cor responding IFI ca lcula ted at the best focus pos i t ion for that telomere. The second method ut i l izes a n u m b e r of different focus p lane images a n d selects the best focus (highest evaluated IFI value) image to calculate the IFI va lue for each telomere i n the image. In the last method of IFI ca lcu la t ion , the IFI value of each telomere i n the image i s determined from the s u m of the cor responding telomere IFI va lues ca lcula ted at each of the mul t i - focus plane images. The methods are elaborated u p o n below. For each image acquired, the telomere segmentat ion and the IFI ca lcu la t ion measurements are performed. E a c h telomere i n each image is then matched w i t h each of i ts corresponding telomere i n the other mul t i - focus plane TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 110 images a n d i s given a un ique label . Since the sample is only moved i n the z- di rec t ion (no other opt ical component is moved d u r i n g the acqu i s i t ion sequence), there i s little registrat ion error. Hence, no correct ion i n the x-y shift i s requi red i n the mul t ip le plane images. There c a n be, however, some differences i n the locat ion of the x-y center of a telomere spot i n different z- plane images. Th i s is because telomeres can be i r regular ly shaped a n d c a n lie i n any or ientat ion i n 3-d imens iona l space. Thus , the center of each telomere spot i n a pa r t i cu la r focus plane image may be different from i ts ne ighbour ing planes . To m a t c h telomeres at different focus planes, the telomere center at a given focus plane shou ld correspond to a point not necessar i ly the center bu t ins ide the area of the corresponding telomere i n a n adjacent focus-plane image. Two nearby telomeres can also be segmented by our a lgor i thm. A s a telomere becomes out-of-focus, it gets b lu r red a n d gradual ly fades into the b a c k g r o u n d . Hence, w h e n two nearby telomeres are in-focus, they m a y be detected as two i n d i v i d u a l objects by our segmentation a lgor i thm. However, w h e n they are out-of-focus, they get b lu r red a n d thus m a y be detected as a single telomere. To overcome this p roblem, the area of each telomere i s ca lcu la ted as we l l as i ts IFI value for each focus plane. If the area of a detected telomere i n one image i s greater t h a n twice that of a cor responding telomere i n another focus-plane image, then two (or more) telomeres m u s t be present. In th i s ins tance, the area corresponding to the telomere i n the image w i t h the highest IFI value (in-focus image) w i l l define the area from w h i c h the IFI i s reca lcula ted i n the other image plane (out-of-focus image). Th i s w o u l d reduce the area w i t h w h i c h the IFI shou ld be ca lcula ted . T h u s the IFI ca lcu la ted TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 111 w o u l d be smal ler t han the "actual" value. Nevertheless, it w o u l d be better to segment nearby telomeres and estimate their IFI va lues t h a n not to i nc lude this da ta i n the total IFI ca lcu la t ion for mul t ip le focus plane images. The IFI va lues ca lcula ted for a telomere from a l l images at different focus p lanes are then s u m m e d to obta in the s u m m e d IFI value of that telomere. The above method i s not prac t ica l to perform i n most s i tuat ions because of the large n u m b e r of images required for each cel l . H a n d l i n g these images w o u l d then require large d i s k storage space a n d long image acqu i s i t ion a n d process ing t imes. To reduce the amount of da ta and acqu is i t ion t ime, the n u m b e r of p lanes acqui red needs to be reduced. To a c c o m p l i s h th is , we m a y keep the same O . l u m spac ing between images bu t only acquire the centra l images of the 2um span . Th i s method mus t accommodate for a large va r i a t ion i n telomere sizes a n d z-focus posi t ions. Another method of r educ ing the da ta i s to acquire images at higher sampl ing spacings (e.g. 0 .2um, 0 .3um, ...) bu t over the 2um span . In this case, it then becomes a matter of de te rmin ing the highest s amp l ing spac ing that do not significantly increase the error i n the IFI ca lcu la t ion of the telomere. Theoretically, i f we sample at ha l f the reso lu t ion of the objective lens, the images acquired shou ld be representative of the object. Th i s corresponds to a sampl ing spac ing of no more t h a n 0 .3um. Th i s spac ing is exper imental ly explored below. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 112 5.5. Algorithm Evaluation And Validation 5.5.1. Overview In th is sect ion, we first estimate the spa t ia l resolu t ion y ie lded b y our a lgor i thm. Th i s evaluat ion gives a n ind ica t ion of how far apart two telomeres have to be before they are classif ied as two dis t inct objects by our telomere segmentat ion a lgor i thm. We then test the va l id i ty of our IFI quant i f icat ion method. For th is , we use test objects w i t h k n o w n IFI values . A theoretical ana lys i s is not easy to perform because of the non-l ineari t ies i n the segmentat ion step (the IFI va lues are based on the area defined by the segmentation result). We use 3 different methods out l ined below to validate our results . The resul ts a n d d i scuss ions of the methods are covered below. One method for va l ida t ion is to use s imula t ion , i.e. to cons t ruc t s imula t ed objects w i th k n o w n IFI values a n d shapes a n d convolve these objects w i t h the P S F of the system. The var ious IFI quant i f icat ion a lgor i thms are then appl ied to the generated images. The theoretical P S F of the system a l though s imi l a r to that of the microscope system used , m a y not be a n accurate representat ion. Nevertheless, th is P S F is sufficient for va l ida t ion purposes since the funct ion used for the system P S F needs only be spat ia l ly invar ian t as s h o w n i n equat ion 5-5. U s i n g this method, we mus t choose objects w h i c h are representative of the shapes and d is t r ibut ions of telomeres. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 113 Another method for va l ida t ing our quant i f icat ion method i s to use different sizes of s m a l l fluorescence beads as test objects. The images of these beads are acqui red a n d our IFI quantifying process is appl ied . The measured IFI va lues are then compared w i th the estimated IFI of the beads based o n the a s s u m p t i o n that the IFI of the bead is propor t ional to its vo lume. A precau t ion i n u s i n g th is method is that the telomeres un l ike beads are generally not spher i ca l i n shape. Another approach for va l ida t ion i s to use telomere objects of k n o w n lengths. P l a s m i d s (circular pieces of DNA) can be used for th is purpose . D N A w i t h telomere sequences ( T T A G G G repeats) of k n o w n lengths a n d u p to 3200 base-pai rs c a n be inserted into a p l a smid . A precaut ion of th is app roach is that these telomere lengths are m u c h smal ler t han those i n h u m a n chromosomes . A s a result , the fluorescence s ignals are weaker , harder to detect, a n d m a y not span over a s imi la r n u m b e r of focus planes . Another p recau t ion i s due to other var ia t ions i n the system (such as b io logica l a n d sample prepara t ion variations) w h i c h may interfere w i t h the accuracy of the va l ida t ion process. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 114 5.5.2. Spatial Resolution of IFI Segmentation Simula t ed test objects are used to estimate the spa t ia l reso lu t ion of u s i n g our a lgor i thm to separate two nearby telomeres (Figure 5.4). We use the theoret ical P S F of the system a n d convolve i t w i t h our s imula ted test objects to generate the images of the test objects. The test objects consis ts of two point sources w h i c h are incremental ly moved apart from one another i n steps cor responding to one p ixe l spac ing of the detector. O u r a lgor i thm i s then appl ied to the generated objects to determine how far the test objects have to be away from each other before they are treated as be ing two objects, It c a n be seen that one c a n j u s t v i sua l ly d i s t ingu i sh 2 points i n the s imula ted images w h e n the points sources are 4 pixels (the approximate reso lu t ion of our microscope a n d camera system) away from each other. A t 5 p ixels away from each other, the h u m a n eye and our a lgor i thm can bo th clearly separate the two objects. Th i s dis tance corresponds to approximate ly 0 .54um i n the object p lane at 6 3 x magnif icat ion. Th i s distance i s also twice the theoret ical reso lu t ion of j u s t the microscope. Thus , i t i s est imated that our segmentat ion a lgor i thm c a n separate objects that are 5 or more pixels (> 0.54um) apart . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 115 Figure 5.4. S imula t ed test objects for spat ia l resolu t ion estimates. The top row shows the character is t ics of the test objects. The middle row shows the ca lcu la ted in-focus image. The last row shows the borders of our telomere segmentat ion a lgor i thm. O u r a lgori thm separates the two point sources when they are at least 4 pixels apart. 5.5.3. S i m u l a t e d O b j e c t s To val idate our telomere IFI quantif icat ion a lgor i thm, we used different s imula ted test objects where the relative IFI value of each is k n o w n . The s imula ted objects have vary ing shapes and intensi ty d is t r ibut ions (to s imula te va ry ing shape a n d intensi ty of telomeres) but the same IFI value. These objects are used to test the robustness of our IFI a lgori thm to see i f s imi l a r IFI va lues are generated. The s imula ted objects consis t of i) a single point source whose d imens ions are 1 p ixe l (object #1 i n Figure 5.5), a n d ii) point sources whose d imens ions are greater t han 1 p ixe l i n the x direct ions (objects #2, #3, a n d #4 TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 116 i n Figure 5.5) a n d greater t han 1 p ixe l i n the x and y direct ions (object #5 i n F igure 5.5) a n d greater t h a n 1 p ixe l i n the z di rect ion (objects #6, #7, #8, a n d #9 i n Figure 5.5). S imula t ed images were generated by convolving test objects w i t h the theoret ical P S F of the system as descr ibed i n Chapter 5.3. W h i l e the spa t ia l d i s t r ibu t ion of each s imula ted object is different from the others, i ts IFI va lue i s the same as the others (i.e. the s u m of p ixe l in tensi ty va lues for each object i s the same as the s u m of p ixe l intensi ty va lues for other objects). The va lues a n d shapes of the s imula ted test objects (in the x -z plane) are s h o w n i n Figure 5.5. The resul ts of the IFI ca lcula t ions (using equat ion 5-9) at each differently focus image are shown i n Table 5 .1 . A plot of the IFI var ia t ions for some of these s imula ted objects as a funct ion of focus i s s h o w n i n F igure 5.6. The total IFI es t imat ion of each object i s then determined from the s u m of the IFI va lues over different focus images (sum of row values i n Table 5.1) a n d are s h o w n i n Table 5.2. F i rs t , the single point source image is chosen as a reference for compar ing IFI values . After normal iza t ion , the IFI of the in-focus image of th is single poin t object i s set to 100%. F r o m the resul ts (Table 5.1 a n d Figure 5.6), i t c a n be seen that the IFI value calcula ted (at the best focus plane) of each s imula t ed object c a n vary by 9% amongst objects. The var ia t ion of the IFI of a n object over ± 0 . 1 um from its best focus plane c a n be as h i g h as 10% (e.g. object #6). The var ia t ion are higher at higher de-focus amounts . A t ± 0 . 2 u m de-focus (a typ ica l z-distance for metaphase telomeres prepared on a slide), the ca lcu la ted IFI c a n be h igh (approximately 30%). F r o m the theory (equation 5.7), the IFI ca lcu la t ion at any focus plane {zl of z 2) s h o u l d be the same. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 117 However, due to the difficulty i n defining the exact border of the telomere objects i n different images as the objects become out-of-focus, the ca lcu la ted IFI va lues are different. Thus , the resul ts suggest that the focus plane from w h i c h the image of the object i s captured p lays a n impor tant role i n the accuracy of m e a s u r i n g the IFI va lue of the object. If the IFI value obtained at each focus plane (calculated u s i n g equat ion 5.9) i s s u m m e d to give the total IFI value for the object (e.g. c o l u m n 1 of Table 5.2), then the var ia t ion i n the total IFI value i s on ly 3%. T h u s , s u m m i n g i n d i v i d u a l IFI values ca lcula ted from a series of images cap tured at different focus planes generates more precise resul ts t h a n the IFI va lue ca lcula ted from a n image captured at only a single focus plane. Instead of s u m m i n g p lanar IFI values of images t aken at a l l 0.1 um z- spacings , the s u m m e d IFI values are ca lcula ted from images captured at larger sample spacings (e.g. 0.2, 0.3, and 0 .4um z-focus apart). A s less z-focus p lanes are used i n the summat ion , the resu l t ing IFI va lue w i l l be proport ionately smal ler (e.g. only ha l f of the O . l u m z-spacing images are u s e d i n the 0.2urn sample spac ing ca lcu la t ion and hence the generated IFI s u m is ha l f that of the O . l u m sample spac ing calculat ion). Hence, the s u m m e d va lues are mu l t i p l i ed by the s ampl ing frequency (e.g. 2, 3, a n d 4, respectively) to resul t i n a s imi l a r magni tude i n total IFI as that of the O . l u m spac ing . The resul ts of different var ia t ions of the s ampl ing j u s t descr ibed are s u m m a r i z e d i n Table 5.2. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 118 Object #1 1.00 Object #2 0.50 0.50 Object #3 0 .3333 0 .3334 0 .3333 Object #4 0.25 0.25 0 .25 0 .25 Object #5 0.13 0.25 0.06 0.22 0.15 0.19 Object #6 0.50 0.50 Object #7 0 .3333 0 .3334 0 .3333 Object #8 0.25 0.25 0.25 0 .25 Object #9 0.20 0.20 0.20 0.20 0.20 Figure 5.5. S imula t ed test object values and shapes. For a l l objects, the hor izonta l d i rect ion represent the extent of l uminance i n the x- direct ion. For object #5, the ver t ical direct ion represent the extent of l u m i n a n c e i n the y-direct ion. For objects #6, #7, #8, a n d #9, the ver t ica l d i rec t ion represent the extent of l uminance i n the z-direct ion. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 119 Obj. # Normalize Calculated IFI Percentage at Sub-Micron Z-Spacing -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 1 15 40 61 81 95 100 96 81 62 42 15 2 16 35 61 80 93 98 94 81 62 37 18 3 18 38 63 80 94 98 90 77 60 40 19 4 20 40 60 77 90 95 91 78 61 42 22 5 20 40 60 78 91 95 91 79 61 42 22 6 0 24 51 71 88 98 97 88 72 53 25 7 16 40 60 79 92 96 92 80 62 42 17 8 26 50 70 85 92 93 85 70 51 26 11 9 19 42 59 75 86 91 87 76 60 42 20 Table 5 .1 . IFI values of typ ica l h u m a n telomeres at different focus levels. The bo ld face va lues represent the highest IFI va lues over the range of focus levels for each telomere. -0.6 -0.4 -0.2 0 0.2 0.4 0.6 Z - f o c u s p l a n e ( m i c r o n s ) Figure 5.6. Normal ized IFI values at va ry ing focus of different objects. