UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Detection and localization of pre-cancerous lesions and early lung cancer using tissue autofluorescence Hung, Yip-Chan Jacyln 1992

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1992_fall_hung_jaclyn_yip_chan.pdf [ 2.25MB ]
Metadata
JSON: 831-1.0085057.json
JSON-LD: 831-1.0085057-ld.json
RDF/XML (Pretty): 831-1.0085057-rdf.xml
RDF/JSON: 831-1.0085057-rdf.json
Turtle: 831-1.0085057-turtle.txt
N-Triples: 831-1.0085057-rdf-ntriples.txt
Original Record: 831-1.0085057-source.json
Full Text
831-1.0085057-fulltext.txt
Citation
831-1.0085057.ris

Full Text

DETECTION AND LOCALIZATION OF PRE-CANCEROUS LESIONS AND EARLY LUNG CANCERUSING TISSUE AUTOFLUORESCENCEbyJaclyn Yip-Chan HungB.Sc., The University of Prince Edward Island, 1985M.Sc., The University of British Columbia, 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of PhysicsWe accept this thesis as conformingto the required standardflTHE UNIVERSITY OF BRITISH COLUMBIAJune 1992© Jaclyn Yip-Chan Hung, 1992Signature(s) removed to protect privacyIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for schola1y purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department ofPhysics(SignatureThe University of British ColumbiaVancouver, CanadaDate iu1v 23rd 1992DE-6 (2/88)Signature(s) removed to protect privacy11ABSTRACTIn this work, two different yet related hypotheses were tested byexperimental means as follows: i) pre-cancerous and non-invasive (early)lung cancer can be detected and localized using the fluorescence propertiesof tumour localizing drugs at non-photosensitizing doses to skin tissue;ii) significant differences exist in laser-induced autofluorescence betweennormal, pre-cancerous and cancerous tissues such that these differencesalone can be exploited to detect and delineate early lung cancer withoutusing exogenous drug(s).For most cancers, including lung cancer, a five-year survival of over90% can be realized if the malignancy is diagnosed at the carcinoma in situstage. However, current techniques: chest x-ray, sputum cytology, andconventional bronchoscopy alone or in combination cannot detect these verysmall lesions which are usually only a few cell layers thick and a fewmillimeters in surface diameter. Exogenous fluorescent tumour markers suchas hematoporphyrin derivatives (e.g. Photofrin) have been used to enhancethe detection of these occult lung lesions. Photofrin is preferentiallyretained in tumour tissues compared to the surrounding normal tissues; itfluoresces at 630 rim and 690 run when excited at —405 rim. based on thisprinciple several imaging and non-imaging devices have been developed.However, wider clinical applications were limited due to the skinphotosensitivity property of Photofrin. We have postulated that this couldbe solved by employing a much lower dose of photofrin (0.25 mg/kg) whichwas believed to be less photosensitizing to human patients. This postulatewas experimentally tested by ratio fluorometry and early lung cancers wereiiidetected with no false negative results and no apparent skinphotosensitivity. An important finding in this study was that themechanism for detection of early cancer was mainly due to the differencesin the green autofluorescence between normal and malignant tissues, ratherthan fluorescence of tumour localizing drug.This discovery led to the second postulate of this thesis that tissueautofluorescence alone can be exploited for the detection of early lungcancer. In vivo spectroscopy using an optical multi-channel analyzershowed an overall decrease in autofluorescence in pre-cancerous andcancerous lesions when excited by 405 nm or 442 nm laser light. A stepwisediscriminant function analysis was performed on a database of nearly 300patients spectra to determine the optical emission wavelength(s) andalgorithm(s) at which the normal, pre-cancerous and cancerous tissues canbe differentiated. The results indicated that algorithm(s) could bedeveloped to clearly delineate early lesions from the normal tissues.Several algorithms were then tested using a non-imaging ratio fluoronieterdevice and a prototype imaging fluorescence system to detect early lungcancer and dysplasia during standard bronchoscopy, therefore confirming theinitial