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 120 It c a n be seen that i f the IFI ca lcu la t ion were s u m m e d over every other 0.2 or 0.3 urn z-spaced images, the s u m m e d IFI value can vary by 5% amongst objects. The var ia t ion i n the s u m increased to 9% w h e n every other 0.4 z- spaced images are s u m m e d . Th i s suggests that images c a n be spaced at u p to 0.3 um apar t from one another wi thout significantly inc reas ing the error i n the ca lcu la t ion of the s u m m e d IFI value of the s imula ted object. Obj. # Normalized Calculated IFI Percentage Summed Over Different Z-Spacings (1:0) (2:0) (2:1) (3:0) (3:1) (3:2) (4:0) (4:1) (4:2) (4:3) 1 100 100 100 98 101 101 94 100 105 100 2 98 96 100 96 99 99 94 100 99 101 3 98 97 100 96 100 99 91 100 103 100 4 98 97 100 94 100 100 90 100 103 101 5 99 97 101 95 101 101 91 100 103 101 6 97 97 97 96 97 98 93 93 101 101 7 98 98 98 95 100 99 92 98 104 98 8 97 97 98 98 97 98 97 99 98 97 9 97 97 96 98 97 98 96 96 101 96 Table 5.2. IFI va lues of s imula ted test objects ca lcula ted at different focus level s a m p l i n g spacings . The s u m m e d value i s obtained by i n c l u d i n g images spaced at 1, 2, 3, or 4 (O.lum) z-focus spacings . The first n u m b e r i n parenthes is (in the c o l u m n heading) represents the z-focus spac ing used to generate the s u m m e d value. The second n u m b e r i n parenthesis represent different s tar t ing point or offset (for each focus spacing) i n w h i c h the s u m i s ca lcu la ted . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 121 5.5.4. Fluorescent Beads Images of different size fluorescence beads (0.1, 0.2, 0 .5, a n d 1.0 um) are acqu i red a n d the IFI value at each focus plane a n d the total IFI va lue of a l l focus p lanes are generated for each bead. If we assume that the fluorescence in tens i ty of the bead is related to the 3-d imens iona l size of the bead, then the IFI of the bead w o u l d be propor t ional to the cube of i ts 1-dimensional size (diameter). The resul ts of this experiment and the normal ized expected theoret ical IFI va lues are summar ized i n Table 5.3 a n d plotted i n Figure 5.7. Diameter of B e a d (um) Bes t Focus IFI (Mean 8B Standard Devia t ion Bes t Focus Normal ized IFI S u m Z- Planes IFI (Mean 85 S tandard Devia t ion S u m Z- Planes Normal ized IFI Theor. IFI 0.1 9.24 ± 2.64 1.00 ± 2 9 % 1 1 0 ± 3 1 1.00 ± 2 8 % 1 0.2 99.1 + 13.3 10.7 ± 1 3 % 968 ± 1 3 0 8.8 ± 1 3 % 8 0.5 1162 ± 3 7 126 + 3 % 12503 ± 530 113 + 4 % 125 1.0 9 3 4 4 ± 203 1011 ± 2 % 146189 ± 3224 1329 ± 2 % 1000 Table 5.3 IFI va lues of different size beads. It c a n be seen from the graph (Figure 5.7) that the ca lcu la ted IFI va lues for the beads correspond closely to the theoretical IFI values . It c a n also be TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 122 seen that the IFI value calcula ted at the best focus plane gives s l ight ly more accurate resul ts t han the s u m m e d value for a l l focus p lanes i n the cases of the larger size beads (0.5 and 1.0 um) The resul ts show that our IFI a lgor i thm gives a good estimate of the IFI of the bead. T5 10 00 T3 0) N "(5 E 10000 1000 100 10 • Best Focus Sum of Z-Planes Estimated Size 10 100 Estimated IFI (normalized to 0.1 micron bead) 1000 Figure 5.7. IFI d i s t r ibu t ion of different size beads. 5.5.5. Plasmids S i m i l a r to the analys is for the beads, images of different size telomeres w i t h i n p l a s m i d s (150, 400 , 800 , and 1600 base-pairs) are acqui red a n d the total IFI value is generated for each p l a smid . If we assume that the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 123 fluorescence intensi ty of the p l a s m i d is related to the n u m b e r of telomere base- pa i r s present (like our a s sumpt ion for telomeres i n chromosomes) , then the IFI of the p l a s m i d w o u l d be propor t ional to the n u m b e r of base-pairs i n the p l a s m i d . The resul ts of this experiment a n d the normal ized expected theoret ical IFI va lues are summar ized i n Table 5.4 a n d plotted i n Figure 5.8. C o l u m n 1 C o l u m n 2 C o l u m n 3 C o l u m n 4 C o l u m n 5 C o l u m n 6 Size of P l a s m i d (base pairs) Normal ized to 150 base pa i rs Best Focus IFI (Mean 8s S tandard Devia t ion Best Focus Normal ized IFI S u m Z - Planes IFI (Mean 8s S t anda rd Devia t ion S u m Z- Planes Normal ized IFI 150 1.0 4.17 ± 1.31 1.00 ± 3 1 % 4 8 . 0 ± 18 1.00 ± 3 7 % 400 2.7 11.2 ± 3.4 2.69 ± 2 9 % 119 ± 4 0 2.47 ± 3 3 % 800 5.3 24.3 ± 4.6 5.83 ± 19% 221 ± 4 1 4 .60 ± 19% 1600 10.7 44.7 ± 5.7 10.7 ± 1 3 % 464 ± 64 9.67 ± 1 3 % Table 5.4 IFI va lues of different size p lasmids . A g a i n , as w i t h the resul ts w i t h the beads, the p l a s m i d resul ts also show good correla t ion between the calcula ted a n d the expected theoret ical va lues (compare c o l u m n s 4 or 6 w i th c o l u m n 2 of Table 5.4). That is , our IFI a lgor i thm gives a good estimate of the length of telomeres. The best focus IFI va lues are better matched to the theoretical va lues t h a n the s u m m e d va lue for a l l focus planes . The var iance of the calcula ted va lues are h igher for the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 124 smal le r p l a smids . Th i s suggests that there are larger var ia t ion i n the size of the telomere repeat sequence i n smal ler p lasmids . o -I 1 1 1 1 1 1 1 1 0 200 400 600 800 1000 1200 1400 1600 Size of Plasmid (base pairs) Figure 5.8. IFI D i s t r i bu t ion of Different Size P lasmids . 5.5.6. Summary of Algorithm Validation Results The previous three methods used to evaluate the IFI value of the object showed that the IFI of the object shou ld be evaluated u s i n g images captured at different focus planes . The IFI value evaluated u s i n g the IFI value from the best focus plane image for the object p roduced s imi la r resul ts to the IFI value evaluated u s i n g the s u m of IFI values from images at different focus p lanes . For the s imula ted objects, the best focus method resul ts i n 3 % var i a t ion i n IFI va lue amongst different objects whi le the s u m m e d IFI va lues (up to sample spac ing of 0.3um) from mul t ip le focus plane method resul ts i n 5% var ia t ion . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 125 For the beads and p lasmids , bo th methods produce s imi la r resul ts . A s there are more var iab i l i ty i n the smal ler size beads and p l a smids compared to the larger objects, the var ia t ions i n their sizes are also higher. If only one image near the best focus plane for a l l objects i s used , the focus of i n d i v i d u a l objects i n the image can be ± 0 . 2 u m from the best focus pos i t ion for that object (as shown i n Figure 5.6 for the s imula ted objects, Table 5.5 for the beads a n d p lasmids , a n d later i n Table 5.6 for the telomeres). The IFI va lue evaluated from a single image plane cou ld then vary by 2 0 % from the IFI va lue est imated from the best focus image. Object Z-focus position (microns) -0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.1 (im bead 0.77 0.85 0.98 0.98 0.98 1.00 0.89 0.85 0.79 0.53 0.43 0.2\m\ bead 0.33 0.52 0.69 0.84 0.96 1.00 0.95 0.87 0.76 0.60 0.48 0.5(xm bead 0.42 0.57 0.72 0.85 0.93 1.00 0.97 0.89 0.80 0.69 0.57 1 .Oum bead 0.69 0.79 0.87 0.93 0.98 1.00 0.99 0.95 0.88 0.78 0.66 1 50bp plasmid 0.48 0.64 0.66 0.68 0.85 1.00 0.82 0.70 0.61 0.59 0.54 400bp plasmid 0.67 0.77 0.86 0.97 0.98 1.00 0.95 0.76 0.66 0.48 0.35 800bp plasmid 0.48 0.62 0.78 0.91 0.97 1.00 0.89 0.67 0.57 0.47 0.33 1600bp plasmid 0.63 0.70 0.83 0.88 0.94 1.00 0.93 0.84 0.59 0.41 0.27 Table 5.5 Normal ized IFI values of objects at different focus pos i t ions . These different methods of IFI ca lcu la t ion is next app l ied to telomere images to see i f s imi la r resul ts are obtained. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 126 5.6. Human Telomeres Results 5.6.1. Number of Focus Planes Required A s imi l a r type of analys is as that performed on the s imula ted objects i s made on the telomere images (using equat ion 5-9 for each focus plane). M u l t i - focus p lane images of telomeres are acquired at O . l u m z-spacing. There were 125 telomeres analyzed from the images. The IFI for each focus p lane for each telomere is generated. The resul ts of some of the 125 telomere IFI's are s h o w n i n Table 5.6. T e l o m e r e N u m b e r Z - P o s i t i o n ( m i c r o n s ) - 0 . 9 - 0 . 7 - 0 . 5 - 0 . 3 - 0 . 1 0 . 1 0 . 3 0 . 5 0 . 7 0 . 9 1 63.3 127.3 212.1 324.8 3 8 3 . 6 381.2 307 .1 235.2 158.5 97.4 2 3.0 21.8 42.5 79.4 111.2 1 2 5 . 2 110.5 93.8 70.0 42.7 3 7.7 35.3 73.7 139.6 202.0 2 3 8 . 3 231.7 201.3 156.0 104.9 4 28.8 67.0 113.3 196.7 261.9 2 8 9 . 6 267.7 229.0 184.4 120.1 5 0.4 18.9 41.6 90.2 149.4 183.5 1 9 5 . 6 170.0 136.5 95.5 6 19.2 53.2 97.8 162.6 234.7 2 5 6 . 5 246.0 213.4 170.4 113.0 7 23.6 63 .1 107.4 183.1 254.6 2 7 9 . 0 231.8 183.3 134.9 71.4 8 16.8 69.5 137.3 239.2 337.3 3 7 4 . 2 331.4 267.4 197.9 122.4 9 14.1 40.1 79.5 139.9 220.6 266.8 2 6 7 . 9 236.0 193.0 135.2 1 0 51.3 116.6 195.3 301.0 3 5 0 . 6 334.9 270.8 187.2 118.9 65.6 Table 5.6. IFI va lues of typ ica l h u m a n telomeres at different focus levels. The bo ld face va lues represent the highest IFI va lues over the range of focus levels for each telomere. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 127 It c a n be seen i n Table 5.6 that the pos i t ion of the in-focus p lane (highest IFI value) for each telomere i n a metaphase chromosome sample var ies by at least 0.4u,m. In addi t ion , there i s at least a three fold difference between the lowest a n d highest value amongst the telomeres i n the example. -1 -0.5 0 0.5 1 Distance from In-Focus Position (microns) Figure 5.9. IFI of typ ica l h u m a n telomeres at different focus levels. The IFI va lues at different focus planes are then normal ized so that they c a n be compared w i t h those of other telomeres. To normal ize the IFI values , the highest IFI value for each telomere is set to 1.00. The IFI va lues at other focus p lanes are then divided by the highest IFI value for that telomere to give TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 128 the normal i zed resul ts . A plot of the IFI var ia t ions for the average of 125 normal i zed telomere IFI values as a funct ion of focus i s s h o w n i n Figure 5.9. It c a n be seen from Figure 5.9 that i f a single focus plane i s u s e d to represent the IFI value of the telomere, there c a n be at least 10% error i n the ca lcu la ted IFI i f the telomere is 0 .2um out-of-focus. That i s , i f a single image is u s e d to represent the best focus image for a l l telomeres, there c a n be some telomeres w h i c h are 0 .2um out-of-focus (Table 5.6). Hence, the ca lcula ted telomere IFI value cou ld be 10% less t han their in-focus value . S i m i l a r to the s imula ted objects, the total IFI i s ca lcu la ted as the s u m of va lues at three differently spaced z-intervals (0.2, 0 .3 , a n d 0.4um). The resul ts of different var ia t ions of the sampl ing for the telomeres are s u m m a r i z e d i n Table 5.7. It c a n be seen from Table 5.7 that the s tandard deviat ion of the total IFI va lues for the 125 telomeres i s approximately 18% of the average IFI va lue . T h i s impl ie s that i f the total IFI is not ca lcula ted over mul t ip le focus p lanes , there c a n be 18% difference between the mul t i - focus plane IFI va lue a n d the best focus IFI value . The difference w o u l d be higher i f the compar i son is made w i t h the IFI value calcula ted from the image where the telomere is 0 .2um out- of-focus (i.e. i f only one image plane is used i n the ana lys i s as i n most quanti tat ive cytometry studies). It c an also be seen that s u m m i n g images at every 0.2 or 0 .3um z-spacing produce s imi la r resul ts w i t h error of approximate ly 1.5% of the total IFI value. The error more t h a n doubles to 3 .8% w h e n the s u m m i n g step size is increased to 0 .4um. Hence, a 0 .3um step TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MCROSCOPY 129 size as predicted i n Sect ion 5.4 is u sed for acqu i r ing mul t i -p lane telomere images. The total IFI values for each telomere can then be obtained from these images. Alternat ively, as shown i n the 3 IFI va l ida t ion experiments , the best focus image (selected as being the focus plane that gives the highest IFI value for that object over O . l u m z-spaced images) cou ld be u s e d to calcula te the representative IFI of the object. Z - S p a c i n g (um) Offset (um) Ave rage IFI S t a n d a r d D e v i a t i o n A v e r a g e C V 0.1 0.0 10.863 1.802 0 . 0 0 0 0 0.2 0.0 10.856 1.802 0 . 0 1 4 8 ± 0 . 0 0 8 6 0.2 0.1 10.867 1.812 0.3 0.0 10.843 1.816 0 . 0 1 5 7 ± 0 . 0 0 8 8 0.3 0.1 10.875 1.802 0.3 0.2 10.870 1.810 0.4 0.0 10.992 1.887 0 . 0 3 8 2 ± 0 . 0 1 1 1 0.4 0.1 10.894 1.858 0.4 0.2 10.720 1.774 0.4 0.3 10.840 1.834 Table 5.7. IFI va lues of h u m a n telomeres ca lcula ted at different focus level s a m p l i n g spacings . For each z-spacing, there can be a n u m b e r of different s tar t ing poin ts or offsets from w h i c h the s u m i s generated. The s u m m e d IFI value i s generated by s u m m i n g the IFI values from images at different z- spac ings for each telomere. The average of 125 s u m m e d IFI va lues i s ca lcu la ted . For each z-spacing category, the average coefficient of va r i a t ion (CV) i s ca lcu la ted (i.e. C V is the difference between the s t andard devia t ion a n d i ts m e a n s tandard deviat ion determined from w i t h i n the z -spac ing group d iv ided by the m e a n s tandard deviation). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 130 5.6.2. Telomere Distribution in Cells The d i s t r ibu t ion of telomere IFI values i n a cel l i s s h o w n i n Figure 5.10. It c a n be seen that the plot is asymmetr ic about i ts peak a n d resembles a Po i s son d i s t r ibu t ion rather t h a n a G a u s s i a n d i s t r ibu t ion . A n explana t ion for th i s is that there are var ia t ions i n telomere lengths i n a given cel l as s h o w n prev ious ly u s i n g the Sou thern analys is (Allshire et a l . 1988). E v e n i f a l l telomeres i n the cel l have the same telomere length to begin wi th , after a n u m b e r of cel l d iv i s ion , the majority of the telomere lengths w i l l be at a r o u n d the same value . There are however, others w h i c h do not have a s imi l a r rate of reduc t ion i n telomere lengths. Thus , a spread of long telomeres c a n be observed. The n u m b e r of very short telomeres compared to the majori ty of telomeres i n the cel l i s s m a l l because the cel l has reached its c r i t i ca l state a n d tends not to divide. If the n o r m a l cel l happens to divide, on ly those cel ls w h i c h have longer telomeres w i l l tend to survive. B y ana lyz ing less t h a n 30 metaphases samples , a s ta t is t ical interpretat ion of the IFI d i s t r ibu t ion of the cel l popu la t ion c a n be obta ined. Previous ly , u s i n g the Southern analys is , approximate ly 100,000 cells were requi red to give s imi l a r resul ts as ours . In addi t ion , our IFI ana lys i s c a n give in format ion of the IFI d i s t r ibu t ion w i t h i n each i n d i v i d u a l cells . If ka ryo typ ing (the ident i f ica t ion/c lass i f ica t ion of each chromosome w i t h i n a cell) i s performed on the chromosomes , the IFI of each of the 24 different types of chromosomes i n the cel l and i n a popula t ion of cells can be obtained. We have performed TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 131 these s tudies on mouse telomeres (Zijlmans et a l . ; 1997) and the p - a r m of chromosome 17 (Martens et a l ; 1997). 35 cu 30 u 2 25 l 20 1 15 c CU 3 O" cu LL 10 0 r- O O co C Ĵ CO CM CO CO CsJ co CO LO CO CO to r--- CO CO C O CO Lf) CD Lf) CD IFI Va lue (grey levels) Figure 5.10. Telomere IFI d i s t r ibu t ion i n a cel l . 5.7. Chapter Summary In th is Chapter , we descr ibed what the IFI of a n object is a n d h o w it c an be theoret ical ly evaluated. We have shown that a l though objects occupy a 3 D space, only one image obtained at the focus plane is sufficient to determine the IFI of the object. We next descr ibed the difficulty i n ca lcu la t ing the IFI va lue as telomeres are i n close v ic in i ty of one another and hence their s igna l intensi t ies overlap. The errors i n IFI ca lcu la t ion as a resul t of segmentat ion are then d i scussed . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 132 We then in t roduced our telomere segmentation a lgor i thm. Th i s a lgor i thm is capable of separat ing images of point sources whose locat ions are 4 p ixels apart from one another. O u r a lgor i thm also segments objects over no i sy a n d va ry ing (not flat or s imi la r i n intensity) b a c k g r o u n d intensi t ies . Three methods of evaluat ing the IFI value of a n object are in t roduced . The first ca lcula tes the IFI value of every objects w i t h i n the image u s i n g the same focus plane image. The latter two methods use images captured from mul t i - focus planes . One method selects the best focus (highest IFI value) image for every telomere to evaluate the IFI of the object. The other method evaluates the IFI value of the object as the s u m of the cor responding IFI va lues for the object determined from images at equally spaced mul t i - focus p lanes . We used s imula ted objects, beads, a n d p l a smids to evaluate a n d val idate our IFI segmentation and quantif icat ion methods. We then appl ied our IFI a lgor i thm on telomeres and compared the resul ts . We have s h o w n that our IFI quant i f ica t ion a lgor i thm can be used to estimate the lengths of telomeres (verified by the beads and p l a s m i d experiments). We have also s h o w n that images from more t h a n one focus plane are required to reduce the errors i n the IFI ca lcu la t ion . The best focus IFI value or the s u m of IFI va lues evaluated from mul t i - focus plane images give s imi la r acceptable resul ts . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 133 Chapter 6. Segmentation of Chromosomes 6.1. Overview Thi s chapter describes how we segment fluorescence microscopy images of metaphase (chromosomes w h i c h have dupl ica ted bu t have not separated i n a d iv id ing cell) chromosomes. The objective here is to determine the regions occup ied by each chromosome i n the image. B y segmenting the chromosomes , the telomere IFI values (obtained from Chapter 5) c a n be l i n k e d to specific segmented chromosome objects. The user can then classify the chromosome type a n d the length of the telomeres of each chromosome c a n be obta ined. The var iab i l i ty i n the chromosome texture (intensity) w i t h i n i n d i v i d u a l chromosomes a n d amongst different chromosomes make i t difficult to f ind the exact borders of each chromosome. In addi t ion , the h igh noise levels associated w i th low light level fluorescence images pose another difficulty for segmentat ion. Yet another segmentation difficulty l ies i n def ining the boundar ies of touch ing and overlapping chromosomes. A l t h o u g h one c a n select metaphases where a l l the chromosomes are isolated from one another, s u c h images are rare to f ind. Hence, one typical ly scans a sl ide to f ind a metaphase where most of the chromosomes are isolated from one another a n d some of w h i c h are touch ing or overlapping. Images of these are then acqui red a n d used for our analys is . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 134 We have appl ied several k n o w n edge detection filters to segment chromosomes . The edge detection filters tr ied i nc luded the t rad i t iona l difference of G a u s s i a n s (DoG) filter (Marr 1982) and the C a n n y filter (1983). These edge detectors generate incomplete or d i scon t inuous edge boundar ies as we l l as edges w i t h i n and outside the chromosome region. Fur thermore , add i t iona l process ing (which c a n be quite computa t iona l ly involved) are requi red to select and fine tune the edge pixels to form a con t inuous edge a r o u n d each object. Thus , we developed a method w h i c h remedies the b o u n d a r y d iscont inu i ty problem a n d give the required chromosome boundar ies . O u r method employs only integer operations a n d hence i s faster to compute . Th i s method involves a novel filter ca l led the R a n k Difference filter w h i c h c a n be used as a n edge detector or a morphologica l filter. D u e to the difficulties i n chromosome segmentat ion ment ioned above, there are no single or combina t ion of segmentat ion techniques w h i c h c a n correct ly segment a l l chromosomes. J i (1994) developed a method to segment chromosome images from brightfield microscopy (which has better contrast a n d less noise t h a n images from fluorescence microscopy, the mic roscopy mode u s e d i n our study). H i s method uses a n iterative ru le-based app roach to ob ta in the required n u m b e r of segmented chromosomes per image. We d i d not use J i ' s method because it c an take a long time to compute (as m a n y i terat ions m a y be required) a n d h i s method is not designed for chromosomes w h i c h has been s ta ined to highl ight their band ing s t ructures (such as the fluorescence D A P I s ta in that i s u sed i n our study). We require th is b a n d i n g s t ructure informat ion for ka ryo typ ing (identify the type of chromosome i n a cell) s u c h TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 135 that telomere length informat ion of a par t icu la r chromosome i n a ce l l c a n be obta ined. C o m m e r c i a l chromosome analys is systems (Applied Imaging Inc., B io log ica l Detect ion Systems Inc., V y s i s Inc., and Oncor Ins t rument Systems) tend to use a s imple semi-automated segmentation a lgor i thm to generate a n in i t i a l estimate of the chromosome borders. The a lgor i thms typ ica l ly u s e d for th i s step consis ts of first interactively or automat ica l ly th reshold ing the image a n d then extract ing the chromosomes borders from the thresholded image. Once the in i t i a l borders are found, researchers then m a n u a l l y verify a n d correct the segmentat ion results . B y u s i n g our method, mos t of the chromosomes are correctly segmented. Hence, less user in teract ion i s requi red i n the m a n u a l verif icat ion process resul t ing i n a less tedious a n d a more economica l overal l interactive analys is . O u r chromosome segmentation a lgor i thm consis ts of a combina t ion of different segmentat ion methods since a single technique does not p roduce good resul ts . E a c h method or step i n the sequence improves on the resul ts obta ined by the previous step. Threshold ing is first u sed to define the first app rox ima t ion of the regions occupied by chromosomes. Texture informat ion i n the segmented region i s then used to generate the second approx ima t ion of the chromosome region. In this step, we first detect the loca l h igh in tens i ty p ixels . We then use our R a n k Difference filter as a morpholog ica l operator to merge detected pixels into different chromosome regions a n d at the same t ime separate touch ing chromosomes. S tandard d i la t ion a n d erosion morpholog ica l filters were not u sed i n th is step because they do not perform as we l l as our TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 136 filter for bo th merging and separat ing the different chromosome regions. The f inal step first f inds the edges of the chromosome regions u s i n g our R a n k Difference filter again, bu t this t ime i t functions as a n edge detector. O u r R a n k Difference filter is s imple , employing only integer operations, to implement a n d generates resul ts comparable to or better t han the more complex edge detectors (difference of G a u s s i a n and Canny) . The edges are then used to refine the borders for each detected chromosome region. Feature va lues are extracted from each chromosome region and these extracted va lues are used to reject objects whose s igna l intensit ies are too weak a n d whose sizes are too s m a l l to be chromosomes . F ina l ly , the segmented telomere image i s u sed to conf i rm that the segmented regions are indeed chromosomes wi th telomeres located at their ends. A s the R a n k Difference filter is used i n two stages of our segmentat ion a lgor i thm, we first describe the formulat ion and usefulness of th is filter (Section 6.2). We then compare the edge detection propert ies of our R a n k Difference filter w i th those of t radi t ional edge detectors (Section 6.3). F ina l ly , we describe the details of our chromosome segmentat ion a lgor i thm as ou t l ined above (Section 6.4-6.8). 6.2. Rank Difference Filter O u r R a n k Difference filter is formed from a difference of two r a n k filters. R a n k filters are widely used i n image processing. A popu la r a n d c o m m o n l y u s e d r a n k filter, the m e d i a n filter, removes r a n d o m noise a n d preserves edges i n images (Huang, 1979, R u s s and Russ , 1986). Another commonly u s e d r a n k TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 137 filter, the m a x i m u m or m i n i m u m r a n k filter, i s u sed to ob ta in smooth b a c k g r o u n d regions (Bright and Steel, 1986). B y c o m b i n i n g the m a x i m u m a n d m i n i m u m r a n k filtered images and ca lcu la t ing the difference, another image con ta in ing the m a x i m u m loca l contrast i n the or ig ina l image (Russ, 1990) i s generated. F r o m th is latter filter, we obta in the R a n k Difference filter b y extending the choice of r a n k filters i n the difference to inc lude those i n between the m i n i m u m a n d m a x i m u m ranked values . A s a resul t (as s h o w n later i n th is section), the choice of r a n k filters to use i n the difference determines the filter's tolerance to r a n d o m noise i n the image. It i s also s h o w n later i n th is sect ion that our R a n k Difference filter can be used as a selective morphologic filter. Th i s filter has properties w h i c h out-performs those of other filters for use i n chromosome segmentation. In formula t ing the R a n k Difference filter, we first define a region S(i(x,y)) i n the image i(x,y) conta in ing a total of v p ixels and a p ixe l (x,y) w h i c h m a y or m a y not be ins ide S(i(x,y)). The magni tudes of the p ixels i n S(i(x,y)) are ordered s u c h that the smallest value is denoted by R^Sfifcy))] a n d the largest va lue is denoted by R^Sfifcy))]. The r a n k filter R^S(i(x,y))] i s the filter whose output is the ith smal les t value of the pixels i n S(i(x,y)). O u r R a n k Difference filter Ru [S(i(x,y))\ s imply combines the difference of two r a n k filters, a n uppe r r a n k Ru[S(i(x,y))] and a lower one R^Sfayfl, into one filter as follows: K,,[S(i(x,y))} = Ru[S(i(x,y))} - RlSfifcy))] where 1<l< u< v (6-1) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 138 The output value of the R a n k Difference filter is t hus the difference between the uth smal les t value and the Ith smal les t va lue of the p ixe ls i n S(i(x,y)). Th i s value i s assigned to the corresponding p ixe l (x,y) i n the filtered image. The R a n k Difference filter image i s then generated by performing the above operat ion on a l l p ixels i n the image. The behaviour of th is filter i s governed by a n u m b e r of parameters: i) the loca t ion of p ixe l (x,y) i n the region S(i(x,y)), ii) the shape a n d size of S(i(x,y)), a n d iii) the va lues of the upper and lower r a n k number s u a n d I. F i rs t , the loca t ion of p ixe l (x,y) w i th respect to the region S(i(x,y)) determines the loca t ion of the resu l t ing edge i n the generated edge image w i th respect to those i n the or ig ina l image. Th i s phenomena is i l lus t ra ted i n Figure 6 .1 . P ixe l (x,y) c a n be located anywhere w i t h i n or outside S(i(x,y)). Th i s filter treats a l l locat ions i n S(i(x,y)) equal ly s ince only the magni tudes of the intensi t ies i n S(i(x,y)) determine the outcome. For the generated edges to be we l l registered w i t h respect to the or ig inal image, (x,y) shou ld be located as close to the center of S(i(x,y)) as possible (Figure 6.1b). In this case, the magni tudes of the p ixe ls i n the v ic in i ty su r round ing the p ixe l (x,y) determines the fate of that p i x e l a n d hence, there is no p ixe l shift i n the resul t ing edge. If the p ixe l (x,y) i s away from the centra l point of the region S(i(x,y)) (i.e. outside the region), then a n image shift occurs i n the resul t (Figure 6.1c). In this ins tance, p ixe l (x,y) is replaced by the resul t of the R a n k difference operat ion ca lcula ted over a region w h i c h does not inc lude p ixe l (x,y). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 139 Figure 6 .1 . Cen t ra l p ixe l of r a n k difference filter. Second, the shape of region S(i(x,y)) determines w h i c h spat ia l or ientat ion of edges are emphas ized more than others. Th i s region can assume any shape. W i t h a square shaped region, the filter emphasizes those edges w h i c h are at a n angle more than those edges w h i c h are perpendicular or para l le l to the square edges. Th i s phenomena is i l lus t ra ted i n Figure 6.2 where c i r cu la r and rec tangular test objects located at different angles from the hor izonta l ax is are used . Hence, a square shaped region wou ld be useful for detecting objects i n the image w h i c h lie i n one general orientat ion (Figure 6.2b). W i t h a c i r cu la r shaped region, edges i n a l l d i rect ion w i l l be equal ly emphasized (Figure 6.2c). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 140 That i s the detected edges of the circle and diagonal l ines have s imi la r widths . S ince the chromosomes i n our analys is can lie i n any orientat ion, a c i rcu la r region is preferred. (a) (b) (c) Figure 6.2. Shape of r a n k difference filter region. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 141 Figure 6.3. Size of r a n k difference filter region. The or ig inal image i s i n row (a). Operat ions w i t h the 3 x 3 , 5x5 , and 7x7 filter regions are shown i n rows (b- d), respectively. C o l u m n (i) shows the filter region. C o l u m n (ii) i s the or ig ina l image. C o l u m n (iii) a n d (iv) have added uni form noise of s tandard deviat ion of 0.5 and 5.0 added, respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 142 The size of the region s(x,y) is found to determine the spa t ia l reso lu t ion of the deta i ls /edges i n the image. For example, i n a 3x3 square p ixe l region, edges w h i c h s p a n over 3 or less pixels wide can be detected. S m a l l region sizes are preferred over larger ones because they require less t ime to process. W i t h a smal le r region size, however, i t i s more difficult to obta in a c i r cu la r l ook ing region i n a rectangular p ixe l gr id . For example, i n a 3x3 p ixe l rec tangular area, a l l 9 p ixe ls i n the square region w o u l d be used to approximate a c i r cu l a r shape. Hence, there is a compromise as to w h i c h filter size s h o u l d be u s e d i n a given appl ica t ion . Figure 6.3 shows the resul ts of app ly ing different size filter regions to the test image. W i t h noise free images as s h o w n i n Figure 6 .3 i i , the edges i n the resu l t ing image tend to be th ick and s imi la r i n w i d t h to that of the filter's size. W i t h no isy images, the larger size filters tends to do a better job of select ing the edges (Figure 6 .3 i i i a n d iv). For example, even w h e n un i fo rm noise of s t andard deviat ion of 5.0 i s added to the test image s u c h that the objects cannot be v i sua l ly identify amongst the noise i n the image, our 7x7 R a n k Difference filter is able to identify the edges (Figure 6.3d,iv) The upper and lower r a n k numbers , u a n d I, determines the filter's tolerance to noise i n the image and the th ickness of the edge i n the resu l t ing image. B y choos ing the highest n u m b e r (v) for the upper r a n k a n d the lowest n u m b e r (1) for the lower r a n k number , the filter generates a h image cons i s t ing of the largest l o c a l intensi ty differences (Figure 6.4). T h i s spec ia l case of ou r R a n k Difference filter becomes the filter used by R u s s (1990). A s the upper a n d lower r a n k number s are moved away from their extreme values , the edges generated by the filter are th inner and less intense (Figure 6.4c). There i s a po in t w h e n the edges become too t h in that they d isappear resu l t ing i n a TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 143 d i scon t inuous border a round the object (Figure 6.4d). If the m a x i m u m and middle r a n k number s are used as the upper and lower r a n k number s i n the R a n k Difference filter, the outer edge of the object (increased by ha l f the w id th of the filter) resul ts (Figure 6.4e). S imi la r ly , i f the middle a n d lowest r a n k n u m b e r s are used , the inner edge of the object resul ts (Figure 6.4f). Th i s is t rue even i n the presence of noise as shown i n Figure 6.5. Figure 6.4. Effect of va ry ing upper and lower r a n k numbers . The or ig ina l image i s s h o w n i n (a). A c i rcu la r 7x7 R a n k Difference is appl ied (as s h o w n i n Figure 6.3di). The m a x i m u m r a n k of 37 and m i n i m u m r a n k of 1 i s u sed i n (b). The upper and lower r a n k number s are (29,9), (22, 16), (37,30) a n d (8,1) for (c), (d), (e), and (f), respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 144 Figure 6.5. Effect of addit ive noise and vary ing upper a n d lower r a n k numbers . Un i fo rm noise of s tandard deviat ion of 1.0 i s added to the or ig ina l image and the resul t i s shown i n (a). A 3x3 R a n k Difference i s appl ied . The m a x i m u m rank of 9 and m i n i m u m r a n k of 1 is used i n (b). The upper a n d lower r a n k numbers are (7,3), (6,4), (9,5) and (5,1) for (c), (d), (e), a n d (f), respectively. 6.3. Comparison of Edge Detectors In th is sect ion, the performance of our R a n k Difference filter i s compared w i th other edge detectors: the difference or L a p l a c i a n of G a u s s i a n s a n d the Canny . A variety of different test images are used i n the compar i son . The first test image consis ts of rectangular and c i rcu la r shaped objects. A s TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 145 expla ined i n sect ion 6.2 (Figure 6.2), the rectangular objects are oriented at va ry ing angles from the hor izonta l axis of the p ixe l gr id s u c h that the filter's performance to edge orientat ion can be tested. The resul ts of the edge detection filters are s h o w n i n Figure 6.6. It c a n be seen that the D o G filter performs the best i n defining the edges of the test object. O u r R a n k Difference filter performs s imi la r ly w i th edges w h i c h are s l ight ly wider (2 to 3 p ixe ls i n width) t h a n that of the D o G result . The C a n n y filter d i d not perform as we l l as the others i n defining the corners of the square a n d rectangular objects. i \ ^ i—i i' \^ r P ^ ! ^ , - 1 j • • v _ y i 5 Figure 6.6. Performance of edge filters on test object. The or ig inal image is s h o w n i n (a). The resul ts of the 3x3 m e d i a n a n d then R a n k Difference filter, R(8,2), i s shown i n (b). The resul ts of the D o G (a = 16) a n d C a n n y (a = 3) filters are shown i n (c) and (d), respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 146 Figure 6.7. Performance of edge detectors i n presence of noise added to the test object. The top row represent the or ig inal test objects. F r o m left to right, the images contains additive G a u s s i a n noise of 0.5 a n d 1.0 s tandard deviat ion, and additive un i form noise of 1.0 a n d 7.0, respectively. The second row is the results of app ly ing the m e d i a n a n d ou r R a n k Difference filter (7x7 c i rcu la r R(31,7), 7x7 c i r cu la r R(31,7), 3x3 square R(9 , l ) , 7x7 c i rcu la r (37,1). The th i rd and fourth rows are the resul ts of the D o G (a = 16, 16, 14, 16) and C a n n y (a = 3, 4, 4, 5) filters, respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 147 R a n d o m noise is then added to this image to test the filter's tolerance level to noise. The resul ts are shown i n Figure 6.7. To filter some of the noise i n the images, we first appl ied a s imi la r size med ian filter before we appl ied our R a n k Difference filter. The use of the med ian filter s ignif icant ly reduce the n u m b e r of false edges in t roduce by ou r filter. F r o m the resul ts s h o w n i n F igure 6.7, i t c a n be seen that the D o G filter performs the worst i n defining the edges of objects. The C a n n y filter perform the best on images w i t h addit ive G a u s s i a n noise . O n the other h a n d , our R a n k Difference filter performs the best i n s i tua t ions where un i fo rm noise i s added to the or ig ina l image. The corners of the objects are generally more preserved u s i n g our filter. O u r filter was also able to detect most of the edges i n the noisy un i fo rm image where i t i s even difficult for the eye to d i s t ingu i sh a l l the edges (Figure 6.7, top 2 , r ight images). Images of rea l objects are then used to evaluate the performance of the edge filters. Fi rs t , a n image of peppers is used to compare the va r ious edge filters (Figure 6.8). The D o G filter again has the worst performance. The loca t ion of the edges do not exactly correspond to the loca t ion of the edges i n the image bu t are i n the v ic in i ty of the true edges. The C a n n y filter performs the best. Sha rp , single p ixe l wide borders i n the resu l t ing image cor respond closely to the true edges of the image. O u r R a n k Difference filter performs a lmost as good as the C a n n y filter. The edges are wider a n d more fuzzy a n d they also m a t c h the locat ion of the true edges i n the or ig ina l image. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 148 Figure 6.8. Performance of edge detectors on the peppers image. The or ig ina l image is s h o w n i n (a). The results of our 3x3 R a n k Difference (R(9,l)), the D o G (CT =20), and the C a n n y (a = 1) filters are shown i n (b), (c), and (d), respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 149 The effects of the filters performance to addit ive G a u s s i a n noise are then examined (Figure 6.9). A degradation i n defining most of the edges a n d their t rue locat ions i s evident compared to the filters' performance on the or ig ina l image. More false edges (as a resul t of the added noise) are also detected. Th i s i s especial ly true for our R a n k Difference filter. A g a i n , the C a n n y filter perform the best i n defining the major edges of the image. O u r R a n k Difference filter, however, does a better job of defining the true locat ion of the borders as seen i n the border between the long pepper and the l ight coloured pepper. It i s also observed that more noise i n the image typical ly requires more pre-fi l tering. That i s the noise parameter needs to be increased i n the D o G a n d C a n n y filters a n d the ke rne l size needs to be increased i n our R a n k Difference filter to ob ta in favourable segmentat ion results . The effects i n the performance of the filters to addit ive un i fo rm noise i s also examined (Figure 6.10). A s imi la r resul t is seen as that for the case of addit ive G a u s s i a n noise. The exception is that our R a n k Difference filter performs better w i t h additive un i form noise than G a u s s i a n noise. In th is ins tance, more p ronounced edge pixels correspond closely to the exact pos i t ions of the edges i n the or ig ina l image. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 150 Figure 6.9. Edge filter performance of peppers image w i t h additive G a u s s i a n noise. G a u s s i a n noise of s tandard deviat ion of 0.5 is added to the or ig ina l image to resul t i n (a). The resul ts of the med ian a n d our 7x7 c i rcu la r R a n k Difference (R(31,7)), the D o G (a =24), and the C a n n y [a = 3) filters are shown i n (b), (c), and (d), respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 151 Figure 6.10. Edge filter performance of peppers image w i t h addit ive un i form noise. Un i fo rm noise of s tandard deviat ion of 3 is added to the or ig inal image to resul t i n (a). The resul ts of the m e d i a n a n d our 7x7 c i rcu la r R a n k Difference (R(37,l)), the D o G (a =25), a n d the C a n n y (a = 6) filters are shown i n (b), (c), and (d), respectively. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 152 A n image of metaphase chromosomes i s then used to compare the edge filters (Figure 6.11). A l so shown i n this figure are the filters performance on the G a u s s i a n a n d un i fo rm noise added to the or ig ina l image. L i k e the other examples , the C a n n y filter performs the best overal l i n terms of detect ing the chromosome edges. O u r R a n k Difference filter gives th icker boundar i e s a n d also has better local iza t ion of the true edges of the chromosomes i n the image. O u r R a n k Difference filter also perform wel l on images w i t h addit ive un i fo rm noise a n d not as we l l on images w i th additive G a u s s i a n noise. W i t h a l l three filters, as more noise i s added to the image, more process ing t ime i s required. Larger filter sizes are generally required to smooth out the noise i n the image. A s the filter size increases, the process ing t ime also increases i n a re la t ionship of approximately the square of the size of the filter. The operat ions i n the R a n k Difference filter are based m a i n l y on the sor t ing of integers. O n the other h a n d , bo th the D o G a n d C a n n y filter u t i l izes f loating poin t ar i thmet ic w h i c h takes m u c h longer t ime to process. A l t h o u g h the C a n n y filter generates the best edges of the three edge detectors t r ied, the resul ts are not good enough for our purpose as m a n y touch ing chromosomes are not properly segmented a n d m a n y edges are not connected (discont inuous object boundaries) . Thus , we have to develop a better technique w h i c h overcomes these problems. O u r technique conta ins a n u m b e r of segmentat ion steps and inc ludes the use of our R a n k Difference filter i n two of these steps. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 153 Figure 6 .11 . Edge filter performance on chromosome image. The or ig ina l images are shown i n the top left. The top r ight image have G a u s s i a n noise w i th a s tandard deviat ion of 0.3 added. The second row are the cor responding resul ts of the R a n k Difference filter (3x3 square R(7,3) a n d 7x7 c i r cu la r R(31,7)). The th i rd row are the corresponding resul ts of the D o G filter (a =5 and 11). F ina l ly , the last row are the resul ts of the C a n n y filter (a =1 a n d 3). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 154 6.4. First Approximation to Edges: Thresholding In th is and the following two sections, we describe the details of each of the three steps of our segmentation a lgor i thm. The first step i n our segmentat ion a lgor i thm generates a n approximate chromosome region or m a s k by de termining w h i c h pixels belong to the background . These p ixe ls are then exc luded from further analys is . Threshold ing i s chosen for th is step because it i s s imple to implement . Threshold ing performs wel l on most isolated chromosomes bu t behaves poorly i n segmenting touch ing a n d a lmost t ouch ing chromosomes (Figure 6.12f,g,h). The chromosome regions are faded into the b a c k g r o u n d as the threshold level i s increased. It c a n be seen that no one th reshold c a n be used to segment a l l chromosomes, especial ly those w h i c h are touch ing . E v e n adaptive threshold ing does not separate nearby chromosomes as the in tensi ty levels of the over lapping and touch ing pixels are m u c h s imi l a r i n va lues to those p ixels w i t h i n the chromosomes. Th i s s imi la r i ty i n in tens i ty levels c a n be seen i n Figure 6.12 where the or ig ina l image is thresholded to generate thresholded images at var ious intensi ty levels. A s seen i n the thresholded images, no single threshold can be u s e d to separate a l l chromosomes . A s we are only interested i n a n approx ima t ion of the chromosome region i n th is step, we chose a conservative threshold level s u c h that the b a c k g r o u n d region constitute a large por t ion bu t more impor tan t ly no chromosome region is e l iminated from the threshold ing process. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 155 Figure 6 .12. Chromosome image at var ious thresholds. The or ig ina l image (a) is thresholded at 90 to 150 i n increments of 10 grey levels to generate b ina ry images (b) to (h). In the b inary images, gray levels above the threshold are represented by white. Image (i) shows the resul t of sett ing a l l va lues of the or ig inal image w h i c h are below the threshold of 110 grey level to a value of 110. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 156 To obta in the value of the threshold, a h is togram of in tensi ty levels of the chromosome image is first generated. Trad i t iona l th reshold ing techniques generally search for the val ley between the peaks i n the h is togram or for a loca t ion at a cer ta in distance from a defined locat ion of the m a x i m u m peak. These t rad i t iona l techniques can not be used for the chromosome images because of the fol lowing reasons. Firs t , even i f smooth ing i s app l ied to remove the noise i n the image, the h is togram m a y conta in more t h a n 2 peaks . Second , the h is togram may conta in only one peak, as the chromosome p ixe ls are spread a lmost un i formly over the intensi ty scale. Las t ly , the m a x i m u m peak i n the h is togram do not a lways correspond to the mode of the b a c k g r o u n d pixels , b u t c a n represent different chromosome regions (i.e. da rk or br ight b a n d s i n the chromosome). Hence, we use a different technique i n ob ta in ing the th reshold level. In our method, we first define the range i n w h i c h the chromosome a n d backg round pixels l ie. Mos t artifact a n d extreme noise p ixe ls are first rejected from th is range. The m i n i m u m level i s then the level where approximate ly 0 .2% of the total n u m b e r of p ixels i n the image have lower intensi t ies . S imi la r ly , the m a x i m u m level is the level where approximate ly 0 .2% of the total n u m b e r of pixels i n the image have higher intensi t ies . The selected m i n i m u m a n d m a x i m u m intensi ty levels then define the range of intensi t ies cor responding to approximately 9 9 . 