hypotheses even in a clinical setting.The major source of the autofluorescence in the normal bronchialtissue was determined to come from the sub-epithelial layers. Themechanism for the decrease in the autofluorescence in pre-malignant andmalignant tissues was explored but not yet completely elucidated. Severalfactors such as decrease in fluorophores, increase in absorption of theexcitation/fluorescent light or different redox state of the fluorophoresmay be responsible for the observed phenomena.ivTABLE OF CONTENTSABSTRACT.iiTABLE OF CONTENTS ivLIST OF FIGURES viiLIST OF TABLES ixACKNOWLEDGEMENTS x1. INTRODUCTION 11.1 Outline 11.2 Lung Cancer 21.2.1 Incidence and Mortality 21.2.2 Classification of Lung Cancer 41.2.3 Natural History of Lung Cancer 51.3 The Bronchial Wall 82. EARLY DETECTION 112 . 1 Introduction 112.1.1 Chest Radiograph and Sputum Cytology 112.1.2 Biochemical, Immunological and Molecular Markers 132.1.3 White Light Bronchoscopy 142.2 Diagnostic Applications of Fluorescent Tumour Markers 152.2.1 Fluorescent Tumour Markers: Hematoporphyrin 162.2.2 Hpd and Its Active Components 172.2.3 Photo-Physical Properties 182.3 Tumour Detection Using Hpd or Photofrin 192.3.1 Mechanism For Detection 202.3.2 Reviews of Fluorescence Endoscopic Detection System.... 21V2.4 Previous Work Using Tissue Autofluorescence In Diagnosis ... 273. LIGHT INTERACTION WITH TISSUES 313.1 Photo-Biological Response 313.2 Optical Properties of Turbid Media: Tissue 323.3 Light Propagation in Tissue: Theory 403.4 Measurements of Tissue Optical Properties 463.5 Photo-Physical Process Following Light Absorption 483.5.1 Fluorescence 503.5.2 Characteristics of Fluorescence 534. EXPERIMENTAL DESIGN AND RESULTS 554.1 Thesis Objective 554.2 Detection of Early Lung Cancer Using Non-SkinPhotosensitizing Dose of Photofrin 554.2.1 Rationale 554.2.2 Mechanism for Detection by Ratio Fluorometry 624.3 Fluorescence Spectroscopy 654.4 Detection of Dysplasia and Carcinoma in situ UsingNon-Imaging and Imaging Methods Using TissueAutofluorescence Alone 864.4.1 Non-Imaging Methods 864.4.2 Fluorescence Imaging 875. DISCUSSION AND CONCLUSION 995.1 Tumour Detection Using Non-Skin Photosensitizing Doseof Photofrin 1005.2 Tissue Autofluorescence in Diagnosis 1015.3 Mechanism for Difference in Tissue Autofluorescence 110vi5.3.1 Preliminary Data 1155.3.2 Differences in Autofluorescence in Other Tissue 1175.4 Development of Lung Imaging Fluorescence Endoscope 1235.5 Direction for Future Studies 1246. REFERENCES 1277. APPENDIX 1477.1 Discriminant Function Analysis 1477.2 In Vitro Spectroscopy 1567.3 Optical Multi-Channel Analyzer 162viiLISTS OF FIGURESFigure 1 Propagation of light in bronchial tissue 35Figure 2 Photo-physical process following light absorption 52Figure 3 Schematic diagram of the ratio fluorometer probe 57Figure 4 Red-green ratios of pathologically confirmedareas from 20 patients who had received 0.25 mg/kgPhotofrin i.v 61Figure 5 Red (690 nm ± 10 nm) fluorescence intensity oftumours as a ratio of the normal background withand without Photofrin 63Figure 6 Green (520 rim ± 10 rim ) autofluorescence intensityof tumours as a ratio of the normal background withand without photofrin 64Figure 7 Schematic diagram of the spectroanalyzer system 68Figure 8 In vivo autofluorescence spectra from a normalbronchus and carcinoma in situ lesion excitedat 405 run 69Figure 9 In vivo autofluorescence spectra from a normalbronchus, dysplasia and carcinoma in situ lesionexcited at 442 rim 70Figure 10 The mean curves for each class normalized in thered region (600 rim - 660 nm) 76Figure 11 The mean curves for each class normalized in thegreen region (530 rim - 570 run) 77Figure 12 The normalized (500 rim - 660 run) average intensityas a function of wavelength for each class 78Figure 13 The unnormalized mean curves for each class 79Figure 14 The p-values for each class normalized in the redregion (600 rim - 660 rim ) at 10 rim interval 80Figure 15 The p-values for each class normalized in the greenregion (530 run - 570 rim ) at 10 rim interval 81Figure 16 The normalized (600 rim - 660 rim) average intensitycurves for each class as a function of wavelength 82viiiFigure 17 Red-green ratios from pathological confirmed areas of82 volunteers 88Figure 18 