6 % of the total p ixe ls i n the image. A threshold level, T, i s then set at 3 / 5 of the in tensi ty range from the m i n i m u m intensi ty level. Th i s threshold level was found to w o r k best for images acqui red under a n u m b e r of different s i tuat ions . Th i s level corresponds mos t ly to those backg round pixels w h i c h are near the borders of chromosomes . The p ixe ls away from the chromosomes generally have intensi ty va lues below TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 157 th i s level. A s there are relatively few pixels i n the image at or near the threshold level, a n error of approximately 10% i n the exact loca t ion of the threshold level w o u l d resul t i n approximately 1 p ixe l shift i n the borders of the segmentat ion resul t (Figure 6.10c,d,e). Th i s shift i s not signif icant s ince the purpose of th is step is to remove the pixels i n the backg round w h i c h are away from the chromosomes. Those background pixels w h i c h are close to the chromosomes w o u l d be segmented i n later steps. The first app rox ima t ion image to the chromosome regions i2(x,y) i s then the thresholded vers ion of the chromosome image i(x,y) and is given by the following: Once the threshold value i s found, a l l pixels i n the image w h i c h are below the threshold are set to the threshold (background) level (Figure 6.10). Note that th is first approx imat ion of the segmentat ion m a y con ta in a few erroneous p ixels both w i t h i n and outside the chromosome. 6.5. Second Approximation to Edges: Texture Detection The second step i n our segmentation a lgor i thm refines the chromosome regions obta ined from the first approximat ion . In th is step, we search i n the previous ly defined regions for texture informat ion w h i c h i s character is t ic of the b a n d i n g s t ructures a n d the texture i n the chromosomes of fluorescence images. Th i s search process i s divided into two parts. The first par t f inds poin ts i n the hix>y) i[x,y) if if i(x,y)>T i(x,y)<T (6-2) T TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 158 approximated region w i t h h igh loca l intensit ies. The second par t then expands these poin ts into their ne ighbours to define the chromosome regions. To f ind the loca l h igh intensi ty points , we use the Average Difference filter descr ibed earlier i n Sect ion 5.3. Th i s t ime, ins tead of u s i n g a 3x3 region, the va lue of the local ly averaged image at (x,y) i s the average value of the intensi t ies over a 5x5 square ne ighbourhood region center about (x,y). Th i s larger filter region smoothes out more noise and texture that are present i n the chromosome images. We then impose a non-negative const ra int on the difference image. The entire process is formulated as follows: •^2 2 (a) j[x,y) = il(x,y)-—- £ ^i^x - m,y - n) m=-2 n=-2 (b) i f j(x,y) < 0, replace by j(x,y) = 0 (6-3) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 159 Figure 6 .13 . Second approximat ion to edges: texture detection. The thresholded image is generated from the first approx imat ion step a n d is shown i n (a). A 5x5 average filter is then appl ied to resul t i n (b). The average difference is then generated by subtrac t ing image (b) from (a) to resul t i n image (c). The b a c k g r o u n d has a value of 0 . Da rke r regions represents negative va lues whi le brighter regions represent positive values . The positive va lues of image (c) i s then s h o w n i n (d). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 160 A s a resul t of this operat ion, the background regions, w h i c h are close to the chromosomes , and most of the chromosome edges are set to 0 because i ts resu l t ing intensi ty is below the loca l average of the first approximate segmentat ion (6.13). The isolated background regions where va r i a t ion i n in tens i ty c a n resul t i n positive difference values are not considered because these poin ts have been rejected by thresholding i n the previous step. Detected poin ts (those set above 0) are found scattered further ins ide the chromosome as the image intensi ty difference begins to fluctuate between posit ive a n d negative va lues (Figure 6.13d). Th i s f luctuat ion i s due to the b a n d i n g s t ructures of the chromosome and the r andom noise i n areas where the chromosome intensi t ies are s imi lar . These cluster of detected po in t s are u s u a l l y less t han 2 pixels wide. Mos t of the points in-between touch ing chromosomes are also e l iminated, since they have negative difference va lues . The second par t i n the texture detection a lgor i thm i s to expand a n d connect the detected texture points into chromosome regions. Morpho log ica l filters (dilation filter a n d combina t ion of erosion a n d d i l a t ion filters) are the obvious choice of a lgori thms to use. For example, i n a 3x3 d i l a t ion filter, the va lue of the p ixe l is set to 255 i f a selected n u m b e r of the pixel 's ne ighbourhood has a value of 255 . Otherwise, the p ixe l va lue i s set to 0 as mos t of the ne ighbours have a value of 0. Hence, the gaps i n between detected chromosome points w h i c h are most ly 2 or less pixels wide are fi l led a n d set to a value of 255 . Th i s filter also fills the region between touch ing chromosomes . If the erosion filter (the filter where p ixe l va lues are set to 0 i f some of i ts ne ighbours are 0 and to 255 otherwise) i s used pr ior to the d i la t ion filter for the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 161 purpose of separat ing chromosomes, the chromosome i tself (clustered poin ts ment ioned above) m a y become separated by the process. We overcome this challenge of j o i n i n g chromosome points a n d s t i l l be capable of separat ing touch ing chromosomes by u s i n g our R a n k Difference filter w h i c h we developed and descr ibed earlier i n Sect ion 6.2. A s the gaps to be fil led are 2 or less pixels wide, a 3x3 ne ighbourhood region S(j(x,y)) i s chosen. Larger neighbourhoods w i l l increase the process ing t ime w i t h no signif icant improvement on the resul ts . A n upper and lower r a n k n u m b e r of 7 a n d 1, respectively i s chosen. The R a n k Difference image i s then b inar ized by sett ing a l l negative values to 0 and a l l others to 255 . The resu l t ing image i2(x,y) (Figure 6.14b) i s the second approx imat ion to the chromosome region a n d i s given by: a) b) i2(x,y) = R71[S0(x,y)))] if i2(x,y) > 0, replace i2(x,y) with 255 otherwise replace i2(x,y) with 0 (6-4) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 162 Figure 6.14. R a n k difference of the average difference image. The positive por t ion of average difference image is shown i n (a). The thresholded R a n k Difference operat ion on (a) is shown i n (b). Th i s selection of r a n k numbers and b ina r i z ing dictate that i f there are 3 or more p ixels i n the 3x3 neighbourhood w h i c h have a value greater t h a n the lowest va lue i n the region, the value of the p ixe l is set to 255 . Otherwise, the p ixe l va lue is set to 0. Hence, th is filter behaves l ike a selective d i l a t ion a n d erosion filter. It i s ident ica l to the d i la t ion filter i f the lowest value is a lways 0 (which is not the case i n our images) and a l l other values are b inar ized to 255 . Poin ts w i t h s imi la r edge magni tudes i n a ne ighbourhood are eroded a n d set to zero (e.g. edges of chromosomes and some areas i n between chromosomes) . Conversely, va ry ing magni tude points are di lated and set to 255 (e.g. areas w i t h i n the chromosome). The resul t is a more refined m a s k of the chromosome region. It resembles a skeleton outl ine of chromosomes as only the interior of TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 163 the chromosome i s present and most of the edge pixels are e l iminated . A l m o s t a l l b a c k g r o u n d regions a n d regions i n between touch ing chromosomes are removed by th is R a n k Difference filter operation. 6.6. Third Approximation to Edges: Region Refinement and Labeling The f ina l step i n determining the chromosome region is to refine the resul ts of the previous approx imat ion to obta in a better estimate of the chromosome border a n d to labe l or d i s t ingu i sh the region of one chromosome from another. A t th is stage, we can use either the Difference of G a u s s i a n , C a n n y or R a n k Difference filter since for b inary pic tures (Figure 6.6), they a l l produce good resul ts . We chose to use our R a n k Difference filter because of the fol lowing reasons. Fi rs t , the a lgor i thm i s already avai lable w i t h i n the program. Second, the algor i thm uses only integer operat ions a n d i s less complex a n d hence it i s faster to compute. Las t a n d mos t impor tan t ly , the R a n k Difference filter gives th ick edges at the appropriate locat ions s u c h that some of the remain ing touch ing chromosomes w h i c h are not segmented i n previous approximat ions c a n be separated. In this ins tance, the R a n k Difference Fi l ter i s used as a n edge detector ins tead of a selective d i l a t ion filter. The purpose of th is filter operat ion is to determine the borders of the chromosome regions. Th i s filter operates over a 3x3 ne ighbourhood u s i n g 9 a n d 1 as the upper and lower r a n k numbers , respectively. Th i s filter generates the bounda ry image b(x,y) w h i c h is defined as follows: b(x,y)=R9A[S(i2(x,y))\ (6-5) TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 164 Since the inpu t image is b inary , the resu l t ing image i s also b ina ry (values of 0 a n d 255). The filter generates th ick edges (approximately 3 p ixels wide) a r o u n d the boundar ies of the chromosome. Pixels i n between touch ing chromosomes w h i c h have 4 or less con t inuous p ixels now become a n edge p ixe l a n d are set to 255 . A logical ar i thmetic operat ion i s then employed to separate the touch ing chromosomes. In this operat ion, a new chromosome region, m(x,y) i s generated based on the logical A N D (•) of the previous chromosome approx imat ion region, i2(x,y), w i th the logical N O T (—)of the newly ca lcu la ted bounda ry image, b(x,y) as follows: m[x,y) = i2[x,y)»b[x,y) (6-6) The resu l t ing image, m(x,y) then contains regions defined by the second approx ima t ion image i2(x,y) less those boundary p ixels w h i c h lie i n bo th i2(x,y) a n d b(x,y). A l t hough the regions found are smal ler t h a n the ac tua l regions of the chromosomes , they are most ly dis t inct and isolated from one another. E a c h object is next labeled s u c h that each isolated region is given a d is t inc t number . The size of the region of each labeled object is then increased s u c h that it i s representative of the size of the chromosomes. Th i s inc reas ing process i s accompl i shed by d i la t ing each labeled region twice u s i n g a 3x3 d i l a t ion filter. In th is d i la t ion process, the center p ixe l i n the 3x3 region i s set to 2 5 5 i f any of the pixels i n the region has a value of 255 . Otherwise, the center p ixe l i s set to 0. A s different label number s are used i n the d i l a t ion process , regions w h i c h touch one another after the d i la t ion are kept d is t inc t w i t h different label numbers . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 165 Figure 6 .15 . T h i r d approx imat ion to edges: region refinement and label ing . A 3x3 R(9 , l ) R a n k Difference filter is appl ied to the resul t of the second approx ima t ion step (Figure 6.14b) and is shown i n (a). A n inverse of image (a) i s generated and shown i n (b). A logical A N D is performed on the image from the second approx imat ion step (Figure 6.14b) and the image i n (b) to give image (c). Objects w h i c h touches the edges of the image are deleted. The image is then labeled and di lated. A border is then placed a round each labeled chromosome to resul t i n image (d). TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 166 6.7. Feature Extraction and Artifact Removal In a l l previous steps, no a-pr ior i morphologica l informat ion about the chromosome is used . In th is step, we uti l ize features of the "chromosome" objects as parameters for rejecting artifacts from the detected objects from the t h i rd approx ima t ion image. Chromosome objects are generally more intense t h a n artifact objects w h i c h have intensit ies s imi la r to that of the b a c k g r o u n d . Hence, the IFI over the defined area of the object c a n help i n object d i sc r imina t ion . The area i s ca lcula ted from the total n u m b e r of p ixe ls w i t h i n the labeled object. The IFI is ca lcula ted by first s u m m i n g the intensi t ies of a l l p ixe l s i n the detected object a n d then subt rac t ing the average b a c k g r o u n d in tens i ty mu l t i p l i ed by the n u m b e r of pixels i n the detected region. The b a c k g r o u n d is ca lcula ted by determining the average intensi ty of the p ixels w h i c h l ie j u s t outside the object region. F ina l ly , the dec is ion for rejection is then to el iminate objects w h i c h are d i m and have a n average of 5 or less gray levels above the backg round intensi ty level. 6.8. Associate Telomere with Chromosome To further refine both the telomere and chromosome regions, bo th the telomere a n d chromosome images are used . Other chromosome segmentat ion a lgor i thms do not have our added advantage of hav ing cor responding telomere images w h i c h c a n help i n defining the ends of the chromosomes. A s there are no cor responding reference points i n the telomere a n d chromosome images for image registrat ion, a probabi l i s t ic ma tch ing of the two images i s first performed. The telomere image is shifted at 2 p ixe l steps i n bo th the x a n d y TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 167 direct ions from the chromosome image. A t each shift pos i t ion , the n u m b e r of telomeres that are located w i t h i n the chromosome regions are determined. The shift n u m b e r w h i c h corresponds to the largest n u m b e r of telomeres found becomes the chosen shift va lue . Once the shift va lues have been found, the n u m b e r of telomeres w i t h i n each labeled chromosome can then be calcula ted. A s there c a n only be four telomeres i n a chromosome, telomeres or chromosomes w h i c h do not follow th is ru le are highl ighted accordingly s u c h that they c a n be easi ly seen d u r i n g the m a n u a l edi t ing and verif icat ion stage. Those telomeres w h i c h are more t h a n 2 p ixe ls away from any chromosome m a s k are treated as artifacts a n d are rejected from further analys is . A n example of the resul ts of overlaying the telomere borders onto the chromosome image a n d segmented resul ts are s h o w n i n Figure 6.16. In this image, the details w i t h i n the chromosomes are enhanced by contrast s t retching s u c h that they c a n be more readi ly seen. The b a c k g r o u n d i s also set to a gray colour (instead of the n o r m a l black) for v i s u a l enhancement of the details. It c an be seen from the image that some of the telomeres are ly ing par t ia l ly or j u s t outside the chromosome border (e.g. chromosomes #6 a n d #7 i n Figure 6.16). These telomeres are proper ly associated w i t h the correct chromosome. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 168 Figure 6 .16. Chromosome and telomere segmentation results . The borders of the chromosomes a n d telomeres are shown i n b lack . The backg round of the or ig ina l image is changed from b l ack to gray a n d the intensi t ies w i t h i n the chromosome are contrast enhanced and inverted to help v isual ize the details w i t h i n . Chromosomes w h i c h lie on the boundary of the image are not segmented i n the a lgor i thm but their telomere resul ts w h i c h do not lie on the image bounda ry is shown. 6.9. Segmentation Performance O u r segmentat ion method described i s compared to the best of the different edge filters previously described, the C a n n y filter, for chromosome TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 169 segmentat ion. Sample chromosome segmentation resul ts from the C a n n y a n d our method are shown i n Figure 6.17. I ( a ) ( b ) ( c ) Figure 6 .17. C o m p a r i s o n of our segmentation a lgor i thm to the C a n n y filter. The or ig ina l image is shown i n (a). The results of the C a n n y filter is s h o w n i n (b). The resul ts of our a lgor i thm is shown i n (c). In our a lgor i thm, a l l chromosomes have closed boundar ies and chromosomes w h i c h touch the edge of the image are d iscarded. It c a n be seen that our method gives super ior resul ts compared to the C a n n y filter. The borders i n our method encompasses the outer edge of the chromosome. The larger m a s k wou ld help i n associa t ing i n d i v i d u a l telomeres, w h i c h can fall outside the chromosome region, to the cor responding chromosome. The borders shown are cont inuous and they properly describe the region occupied by the chromosome. Even the backg round i n between the a rms of chromosomes are marked as not being part of the chromosome. In the C a n n y image however, some of the chromosomes (in par t i cu la r ly the smal le r ones) do not have cont inuous boundar ies . The edges w i t h i n the chromosomes are also not proper ly jo ined s u c h that these backg round regions are properly classif ied. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 170 It i s difficult to accurately quantify the performance of ou r a lgor i thm. There are a n u m b e r of reasons for this . Firs t , metaphase chromosomes are m a n u a l l y selected. Hence, the performance of our a lgor i thms i s largely dependent on w h i c h metaphases are selected for the ana lys is . Second , we do not have access to commerc ia l chromosome segmentation packages from w h i c h we c a n compare our results . Last ly , there are cases where v i sua l ly , i t i s difficult to d i s t ingu i sh w h i c h por t ion of a touch ing or a n over lapping chromosome belongs to w h i c h of the two chromosome. T h u s , we performed a qualitative evaluat ion of our segmentat ion a lgor i thm. F r o m the hundreds of metaphase chromosomes analyzed, there are a n u m b e r of different types of errors observed. Mos t of these errors c a n be subsequent ly corrected by interactively edi t ing the generated resul ts . Fi rs t , errors m a y arise w h e n some touch ing or over lapping chromosomes are not proper ly separated. In these instances, the intensi t ies at the borders resemble those w i t h i n the chromosomes. Thus , we i nc luded a u t i l i ty to force regions to sp l i t by d rawing cut t ing l ines i n the image before the segmentat ion is performed. Second, chromosomes may be improper ly separated. Th i s u s u a l l y occurs at the bounda ry of chromosomes where the observed intensi t ies i n the over lapping region are more intense than those observed i n non-over lapping regions. Th i s spl i t c an also occur i n a n a r m of the chromosome where a subs tan t ia l ly wide da rk b a n d is present. These spl i t regions c a n be jo ined together d u r i n g the interactive edit ing phase. T h i r d , telomeres m a y be ass igned to the wrong chromosome as they are more i n the v ic in i ty of another ne ighbour ing chromosome or are outside the borders of the telomere search region. F ina l ly , telomeres m a y not be properly pa i red a n d ordered w i t h i n the TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 171 chromosome. W i t h the edit ing features, proper pa i r ing a n d labe l ing of each of the telomeres i n the p and q a rms of the chromosome c a n be made. A l t h o u g h there may be some errors i n segmentat ion u s i n g our segmentat ion a lgor i thm, there are considerable regions w h i c h are correct ly segmented. A s our segmentation a lgor i thm is u sed to pre-process the acqu i red images before m a n u a l edit ing, subs tan t ia l savings i n t ime have resul ted as a majori ty of chromosomes are correctly segmented. W i t h the interactive edi t ing features, near ly a l l chromosomes c a n be separated a n d analyzed. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 172 Chapter 7. Conclusion and Future Suggestions 7.1. Overview We have accompl i shed the objectives set forth for th is project. B a s e d on the work descr ibed i n this thesis, our or ig inal hypothes is , w h i c h postulates that the length of i n d i v i d u a l telomeres i n a cel l c an be determined from digi ta l images of fluorescence in situ hybr id iza t ion prepared cells , i s accepted. Th i s c o n c l u s i o n is based on our studies descr ibed i n Sect ion 7.3. In order to perform these studies, we h a d to develop the hardware system, a lgor i thms, a n d software to a l low for reliable measurements . Conven t iona l systems u s i n g the Sou the rn analys is c an only determine the average length of telomeres of a popu la t ion of cells , bu t can not determine the length of i n d i v i d u a l telomeres be longing to every chromosome i n a cel l . A s u m m a r y of our work a n d the performance of our system are d i scussed i n Sect ion 7.2. O u r sys tem is cur ren t ly be ing u s e d i n the Terry Fox Laboratory at the B . C . Cancer Research Centre a n d i n the Nether lands on a rout ine basis to s tudy the behaviour a n d role of telomeres i n cells. Two s u c h studies are descr ibed i n Sect ion 7.3. There are also p lans to ut i l ize our analys is system a n d extend the telomere s tudies into G e r m a n y a n d the Un i t ed K ingdom. Improvements to the sys tem a n d other areas of development are d i scussed i n Sect ion 7.4. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 173 7.2. System Performance W i t h the telomere analys is system that we developed, the d i s t r ibu t ion of telomere lengths of the cells unde r s tudy can be generated (Lansdorp et a l . , 1997). The advantage of our system is that s ignif icantly fewer cells (less t h a n 30 cells) are required to obta in resul ts compared to the convent ional Sou the rn ana lys i s w h i c h requires analys is of approximately 100,000 cells . T h i s makes it poss ible to car ry biological s tudies when only a l imi t ed n u m b e r of cells are available for analys is . In addi t ion , telomere length s tudies c a n n o w be car r ied out on i n d i v i d u a l cells as wel l as i nd iv idua l chromosomes i n every cel l . No other method is current ly available to determine the length of i n d i v i d u a l telomeres. Hence, no direct method for verifying the accuracy of our a lgor i thms for telomere length measurements is available. For th i s reason, we resorted to indirect methods to validate our fluorescence measurements . For th i s purpose, we u s e d objects of k n o w n fluorescence intensi t ies w h i c h resemble telomeres. These objects i nc luded i) s imula ted objects of different shapes a n d sizes, ii) fluorescence beads of k n o w n size a n d relative fluorescence intensi t ies , a n d iii) p lasmids w i th k n o w n telomere insert lengths (which are typ ica l ly a n order of magni tude less i n length t h a n the telomeres i n the cells). O u r a lgor i thm est imated the integrated fluorescence in tens i ty (IFI) of s imu la t ed objects of va ry ing shapes and sizes to w i t h i n ± 3 % . The es t imated m e a n IFI va lues correlated wel l (correlation coefficient of 0.99) w i t h the size of the f luorescence beads a n d wi th the length of telomere insert i n p l a smids (Martens et a l . , 1997). The s tandard deviat ion i n the es t imat ion ranged from 2% for the l u m beads to 13% for the 0 .2um beads to 29% for the O . l u m beads. The TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 174 s t andard deviat ion was larger for the smal ler beads because i t i s more difficult to fabricate them. The s tandard deviat ion for telomere inser ts i n p l a s m i d s was a r o u n d 2 0 % of the m e a n est imated IFI value. Th i s var iance i s mos t l i ke ly due to the var iable efficiency of the hybr id iza t ion procedure (binding of the probe). We observed a s imi l a r var ia t ion i n hybr id iza t ion on chromosomes after hybr id i za t ion w i t h probes specific for centromere repeat sequences of invar iable length. A l t h o u g h the var ia t ion appears to be large, we observed that by averaging the resul ts of 10 or more cells, a good ind ica t ion of the telomere length on a par t i cu la r chromosome a r m i n a popu la t ion of cells c a n be obta ined. That i s , w i th resul ts from 10 or more cells (>40 telomeres), the W i l c o x o n - r a n k - s u m test showed a significance level of less t h a n 0 .05 i n differentiating between telomere lengths of chromosome groups. It i s impor tan t to note that there are no other methods w h i c h can produce s imi la r , let alone better resul ts . In our ana lys i s of each cel l , we need to capture mul t i - focus p lane images con ta in ing only telomere s ignals and a single image con ta in ing only chromosome signals . The images of telomeres are u s e d to evaluate the telomere IFI va lues w h i c h give a n estimate of the telomere lengths i n the cel l . The cor responding image of the chromosome is required to identify the regions occup ied by the chromosomes. B y identifying a n d class i fying each chromosome i n the image a n d associa t ing i t to i ts cor responding telomere, a n estimate of the length of every telomere of each chromosome i n the ce l l i s obta ined. F r o m ana lyz ing a n u m b e r of cells , the telomere fluorescence d i s t r ibu t ion for each type of chromosome i n the cel l i s real ized. O u r a lgor i thm TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 175 for telomere segmentation, telomere fluorescence measurement , a n d chromosome segmentat ion for each metaphase chromosome sample takes less t h a n 1 minu te to perform on a 1 0 0 M H z Pent ium-based microcomputer . In accompl i sh ing our major goal for the project, we have b u i l t a n d u n i q u e l y character ized a fluorescence microscopy imaging system for cap tu r ing images of metaphase chromosomes and mul t i - focus plane images of telomeres. The character iza t ion of the system was then used to generate s imula ted images of the different shape and size test objects. In order to compare da ta acquired at different intervals i n t ime a n d space, we have developed image process ing techniques to compensate for the spa t ia l a n d tempora l dis tort ions in t roduced by the acqu is i t ion system. The tempora l dis tor t ions is a resul t of the decay i n fluorescence in tens i ty of the probe at tached to the object. We have also developed novel techniques to analyze telomeres a n d chromosomes. In these developments, we have in t roduced a lgor i thms to first segment mul t i - focus plane images of telomeres a n d then extract the integrated fluorescence intensi ty (IFI) va lues for each detected telomere. The IFI value i s propor t ional to the telomere length. In add i t ion , we have developed algori thms to segment chromosomes i n c l u d i n g those w h i c h are j u s t touching. The segmentation resul ts a n d the ca lcu la ted telomere IFI va lues are then presented to the user for verif icat ion a n d edit ing. The au tomat ion of the telomere and chromosome extract ion a n d IFI ca lcu la t ion process ha s s impl i f ied the user verif icat ion a n d edi t ing process. O n average, over 9 0 % of the chromosomes are segmented properly. The success rate i n segmentat ion i s dependent on the metaphase sample w h i c h typ ica l ly conta ins a TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 176 few over lapping chromosomes. The improper ly segmented chromosomes c a n be corrected w i t h i n 5 minutes by the user. In addi t ion , the user does not need to perform the tedious task of defining the exact border for every telomere w h i c h i s required to obta in a consistent telomere IFI value . 7.2.1. Imaging system We have successful ly developed and bu i l t a n imag ing sys tem for fluorescence microscopy. Th i s system is capable of acqu i r ing a large range of s igna l intensi t ies (0.00001 to 100 lux) from very faint to s t rong s ignals . The sys tem c a n also acquire images at different focus p lanes spaced at 0.1 um or more from each other. The cr i t i ca l elements of our system, w h i c h we p a i d spec ia l at tent ion to d u r i n g the component evaluat ion and selection process are the i l l u m i n a t i o n source, fluorescence exci tat ion and emiss ion filters, objective lens a n d h igh reso lu t ion integrat ion camera . The components were selected s u c h that the qual i ty of the captured image i s sufficiently h igh a n d t hus very litt le pre- process ing is required to correct or compensate for the aberrat ions i n the images. The pre-processing a n d other funct ional a lgor i thms w h i c h we determined to be essential to incorporate into the ana lys i s i n c l u d e d faulty p ixe l correct ion, flat-field correct ion, automat ic selection of integrat ion t ime a n d lookup table selection, a n d mul t i - focus plane image acquis i t ion . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 177 7.2.2. System Characteristics We developed a new method for character iz ing the response of our microscope imaging system. We based our method on Cas t leman ' s der ivat ion a n d incorpora ted the cont r ibu t ion of the response funct ion of the image detector. O u r resul ts are more representative of the system behaviour t h a n those u s i n g Erhard t ' s method. O u r theoretical P S F , however, i s on ly a n estimate a n d does not properly characterize the system response of our system. T h i s i s because the so lu t ion used to fix the chromosome onto the microscope sl ide (and to prevent photobleaching of the fluorescent probe) has a refractive index w h i c h is different from that of the cover s l ip for the sl ide a n d the i m m e r s i o n o i l . A s the locat ion (z-direction) of the telomere var ies w i t h i n the so lu t ion , the extent of the b lu r r i ng effects caused by the difference i n refractive ind ices c a n not be predicted for each telomere. Hence, our theoret ical P S F funct ion i s only used for generating s imula ted objects to test the telomere IFI a lgor i thms. 7.2.3. Telomere IFI Value We first performed a n ana lys i s to ga in a n under s t and ing of wha t the IFI va lue represents a n d how this value c a n be theoretically ca lcu la ted . The IFI va lue was found to be propor t ional to the s u m of a l l l ight intensi t ies that originate from the object. We determined that this va lue needs not be s u m m e d over a 3 D space bu t can be obtained from a single image plane as long as a sufficiently large region i s used . The pract ical i t ies i n the sample prepara t ion a n d sys tem s u c h as close p rox imi ty of telomeres, quant iza t ion l im i t a n d errors TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 178 i n the detector, noise i n the image and problems i n segmentat ion l imi t ed u s to s u c h a n analys is . We thus resorted to analyze mul t ip le focus p lane images to estimate the telomere IFI value . We have developed a telomere and IFI quant i f icat ion a lgor i thm w h i c h c a n segment objects w h i c h are spaced at dis tances greater t h a n 0 .54um from each other i n the system. We compared our IFI a lgor i thm w i t h s imula t ed test objects as we l l as w i th experimental resul ts u s i n g beads a n d p l a s m i d s where the relative fluorescence intensit ies are p resumably k n o w n . O u r resul ts correlated wel l w i t h the resul ts of these experiments. We observed that better resul ts c a n be obtained i f the IFI value is chosen from the best focussed image (i.e. image w i t h the highest IFI value) for each object i n the set of mul t i - focus p lane images. A 2 0 % reduct ion i n the IFI value from the best focus va lue c a n resul t i f only a single image is used i n the analys is s ince a n image c a n con ta in objects w h i c h are ± 0 . 