smoking status and prevalence of lesions usingfluorescence bronchoscopy (ratio fluorometer) 89Figure 19 Schematic diagram of a clinical prototype lungimaging fluorescence endoscope (LIFE) 93Figure 20 Fluorescence (a) and white-light (b) image of anormal bronchus 94Figure 21 Fluorescence (a) and white-light (b) image of abronchus with an area of carcinoma in situ lesion 95Figure 22 The prevalence of dysplasia and carcinoma in situlesions using fluorescence bronchocopy (LIFE) 97Figure 23 The non-linear discriminant function values of thered and green intensity for the pathology grades 98Figure 24 In vivo fluorescence spectra from a normal and tumourarea with and without Photofrin 106Figure 25 Effects of freezing and thawing on the autofluorescenceof a normal bronchus fragment 107Figure 26 Effect of freezing and thawing on the autofluorescericeof an excised tumour 108Figure 27 Autofluorescence spectra from a normal site and acarcinoma in situ lesion of a hamster cheekpouch model 109Figure 28 Fluorescence spectra of FAD, FMN and riboflavin 118Figure 29 Fluorescence spectra of collagen IV 119Figure 30 Fluorescence spectra (a) and fluorescence image (b)of basement membrane material excited at 442 rim 120Figure 31 Autofluorescence of a freshly excised human bronchusintact, with it epithelium layer and the submucosalayer removed 121Figure 32 Examples of a quadratic (a), linear (b) bayesiandecision boundary for a two dimensional, two-group andtwo feature case 154Figure 33 Schematic diagram of the set-up for the transmittanceand reflectance measurements using an integratingsphere 160ixLISTS OF TABLESTable I Classification table of the autofluorescence spectra of284 total cases according to the histopathologic gradesand organized into data sets each consisting of onlyone spectrum class 75Table II Autofluorescence spectra features for discriminantfunction analysis 83Table III Summary of the performance of the discriminant functionanalysis on the raw data 84Table IV Characteristics of the subjects for the LIFE study 96Table V Summary of the optical properties of human bronchialfragments and the hamster cheek pouch 122xACKNOWLEDGEMENTSThis thesis could not have been completed without the generouscontributions of many people. I regret the impossibility to mention themall by name. I would especially like to thank my thesis supervisor Dr.Branko Palcic for being so patient with me in the completion of thisthesis. I thank him for his encouragement, guidance and endless enthusiasmduring the course of this work.I owe a special debt of thanks to Dr. Stephen Lam for countless hoursof interesting discussions, advice and for sharing his ideas and knowledgewith me. In particular, I would like to express my sincere thanks to himfor all his contributions to this work. This thesis could not have beencompleted without him making the clinical material available. I also liketo thank him for his excellent supervision and inestimable patience duringthe course of this research.Special appreciation goes to Alfred Pon for his excellent engineeringsupport. I would like to acknowledge the advice of Dr. Caluin MacAuley andDr. Ingrid Spadinger in the discriminant function analysis. I would alsolike to thank Susan Crose for her excellent secretarial support.I am grateful to my family and those many friends who have helped seeme through these years.Finally, my most personal thanks I owe to my parents, for theirunfailing support throughout the course of my studies.1CHAPTER 1: INTRODUCTION1.1 OutlineThe investigations were carried out to understand the mechanisms ofearly lung cancer detection within the context of bronchoscopy. Thisthesis addressed the hypothesis that significant differences inautofluorescence exist between normal, pre-cancerous and early cancerousbronchial tissues.The text of this thesis is divided in several chapters. First, thebackground and rationale for this thesis are presented. In Chapter 2, thelimitation of current detection techniques and the use of exogenousfluorescent tumour markers, in particular Photofrin are described. A briefreview of the imaging and non-imaging devices is included in this chapteras well as the diagnostic applications of autofluorescence. Chapter 3reviews the optical process associated with