2 u m i n z-focus away from the best focus image. Alternat ively, resul ts w i t h i n our acceptable accuracy l imi t c a n also be obta ined i f the s u m of the IFI va lues from a s tack of images spaced no more t h a n 0 .3um from each other is ca lcula ted. 7.2.4. Chromosome Segmentation We have developed a chromosome segmentation a lgor i thm w h i c h i s successfu l i n determining the regions w h i c h belong to chromosomes . The resu l t ing au tomat ion i n the segmentation el iminates the lengthy t ime required by the use r to m a n u a l l y define the borders of each chromosome i n the image. Once the image is segmented, karyotyping (chromosome type identification) c a n TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 179 be performed. In conjunct ion w i th the telomere IFI resul ts obta ined earlier, the length of every telomere i n each chromosome i n the cel l i s obtained. For our segmentation a lgor i thm, we first in t roduced the R a n k Difference filter w h i c h i s u sed i n each of the two steps of our segmentat ion a lgor i thm. Th i s R a n k Difference filter c an act as a n edge detector or a morpholog ica l filter. A s a n edge detector, our R a n k Difference filter gives better loca l iza t ion of the edges t h a n the Difference of G a u s s i a n s or C a n n y filters. It a lso performs better t h a n the other filters on images w i t h additive un i fo rm noise. , For chromosome images, our segmentation a lgor i thm outperforms other edge detectors i n defining cont inuous regions a n d i n separat ing t ouch ing chromosomes . Since our a lgor i thm uses only integer operations, i t performs faster t h a n the Difference of G a u s s i a n s or C a n n y filters w h i c h use floating poin t ar i thmetic . A s a vast majority of chromosomes (typically >90%) are proper ly separated by our a lgor i thm, less user in teract ion i s required to correct a n d edit those chromosomes w h i c h are improper ly segmented by the a lgor i thm. Hence, greater product iv i ty i n the analys is is obtained. 7.3. Current Biological Studies The sys tem is current ly be ing used on a da i ly bas i s to analyze telomeres at the Terry Fox Laboratory of the B . C . Cancer Research Centre. A dedicated sys tem i s u sed for acqu i r ing the telomere and chromosome images. A t least two other process ing systems are be ing used to analyze the acqui red images u s i n g our ana lys i s software. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 180 Two biological s tudies have been completed to-date u s i n g our sys tem. The first s tudy (Zijlmans et a l . 1997) investigates the telomere length d i s t r ibu t ion i n mice . Previous s tudies i n th is area have s h o w n that mice have long telomeres w h i c h do not appear to shorten as they age. T h i s p h e n o m e n a contradic ts the concept that telomeres shorten w i t h age. U s i n g our ana lys i s sys tem, we have found that there is a large var ia t ion i n the telomere lengths i n mouse cells . We also observed that there are specific chromosomes i n bone mar row a n d s k i n fibroblast cells i n i n d i v i d u a l mice w h i c h have s imi la r telomere lengths. We also observed the presence of very short telomeres w h i c h m a y be the c r i t i ca l l i n k i n l im i t i ng the cel l repl icat ion process (aging process) i n mice . The second s tudy (Martens et a l . 1997) investigates the telomere length d i s t r ibu t ions i n h u m a n cells. The resul ts of the analys is of one of the metaphase samples is shown i n Figure 7.1. O u r image analys is generates the borders of the telomeres from the telomere (figure labeled CY3) images a n d the borders of the chromosomes from the chromosome (figure labeled DAPI) image. The segmentat ion resul ts are super imposed onto the processed chromosome image (figure labeled Image Analys is ) . The "X" chromosome is h ighl igh ted i n the example to i l lust ra te its telomere IFI values on the p a n d q chromosome arms. A Pseudo-Colour image is also generated from the chromosome a n d telomere image. Th i s pseudo-Colour image is then u s e d to generate the Karyogram image w h i c h sorts and identifies the different chromosome types (chromosomes #1 to #22 and chromosomes X and Y). The respective telomere IFI va lues of each chromosome a r m is also generated (figure labeled w i t h Telomere Karyogram) to facilitate i n the biological ana lys is a n d interpretat ion. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 181 DAPI a * W C <% * t*. I f / * . Image ^ Analysis, ^ Cy-3 0 Pseudo- colour Karyogram j b r S | H X 44 R « # CV* />/ p2 ql q2 X 324 266 166 197 93 33 Telomere Karyogram p-arm q-arm h A | L L L g " 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 1 1 2 2 3 : LJl 12 131314 14 15 15 16 16 17 17 1818 19 19 20 20 21 21 22 22 X Y """ l | B 'U" |F" 'P Chromosomes Figure 7 .1 . Telomere lengths of i n d i v i d u a l chromosomes i n a cel l . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 182 In this second study, we found that for a n i n d i v i d u a l , the telomere lengths i n a specific chromosome from a cer ta in t issues are very s imi l a r to those of other t issues. However, the telomere lengths do vary from a n i n d i v i d u a l to another. We also noted from a s tudy of 11 unre la ted i nd iv idua l s that the telomeres on the arms of chromosome 17p are consis tent ly among the shorter telomeres i n the cell . Studies i n this area may give a n ins ight in to w h y cancer cells frequently lose the ends of chromosome 17p. U s i n g our telomere analys is system, other experiments a n d invest igat ions can now be performed to s tudy a n d determine the role telomeres p lay i n the aging process a n d i n patients w i th cancer or genetic disorders . 7.4. Future Suggestions and Applications A l t h o u g h the system is current ly used on a dai ly bas is , there are a n u m b e r of improvements w h i c h can be made to the system. In terms of the hardware , a cooled integrat ing C C D camera w o u l d be useful i n ob ta in ing better qual i ty images. W i t h s u c h a camera, less faulty pixels w o u l d be present. Hence , the accuracy of the IFI a lgor i thm w o u l d be improved s ince i t i s no longer necessary to estimate the value of the faulty p ixe l by t ak ing the average of i ts su r round ings . Another hardware component w h i c h w o u l d benefit the sys tem a n d improve the ca lcula ted IFI value is a more accurate a n d repeatable z-focuss ing m e c h a n i s m . We recently acquired a piezo-electric mot ion control ler w h i c h adjust the pos i t ion of the objective lens. T h i s sys tem has less b a c k l a s h a n d i s more accurate t han the exis t ing mechan i ca l motor a t tached to the focuss ing knob of the microscope. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 183 One method of improv ing the th roughput of ana lyz ing specimens is to automate the karyotyping process. Th i s process is current ly the mos t t ime c o n s u m i n g step as each chromosome image i s m a n u a l l y sorted by a cytogenetics t echnic ian . The automated process developed s h o u l d be s u c h that the sorted chromosomes is easier to l i n k to the cor responding telomere IFI va lue generated by the current program. F ina l ly , research into segmenting telomeres i n interphase n u c l e i (a c i r cu la r shaped nuc leus where chromosomes are c l u m p e d a n d are ind i s t ingu i shab le from one another) w o u l d significantly increase the n u m b e r of samples w h i c h can be analyzed. In addi t ion , the t ime to acquire images w o u l d be reduced as it i s no longer necessary to only scan a n d select metaphase chromosomes from the slide. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 184 Chapter 8. Bibliography A g a r d D . , H i r a o k a Y . , a n d Sedat J . (1989a) Three-dimensioa l microscopy: image process ing for h igh resolu t ion subce l lu la r imaging, i n New Methods i n Mic roscopy a n d L o w Light Imaging, ed., J a m e s Wampler , S P I E 24-30 . A g a r d D . , H i r a o k a Y . , S h a w P., and Sedat J . (1989b) Fluorescence microscopy i n three d imensions , i n Methods i n C e l l Biology, ed. L a n s i n g Taylor , 3 5 3 - 377 . A g a r d D . (1984) Op t i ca l sect ioning microscopy: ce l lu lar archi tecture i n three d imens ions . A n n . Rev. B iophys . Bioeng. 13:191-219. A l l sh i r e R . C . , Gosden J .R . , Cross S . H . , C rans ton G . , Rout D . , Sugawara N . , Szos tak J . W . , Fantes P .A . , and Hast ie N . D . (1988). Telomeric repeat from T. the rmoph i l i a cross-hybridizes w i t h h u m a n telomeres. Nature 332:656-659 . A l l s o p p R., V a z i r i H . , Pat terson C. Golds te in S. Y o u n g l a i E . , Fu tche r B . , Greider C . W . , a n d Har ley C . B . (1992) Telomere length predicts replicat ive capaci ty of h u m a n fibroblasts. Proc. Nat l . A c a d . S c i . U S A 89:10114- 10118. A m a d a s u n M . , a n d K i n g R . A . (1988) Low-level segmentat ion of mul t i spec t ra l images v i a agglomerative c lus te r ing of un i fo rm neighbourhoods . Pat tern Recogni t ion , 21(4):261-268. B e n b o w R . M . [1992] Chromosome structures . S c i . Progress Oxford. 76 :425- 450 . Beng t s son E . , E r i k s s o n O. , Ho lmqu i s t J . , a n d Stenkvis t B . (1979) Implementat ion a n d evaluat ion of a diode array scanner for d ig i t iz ing microscopic images, i n The A u t o m a t i o n of Cancer Cytology a n d C e l l Image Ana lys i s , eds. P res smann N . J . , a n d Weid S .L. , 269-289 . Bertero M . , B o c c a c c i P. , Brakenhoff G . J . , Mal fan t i F . , a n d v a n der Voor t H . T . M . (1990) Three d imens iona l image restorat ion a n d super - reso lu t ion i n fluorescence confocal microscopy. J . Mic roscopy 157(l) :3-20. Bertero M . , B o c c a c c i P. , Defrise M . , de M o l C , a n d P ike E . R . (1989) Super - reso lu t ion i n confocal scann ing microscopy: II. the incoherent case. Inverse Problems 5:441-461. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 185 Bertero M . , B r i a n z i P. , a n d Pike E . R . (1987) Super - reso lu t ion i n confocal s cann ing microscopy. Inverse Problems 3:195-212. Br igh t D . S . a n d Steel E . B . (1986) Brightf ie ld image correct ion w i t h va r ious image process ing tools, i n Mic robeam A n a l y s i s , eds. R o m i g A . D . Chamber s W . F . S a n Franc isco Press. 517-520. C a n n y J . F . (1986) A computa t iona l approach to edge detection. I E E E Trans , of Pat tern Ana lys i s a n d M a c h i n e Intelligence. 8(6):679-698. C a n n y J . F . (1983) F i n d i n g edges and l ines i n images. Techn ica l Report T R - 720 , MIT, Cambridge , Massachuse t t s . C a r l s s o n K . (1991) The influence of specimen refractive index, detector s igna l integrat ion, and non-un i fo rm scan speed on the imaging properties i n confocal microscopy. J o u r n a l of Mic roscopy 163(2):167-178. Ca r r ing ton W . A . , L y n c h R . M . , Moore E . D . W . , Isenberg G . , Fogarty K . E . , a n d Fay F . S . [1995] Science 268:1483-1487. Ca r r ing ton W. , Fogarty K . , Lifschi tz L . , and Fay F . (1989) Three-d imens iona l imag ing on confocal a n d wide-field microscopes, i n The H a n d b o o k of B io log ica l Confocal Microscopy , ed., J ames Pawley, I M R Press, M a d i s o n , 137-146. Ca r r ing ton W. , a n d Fogarty K . (1987) Three-d imens iona l d i s t r ibu t ion i n l i v ing cells by deconvolut ion of opt ical sections u s i n g l ight microscopy, i n Proc. 13th A n n u a l Nor th-Eas t Bioengineer ing Conference, ed., K . Foster , 108- 111. C a s t l e m a n K . R . (1993) Color compensat ion for digi t ized F I S H images. B i o i m a g i n g 1:159-165. C a s t l e m a n K . R . (1979) Dig i ta l Image Processing. Prentice H a l l , New Y o r k . de Lange T., Sh iue L . , Myers R . M . , Cos D . R . , Naylor S .L . , K i l l e ry A . M . , a n d V a r m u s H . E . (1990) St ructure and var iabi l i ty of h u m a n chromosome ends. M o l . C e l l . B i o l . 10:518-527. Dav is L . S . (1975) A survey of edge detection techniques. Compu te r G r a p h i c s a n d Image Processing, 4 :248-270. de Lange T. [1995] Telomere D y n a m i c s and Genome Instabi l i ty i n H u m a n Cancer , i n Telomeres, eds. B l a c k b u r n E . H . , and Greider C . W . C o l d S p r i n g Ha rbou r Laboratory Press, New York , 265-293 . E g h o l m M . , B u c h a r d t O. , Chr i s t ensen L . , Behrens C , Freier S., Dr iver D . A . , B e r g R . H . , K i m S.K. , Norden B . and Nielsen P . E . (1993) P N A hybr id izes TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 186 to compl imenta ry oligonucleotides obeying the W a t s o n - C r i c k hydrogen b o n d i n g rules . Nature 365: 566-568. E r h a r d t A . , Z inser G . , K o m i t o w s k i D . , a n d Bi l l e J . (1985) Recons t ruc t ing 3-D l ight microscopic images by digi ta l image process ing. A p p l i e d Opt i c s 24(2): 194-200. F u K . S . , a n d M u i J . K . (1981) A survey of image segmentat ion. Pat tern Recogni t ion , 13:3-16. G a r b a y C , Chasse ry J . M . , a n d B r u g a l G . (1986) A n iterative region-growing process for cel l image segmentat ion based on loca l color s imi la r i ty a n d global shape cri teria. Ana ly t i ca l a n d Quant i ta t ive Cytology a n d Histology, 8:25-34. Go lds t e in S.R., H u b i n T., Rosentha l S., and W a s h b u r n C . (1990) A confocal video-rate laser-beam scann ing reflected-light microscope w i t h no mov ing parts . J o u r n a l of Mic roscopy 157(l) :29-38. G o o d m a n J . W . (1968) In t roduct ion to Four ie r Opt ics . M c G r a w - H i l l , S a n Franc i sco . G r a h a m M . D . , a n d Norgren P . E . (1980) The Diff3 Analyzer : A P a r a l l e l / S e r i a l Go lay Image Processor, Real-Time M e d i c a l Image Processing, eds., Onoe M . , Pres ton K . J r . , a n d Rosenfeld A . , P l e n u m , New Y o r k . Har ley C . B . , Fu tche r A . B . , and Greider C . W . (1990). Telomeres shor ten d u r i n g aging of h u m a n fibroblasts. Nature 345:458-460. Hast ie N . D . , Dempster M . , D u n l o p M . G . , T h o m p s o n A . M . , Green D . K . , a n d Al l sh i r e R . C . (1990). Telomere reduct ion i n h u m a n colorectal c a r c i n o m a a n d w i t h ageing. Nature 346:866-868. H e l l S., a n d Stelzer E . H . K . (1995) Lens aberrat ions i n confocal f luorescence microscopy, i n Handbook of Bio log ica l Confocal Mic roscopy , ed. Pawley J . B . P l e n u m Press, New York , 347-354. H e l l S., Reiner G . , Cremer C , a n d Stelzer E . H . K . (1993) Aberra t ions i n confocal fluorescence microscopy i nduced by mismatches i n refractive index. J o u r n a l of Mic roscopy 169(3):391-405. H i r a o k a Y . , Sedat J . W . , a n d Agard D . A . (1990) De te rmina t ion of three- d imens iona l imaging properties of a l ight microscope system: par t i a l confocal behaviour i n epifluorescence microscopy. B i o p h y . J . 