light-tissue interactions,radiative transport of light in tissues, and optical properties of tissues.The physical processes following light absorption, in particular theprinciples of fluorescence are also described. In chapter 4, theobjectives, experimental design and results are shown. In the lastchapter, the hypothesis(es) for the mechanism of the observed difference intissue autofluorescence between normal, pre-cancerous and cancerousbronchial tissues is presented and discussed. In addition, a summary ofthe project is given and future implications of these are considered.Several aspects of this thesis have been described in scientificpapers and abstracts presented and published as listed in the references.21.2 Lung Cancer1.2.1 Incidence and MortalityLung cancer was a relatively rare disease at the beginning of thiscentury, but is now the leading cause of cancer-related death. The worldwide incidence of lung cancer is increasing at a rate of 0.5% per year(IARC (WHO), 1990), and with an estimated 660,500 new cases yearly, it isexpected to become the predominant cancer world-wide by the end of thiscentury (Parkin et al., 1988). Cigarette smoking is the main etiologicalagent, but dietary effects, exposure to other carcinogens and predisposinggenetic factors have also been associated with lung cancers (Henderson etal., 1991, Weinstein, 1991, Law, 1990, Doll and Peto, 1981,). Smokingalone or acting in concert with other risks is the cause of about 80% ofthe lung cancers today (Shoppand, 1991, IRAC (WHO), 1990, Loeb et al., 1984for review of the literature).It has been estimated that in the United States alone, more than amillion lung cancer deaths will occur before the year 2000 even if alltobacco smoking ceases (Mulshine, et al., 1989). This is because of theexisting pool of current and ex-sniokers and that lung cancer is due to theaccumulative effect of smoking over a period of twenty years or more. Forex-smokers, although the risk of developing lung cancer is reduced aftersmoking cessation, the risk remains elevated for about fifteen years aftersmoking cessation compared to the risk of a lifetime non-smoker (Shopand,1990, IRAC(WHO), 1990). In addition, the overall upward trend in lungcancer incidence continues due to the current prevalence of smoking amongthe female population (Stolley, 1983). In women, lung cancer deaths now3exceeds that attributed to breast cancer (American Cancer Society, 1991).The recent increase in smoking among women will further compound thecurrent situation and its accumulated effect on lung cancer will bemanifested for another 25 years or more (Brown and Kessler, 1988). InBritish Columbia alone, an estimated 1,118 new lung cancer cases will occuramong women in the year 2001, compared to the approximately 600 in 1986,representing more than 80% increase within 15 years (McBride et al., 1989).In less developed countries where smoking continues to prevail, thedevastating effect of lung cancer is likely to continue (Stanley andStjernswàrd, 1989, Stjernswàrd, 1989a). By the year 2025, the World HealthOrganization estimated that there will be 3.5 million lung cancer cases peryear globally (Stanley and Stjernsward, 1989).Despite the growing public awareness of the health risk of smokingand some encouraging progress in the reduction in cigarette smoking inrecent years, the occurrence of lung cancer will remain a major healthissue for decades (Stjernsward, 1989b). Although prevention programmestargeted at tobacco products have resulted in the reduction of lung cancerincidence among young people in the developed countries (Shoppand, 1991,Devesa et al., 1989), the impact of an increased use of new addictiveproducts such as marijuana and cocaine on future lung cancer incidence isnot known yet (Stjernswãrd, 1989a, Dupont, 1980).The high incidence of lung cancer is unfortunately associated with ahigh mortality rate. Less than 15% of patients with lung cancer survivethe disease and this figure has not changed significantly over the pastthree decades. The reason for this is that lung cancer is an insidiousdisease producing few symptoms until the late stages of its development.4At the time of diagnosis, the majority of the patients have advancedinvasive cancer so that curative treatment is no longer possible (Mulshineet al., 1991, Minna et al., 1989, Fontana et al., 1986, Jett et al., 1983).Attempts in early detection using chest x-ray and sputum cytologyexamination have not been shown to reduce lung cancer mortality (Fontana etal., 1991 Fontana et al., 1986).1.2.2 Classification of Lung CancerThe World Health Organization classifies lung tumours into four majorhistologic subtypes: squamous cell carcinoma, adenocarcinoma, large cellcarcinoma, and small cell carcinoma (WHO, 1982). These subtypes accountfor about 90% of the primary lung cancers. The first three histologicsubtypes (squamous, large cell and adenocarcinoma) are collectively groupedtogether and referred to as non-small cell lung cancer (NSCLC) and thetreatment approaches for these tumours are identical. NSCLC accounts forabout 75% of lung cancer cases. Small cell lung cancer (SCLC) is diagnosedin 20 to 25% of lung cancer and is a very aggressive tumour withneuroendocrine features. It is the most common histologic type of lungcancer found amongst uranium miners (Auerbach et al., 1978). SCLC isinitially responsive to combination chemotherapy and radiotherapy, but mostpatients will eventually die of recurrent metastatic disease (Ihde andMina, 1991, Minna et al., 1989, Seifter and Ihde, 1988). There is a highincidence of second primary cancers among long-term survivals of SCLC; themajority of the second primary cancers are NSCLC (Thomas et al., 1991).51.2.3 Natural History of Lung CancerThe carcinogenesis and pathogenesis of cancer is a complex multistage process (Fearon and Vogelstein, 1990, Harris, 1991). In general,malignant cancer is a consequence of genetic damage and other changesresulting in uncontrolled cellular proliferation and transformation(Strauss, 1992). Genetic (DNA) damage may be acquired somatically, inducedby environmental and chemical carcinogens, or alternatively, predisposinggenetic factors may exist, and this may involve gene-environmental/chemicalinteractions. Exposure to a cancer causing agent does not lead to animmediate production of a tumour. The specificity of the mutations thathave been observed in tumours suggests that induction of some of thesemutations are a result of several alterations in the DNA (Vogelstein etal., 1988). Most damage sustained by DNA can be repaired. However therepair process is believed to be gene specific and is associated withcancer susceptibility (Link et al., 1991). The distribution of adductformation (damage to DNA) and repair in the genome is non-random and isinfluenced by its DNA organization and structure (Link et al., 1991,Pienta, 1989, Topal, 1988, Bohr et al., 1987). A number of important non-or pre-cancerous adaptive growth changes may occur in response to theinitial primary step in the carcinogenesis process. Most of these “early”lesions may disappear spontaneously. The primary event is likely to be thedamage to a specific gene in the cell DNA (Philips et al., 1988). Furthersteps may involve more genetic damage in other genes stimulating thesecells to grow thereby conferring a selective advantage to these cells(Strauss, 1992, Loeb, 1991). There is a latency period for the developmentand/or progression of cancers (Sporn, 1991). For example, it has been6documented that the time between the onset of habitual cigarette smokingand the development of a clinically apparent tumour can be as long asseveral decades. There is evidence that several pathogenic events mayoccur during this latency period. Progressive alterations in therespiratory epithelium of cigarette smokers have been clinically observed.In the case of centrally arising lung cancers (squamous cell carcinoma),three types of lesions may exist: i) precancerous lesions (steps includehyperplasia, metaplasia, increasing degree of dysplasia); ii) malignant butnon-invasive lesions (carcinoma in situ); and iii) invasive carcinomas thatmay potentially develop to a further degree of malignancy (Greenberg etal., 1987, McDowell and Trump, 1983, Saccomanno et al., 1974). Thesequences of events for the development of peripherally arisingadenocarcinoma and large cell carcinoma is still not fully established. Atthe present time it is suggested that in situ adenocarcinoma probably doesnot exist (Matthews and