57 :325- 333 . TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 187 Holmes T . J . a n d L i u Y . H . (1991) Accelera t ion of m a x i m u m - l i k e l i h o o d image restorat ion for fluorescence microscopy a n d other non-coherent imagery. J o u r n a l of Op t i ca l Society of A m e r i c a 8(6):893-907. Ho lmes T . J . a n d L i u Y . H . (1989a) R i c h a r d s o n - L u c y / m a x i m u m l ike l i hood image restorat ion a lgor i thm for fluorescence microscopy: further test ing. A p p l i e d Opt ics 28(22):4930-4938. Ho lmes T . J . (1989b) Expec ta t ion-maximiza t ion restorat ion of band- l imi ted , t runca ted point-process intensi t ies w i t h appl ica t ion i n microscopy . J o u r n a l of Opt ica l Society of A m e r i c a 6(7): 1006-1014. H o p k i n s H . H . (1955) The frequency response of a defocused opt ica l sys tem. Proc. R .Soc . L o n d o n A231 :91 -103 . H u a n g T .S . (1979) A fast two-dimensional med ian fi l tering a lgor i thm. I E E E Trans . A S S P 27:13-18. Ingram M . , a n d Preston K . J r . (1970) Automat ic A n a l y s i s of B l o o d Ce l l s . Scientif ic Amer i ca , 223:72-82. Inoue S. (1989) Founda t ions of confocal scanned imag ing i n l ight microscopy, i n Confocal Mic roscopy Handbook , ed., J a m e s Pawley, 1-13. Inoue S. (1986) Video Microscopy. P l e n u m , New York . J agg i B . Pontifex B . , Swanson J . , Poon S.S.S. [1993] Performance E v a l u a t i o n of a 12-Bi t , 8 M p e l / s Dig i t a l Camera . SPIE Proceedings, Cameras Scanners , a n d Image Acqu i s i t i on Systems, 1901:99-108. Jagg i B . , Poon S.S.S. , Pontifex B . , Fengler J . J . P . , M a r q u i s J . , a n d Pa lc ic B . (1991) A quanti tat ive microscope for image cytometry. SPIE Proceedings, C a m e r a a n d Input Scanner Systems, 1448:89-97. J agg i B . , Poon S.S .S . , Pontifex B . , Fengler J . J . P . , a n d Palc ic B . (1990) E v a l u a t i o n of a quanti tat ive microscope for image cytometry i n Advances i n Ana ly t C e l l Pathol . , eds., G . Burger et a l . Elsevier Science Pub l i she r s B . V . , Ams te rdam, 31-32. J a g g i B . , Poon S.S .S . , M a c A u l a y C , a n d Palc ic B . (1988) Imaging Sys t em for Morphomet r i c Assessment of Convent ional ly a n d F luorescent ly S ta ined Cel l s . Cytometry, 9:6. J agg i B . , Poon S., a n d Palc ic B . (1987) Opt ica l memory d i sks i n image data- base management for cytometry. App l i ed Opt ics 26(16):3325-3329. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 188 Jagg i B . , a n d Palc ic B . (1985) The Des ign a n d Development of a n Op t i ca l Scanner for C e l l Biology. I E E E Proceedings, Eng inee r ing i n Med ic ine a n d Biology, 2 :980-985. J a g g i B . , Poon S .S .S . , a n d Palc ic B . (1986) Implementa t ion a n d E v a l u a t i o n of the D m i p s C e l l Analyzer . I E E E Proceedings, Engineer ing i n Med ic ine a n d Biology, 3 :906-911. J i L . (1994) F u l l y automat ic chromosome segmentat ion. Cytometry 17:196- 208. Lansdorp P . M . , Poon S., Chavez E . , Dragowska V . , Z i j lmans M . , B r y a n T., Reddel R. , E g h o l m M . , Bacche t t i S., a n d Mar tens U . (1997) Telomeres i n the hematopoiet ic system. C I B A Founda t i on S y m p o s i u m No:211 . Telomeres a n d Telomerase. J o h n - W i l e y & Sons L t d . Chiches te r U . K . (in press). Lansdorp P . M . , Verwoerd N . P . , v a n de Rijke F . M . , D r a g o w s k a V . , Li t t le M . T . , D i r k s R . W . , Raap A . K . , and Tanke H . J . [1996] Heterogeneity i n telomere length of h u m a n chromosomes. H u m a n Molecu l a r Genet ics 5(5):685- 691 . Levy M . Z . , A l l sopp R . C . , Fu tcher A . B . , Greider C . W . , a n d Har le C . B . [1992] Telomere End- rep l i ca t ion Prob lem a n d C e l l Ag ing . 225 :954-960 . Lied tke C . E . , G a h n T., Kappe i F . , a n d Aeikens B . (1987) Segmentat ion of Mic roscop ic cel l scenes. Ana ly t i ca l a n d Quant i ta t ive Cytology and Histology, 9 :197-211. Locket t S . J . , a n d H e r m a n B . (1994) Automat ic detection of c lustered, f luorescent-stained nuc l e i by digi tal image-based cytometry. Cytometry 17:1-12. Locket t S . J . , O 'Rand M . , Rinehar t C , Kau fman D . , J a c o b s o n K , a n d H e r m a n B . (1990) Automated measurement of D N A plo idy u s i n g fluorescence microscopy i n Opt ica l Mic roscopy for Biology. H e r m a n B . a n d J a c o b s o n K . ed., Wi ley -L i s s , New York , 603-613 . L u n d b l a d V . a n d Wright W . E . (1996) Telomeres a n d Telomerase: Mee t ing Review. A s imple pic ture becomes complex. C e l l 87 :369-375 . M a c A u l a y C , a n d Palc ic B . (1988) A compar i son of some q u i c k a n d s imple th resho ld selection methods for s ta ined cells . A n a l y t i c a l a n d Quant i ta t ive Cytology, 10:134-138. M a n t a s J . (1987) Methodologies i n pattern recogni t ion a n d image ana lys i s - a br ief survey. Pat tern Recogni t ion, 20(1): 1-6. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 189 M a r r , D . (1982) V i s i o n , W . H . Freeman and Company , New Y o r k . Mar t ens U . M . , Z i j lmans J . M . J . M . , Poon S.S.S. , D r a g o w s k a V . , Y u i J . , Chavez E . A . , W a r d R . K . , a n d Lansdorp P . M . (1997) Shor t telomeres on the p - a r m of h u m a n chromosome 17. Nature Genetics (in press). M i n s k y M (1957) Mic roscopy Appara tus , U . S . Patent #3013467. Nederlof P . M . , v a n der Fl ier S., Raap A . K . , and Tanke H . J . (1992a) Quant i f ica t ion of inter- and in t ra-nuclear var ia t ion of f luorescence i n s i tu hybr id iza t ion signals . Cytometry 13:831-838. Nederlof P . M . , v a n der Fl ier S., Vro l i jk J . , Tanke H . J . , a n d Raap A . K . (1992b) Fluorescence ratio measurements of double- labeled probes for mul t ip l e i n s i tu hybr id iza t ion by digi ta l imaging microscopy. Cytometry 13:839- 845 . Nederlof P . M . , v a n der Fl ier S., Verwoerd N .P . , Vro l i jk J . , Raap A . K . , a n d Tanke H . J . (1992c) Quant i f ica t ion of fluorescence i n s i tu hybr id iza t ion s ignals by image cytometry. Cytometry 13:846-852. Nie l sen P . E . , E g h o l m M . , Be rg R . H . and B u c h a r d t O. (1991) Sequence-selective recogni t ion of D N A by s t rand displacement w i t h a thymine-subs t i tu ted polyamide . Science 244:1497-1500. Oh lande r R. , Price K . , a n d Reddy R. (1978) Pic ture segmentat ion u s i n g a recursive region spl i t t ing method. Compute r G r a p h i c s a n d Image Process ing, 8:313-333. Pa lc ic B . , M a c A u l a y C , Sc l ien S., Treurnie t W. , Tezcan H . , a n d A n d e r s o n G . (1992) C o m p a r i s o n of three different methods for au tomated class i f icat ion of cervical cells. Ana ly t i ca l Ce l lu la r Pathology. 4 :429-441 . Pa lc ic B . , Jagg i B . a n d Nord in J . A . (1987) D y n a m i c Microscope Image Process ing Scanner . U . S. Patent #4,700,298. Ph i l i p s J . a n d Lunds teen C. (1985) Semiautomated Chromosome A n a l y s i s : A C l i n i c a l Test, C l i n i c a l Genetics , 27:140-146. P l u t a M . (1988) Advanced Light Microscopy , Pr inc ip les a n d B a s i c Properties. Elsevier , New York . Pontifex B . Poon S., a n d Jaggi B . [1994] Des ign of a H i g h Reso lu t ion Programmable Dig i t a l Camera . SPIE Proceedings, Image A c q u i s i t i o n a n d Scientif ic Imaging Systems, 2173:130-139. Poon S .S .S . , a n d Hun te r D . B . [1994] Elec t ronic cameras to meet the needs of microscopy special is ts . Advance Imaging. 9(7):64-67, 84. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 190 Poon S .S .S . , W a r d R . K . , a n d Palcic B . (1993a) Au tomated image detection a n d segmentat ion i n b lood smears. Yearbook of M e d i c a l Informatics, 271 - 279 . Poon S .S .S . , Locket t S . J . , a n d W a r d R . K . (1993b) Charac te r iza t ion of a 3 D microscope imaging system. SPIE Proceedings, B i o m e d i c a l Image Process ing a n d B iomed ica l V i sua l i za t ion , 1905:121-128. Poon S .S .S . , W a r d R . K . , and Palc ic B . (1992a) Feature extract ion from three- d imens iona l images i n quanti tat ive microscopy. M i c r o n s a n d M i c r o s c o p i c a Ac ta , 23(4):481-489. Poon S .S .S . , W a r d R . K . , and Palc ic B . (1992b) A n a l y s i s of th ree-d imens iona l images i n quanti tat ive microscopy. SPIE Proceedings, B i o m e d i c a l Image Process ing a n d Three -Dimens iona l Microscopy , 1660:178-185. Poon S .S .S . , W a r d R . K . , a n d Palc ic B . (1992c) Au tomated image detection a n d segmentat ion i n b lood smears. Cytometry 13:766-774. Poon S .S .S . , a n d Palc ic B . (1991) Importance of focus i n image cytometry. Cytometry Supplement 5:40 (abstract: conference presentation). Poon S .S .S . (1989a) Algor i thms for detecting a n d segmenting nuc lea ted b lood cells . Thes is disertat ion, Un iv . of B r i t i s h C o l u m b i a . Poon S .S .S . , J agg i B . , Spadinger I., and Palc ic B . (1989b) F o c u s s i n g methods u s e d i n the cel l analyzer. I E E E Proceedings, Eng inee r ing i n Med ic ine a n d Biology, 11(2):754-756. Poon S .S .S . , J agg i B . , a n d Palc ic B . (1987) C e l l Recogni t ion A l g o r i t h m for the C e l l Analyzer . I E E E Proceedings, Engineer ing i n Med ic ine a n d Biology, 3 :1455-1456. P o u l i n N . , H a r r i s o n A . , and Palc ic B . (1994) Quant i ta t ive p rec i s ion of a n automated image cytometric system for the measurement of D N A content a n d d i s t r ibu t ion i n cells labeled w i t h fluorescent nuc le ic ac id s ta ins . Cytometry 16:227-235. P o u l i n N . , Poon S.S.S. , K a n d l a G . , and Palc ic B . (1989) Au toma t i c detection of metaphase chromosomes. I E E E Proceesings, Eng inee r ing i n Med ic ine a n d Biology, 11(2):350-351. Pres ton K . J r . (1976) Dig i ta l p ic ture analys is i n cytology, i n Dig i t a l P ic ture Ana lys i s , ed. Rosenfeld A . , Springer Ver lay , New Y o r k 209-293 . Rigaut J . P . (1991a) A new technology is born: confocal image cytometry. A n a l . C e l l . Pa thol . 3(2): 137-141. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 191 Rigau t J . P . , a n d V a s s y J . (1991b) High- reso lu t ion three-d imens iona l images from confocal s cann ing laser microscopy: quanti tat ive s tudy a n d mathemat ica l correct ion of the effects from b leach ing a n d fluorescence a t tenuat ion i n depth. Cytometry (submitted). Rigaut J . P . , V a s s y J . , H e r l i n P. , D u i g o u F . , M a s s o n E . , B r i a n e D . , Fouc r i e r J . , Carvaja l -Gonzalez S., Downs A . M . , a n d M a n d a r d A . M . (1991c) Three- d imens iona l D N A image cytometry by confocal s c a n n i n g laser mic roscopy i n t h i ck t issue b locks . Cytometry 12:511-524. Rosenfeld A . , a n d K a k A . C . (1982) Digi ta l Picture Process ing. Second E d i t i o n , V o l u m e 2. Academic Press, New York . R u s s J . C . (1990) Compute r Ass i s t ed Microscopy: The Measuremen t a n d A n a l y s i s of Images. P l e n u m Press, New York . R u s s J . C . a n d R u s s J . S . (1986) Image process ing for the loca t ion a n d i so la t ion of features, i n Mic robeam Ana lys i s , eds., R o m i g A . D . Chamber s W . F . S a n Franc i sco Press. 501 . Sheppa rd C . J . R . (1988a) Super- resolu t ion i n confocal imaging . O p t i k 80(2):53- 54. Sheppa rd C . J . R . (1988b) Dep th of field i n opt ica l microscopy. J . Mic roscopy 149(l ) :73-75. Sheppard C . J . R . (1988c) Aberra t ions i n h igh aperture convent iona l a n d confocal imaging systems. App l i ed Opt ics 27(22) :4782-4786. Sheppard C . J . R . a n d Cogswel l C . J . (1991) Effects of aberrat ing layers a n d tube length on confocal imaging properties. Op t ik 87:34-38. Shoemaker R . L . , Bar te ls P . H . , H i l l m a n D . W . , Jones J . , Kess ler D . , S h a c k R . V . , a n d Vukobra tov i ch D . (1982) A n ultrafast laser scanner microscope for digi ta l image analys is . I E E E Transac t ions on B i o m e d i c a l Engineer ing , 29(2):82-91. Spadinger I., Poon S.S.S. , a n d Palc ic B . (1990) Effect of focus on cel l detection a n d recogni t ion by the cel l analyzer. Cytometry 11:460-467. Spadinger I., Poon S.S.S. a n d Palc ic B . (1989) Au tomated detect ion a n d recogni t ion of live cells i n t issue cul ture u s i n g image cytometry. Cytometry 10:375-381. S tokse th P .A . (1969) Properties of a defocused opt ical system. J . Op t i ca l Society of A m e r i c a 59(10): 1314-1321. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 192 Tat i an B . (1965) Me thod for obta in ing the transfer funct ion from the edge response funct ion. J . Opt ica l Society of A m e r i c a 55(8): 1014-1019. T u c k e r J . H . , H u s a i n O . A . N . , Watts K . , Fa r row S., Bayley R., a n d S ta rk M . H . (1987) Au tomated densi tometry of cel l popula t ions i n a con t inuous - mot ion imaging cel l scanner. App l i ed Opt ics , 26(10):3315-3324. T u c k e r J . H . (1979) A n image analysis system for cervical cytology au tomat ion u s i n g nuc lea r D N A content. J o u r n a l of His tochemis t ry a n d Cytomet ry 21( l ) :613-620 . U m e s h R . M . (1988) A technique for cluster formation. Pat tern Recogni t ion , 21(4):393-400. V a n der Voor t H . T . M . , Brankenhof f G . J . , and J a n s s e n G . C . A . M . (1988) Dete rmina t ion of the 3-d imens iona l opt ical properties of a confocal s c a n n i n g laser microscope Op t ik 78(2):48-53. V a z i r i H , D r a g o w s k a W. , A l l sopp , R . C . , Thomas T . E . , Har ley C . B . , a n d Lansdorp P. M . [1994] Evidence for a mitot ic c lock i n h u m a n hematopoiet ic s tem cells: loss of telomeric D N A w i t h age. Proc. Nat l . A c a d . S c i . U S A , C e l l Biology, 91 :9857-9860. V i s se r T . D . , Brakenhoff G . J . and Groen F . C . A . (1991) The one-point f luorescence response i n confocal microscopy. Op t ik 87:39-40. V i s se r T . D . , O u d J . L . , a n d Brakenhof f G . J . (1992) Refractive index a n d ax ia l dis tance measurements i n 3-D microscopy. Op t ik 90:17-19. Vro l i jk J . , Sloos W . C . R . , Verwoerd N . P . , and Tanke H . J . (1994) App l i cab i l i t y of a noncooled video-rate C C D camera for detection of fluorescence i n s i tu hybr id iza t ion s ignals . Cytometry 15:2-11. Wa l ton W . H . (1952) Automat ic count ing of microscope part icles . Nature 169: Weins te in M . , a n d Cas t l eman K . R . (1971) Recons t ruc t ing 3-D specimens from 2 - D sect ion images. Proc. Soc. Photo-Opt. Ins t rum. E n g . 26 :131-138 . W i l s o n E . , Larosche T., a n d Gasser M . (1993) Telomeres a n d the func t iona l archi tecture of the nuc leus . Trends C e l l B i o l . 3 :128-134. W i l s o n T. a n d Sheppard C . J . R . (1984) Theory a n d practice of s cann ing opt ica l microscopy. Academic Press, London . W i l s o n T., a n d Hewlett S . J . (1990) Imaging i n s cann ing microscopes w i t h sl i t - shaped detectors. J . Mic roscopy 160(2): 115-139. Y o u n g J . Z . , a n d Roberts F . (1951) A flying spot microscope. Nature 167:231. TELOMERE LENGTH MEASUREMENTS USING FLUORESCENCE MICROSCOPY 193 Y o u n g T . Y . , a n d F u K . S . , Eds . (1986) Handbook of Pat tern Recogni t ion a n d Image Processing. Academic Press, London . Z i j lmans J . M . J . M . , Mar tens U . M . , Poon S .S .S . , Raap A . K . , Tanke H . J . , W a r d R . K . , a n d Lansdorp P . M . (1997) Telomeres i n the mouse have large in te r -chromosomal var ia t ions i n the n u m b e r of T 2 A G 3 repeats. Proc. of Na t iona l Academy of Science U S A , Genet ics 94 :7423-7428 .

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
China 21 26
United States 6 0
Russia 3 0
Iran 1 1
Japan 1 0
City Views Downloads
Beijing 19 0
Mountain View 3 0
Shenzhen 2 26
Unknown 2 1
Ashburn 2 0
Ryazan 2 0
Sunnyvale 1 0
Tokyo 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items