Linnoila, 1988).It is difficult to evaluate human precancerous lesions in terms ofthe underlying carcinogenic process because so much of our knowledge oftumour development in human is indirect and inferential. To follow orassess these pre-malignant changes as they evolve is difficult because theassumed risk for progression is high enough to motivate removal of theselesions once they are detected. The nature of pre-malignant changes hadbeen addressed by means of experimental animal model systems such as theinjection of 20-methyicholanthrene into the submucosa area of beagle dogsto follow the development of invasive carcinoma endoscopically (Kato etal., 1982). Epidemiological and histological studies of lung cancer inhumans showed that the cytological changes induced by benzo(a)pyrene-ferric7mixture in hamsters, 20-methylcholanthrene in beagle dogs are similar tothat found in cigarette smokers (see below) with lung cancer (Kato et al.,1982, Becci et al., 1978, Schreiber et al., 1974).The most comprehensive work done on humans and much of what is knownabout the development of lung cancer is from the studies by Saccomanno andco-workers using sputum cytology samples of more than 5000 uranium minersover a thirty-year period (Sacconiniano et al., 1974, Saccomanno, 1982).These studies suggested that lung cancer develops over a period of 10 to 20years with a series of progressive histopathologic changes, particularlyfor squamous cell carcinoma. These changes include a less specific basalcell hyperplasia followed by squamous metaplasia, increasing degree ofatypia from mild, moderate to severe dysplasia and then final developmentof carcinoma in situ.Dysplasia is a disturbance of differentiation of squamous epitheliumwhich may be characterized as mild, moderate or severe. The biologicalactivity of these lesions is uncertain with some regressing, othersremaining stable and yet others progressing to epidermal carcinoma in situ.The term dysplasia is architecturally descriptive, relating to therelationship between cells within an epithelium. These changes are usuallyaccompanied by cytologic changes referred to as dyskaryosis. Carcinoma insitu refers to an epithelium within which no differentiation takes placethroughout its whole thickness, i.e. with the exception of some slightflattening in the upper most layers, the cells on the surface are identicalto the cells in the basement layer. Histologically, the alteration withinthe surface epithelium resembles squamous carcinoma but there is noevidence of invasion of the underlying fibromuscular stroma.81.3 The Bronchial Wall*The pulmonary airways are a complex system of branching tubes and arein constant direct contact with gases or particulate matters in theexternal environment that may be cytotoxic. The human bronchus is dividedinto extrapulmonary and intrapulmonary bronchi. The extrapulmonary bronchiconsist of the right and left main bronchi and the truncus intermedius.The right middle lobe and the left and right upper lobe prior to thebifurcation of the segmental bronchi is intrapulmonary in terms ofstructure. The extrapulmonary bronchus has horse-shoe shaped cartilagecrescents which extend around 1/2 to 2/3 of the circumference. Elasticfibres extend around the entire circumference and reach a thickness ofabout 0.008 mm in the membranous portion forming longitudinal folds. Theposterior wall of the bronchus is referred to the membranous portion and isfree of cartilage crescents but has a large amount of smooth muscle.Cartilage crescents are absent in the intrapulmonary bronchi. The layer ofelastic fibre between the epithelium and the submucosa is graduallyreplaced by smooth muscle which extend in rings surrounding the entirecircumference of the bronchus. In the intrapulmonary bronchi, the finecircular folds are formed by thin rings of smooth muscle. The ridges seenrunning in longitudinal directions are the elastic fibres that areconsolidated into bundles at intervals between the epithelium and thesmooth muscle all around the bronchus.When viewed tangentially the wall of an airway can be divided intotwo layers: epithelium, submucosal and cartilage. The epithelial surface*References used in preparation of section 1.3 were Forrest and Lee, 1991,Oho and Amemiya, 1984, and the Atlas of Early Lung Cancer.9throughout the bronchial tree is normally pale pink and semi-transparent inappearance. The very thin surface layer of mucus in which the cilia beatgives the mucosa its characteristic sheen. However, the presence ofsquamous cell metaplasia, dysplasia or cancer causes alteration of therefractive index and may then appear opaque, white or granular.The epithelial cells are attached to the basement membrane by finecytoplasmic foot processes. The lamina propria lies below the basementmembrane and is separated by an elastic tissue layer. The basementmembrane and lamina propria separate the mucosa from the submucosa. Thesurface of the trachea and bronchi are lined by a pseudostratified columnarepithelium which consists of 4-5 principal cell types: superficiallylocated ciliated and non-ciliated columnar cells, basal cell or reservecells, and mucin-secreting goblet cells. Pseudostratified implies that allcells rest on the basement membrane and extend the full thickness of theepithelium. The ciliated columnar cells comprised about 85% of therespiratory epithelium; these cells are terminally differentiated andincapable of cell division. The basal cells and the rare basally locatedendocrine cells, the so called Kulchitsky cells with their dense coregranules and features of neuroendocrine cells are scattered along but restdirectly on the basement membrane. These basal cells gives thepseudostratified appearance of the bronchial epithelium because they do notreach the surface of the airway.The submucosa consists principally of mucous glands, smooth muscleand cartilaginous plates. In the large bronchi, submucosal glandspenetrate the submucosa and open by ducts at the mucosal surface. Thesubmucosal bronchial gland consists of mucous cell, serous cell and10Kulchitsky cells. The smooth muscle varies in size and orientation alongthe airways forming a discontinuous layer in the innermost portion of thesubmucosa and spiraling in both directions of the airways. Theextrapulmonary airways have the highest amount of smooth muscle anddecrease progressively between main-stem bronchi and bronchioles.Cartilage is absent in the small bronchi and bronchioles. The connectivetissues, particularly the cartilage, confer structural support on theairways. Nerves, blood vessels and lymphatics are found in the nondistinctive adventitial layer of the airway.11Chapter 2: EARLY DETECTION2.1 IntroductionPrimary prevention remains the key to the long-term control of lungcancer. As discussed in chapter 1, even if World Health Organization (WHO)objectives to reduce tobacco smoking can be realized, lung cancer willremains a significant health problem well into the next century. Despitethe efforts to improve therapy for advanced lung cancer, the efficacy ismodest at best (Minna et al, 1989, Mulshine et al, 1989). Therefore,increased efforts in secondary prevention and treatment of early lungcancer may be the main element in the effective management of this disease.The key to this approach is the development of new diagnostic methods todetect pre-cancerous and in situ carcinoma lesions in high risk individuals(Mulshine et al., 1991).2.1.1 Chest Radiograph and Sputum CytologyIt has been established that long term survival for lung cancerpatients can be accomplished if the tumour is detected and localized at thecarcinoma in situ or microinvasive stage. A five-year survival of 90% orbetter can be achieved in this group of patients after surgical resection(Melamed et al., 1984).It is very difficult to detect and precisely localize lung cancersthat are still confined to the bronchial wall. Chest x-ray and CT scansare usually negative (Frost et al., 1983, Spiro et al., 1982). Cancersdetected by sputum cytology examination with an apparently normal chest xray are traditionally referred to as x-ray occult carcinoma (Miller et al.,121983). Most lung cancers are not detected on a conventional radiographyuntil the tumour has reached a size of approximately 1.0 cm3. By thistime, approximately 30 doubling of the tumour volume would have alreadyoccurred. Death usually occurs when the tumour has undergone 40 doublings.Since many lung cancers are not diagnosed until the tumour is 3 cm - 4 cmin diameter, it is not surprising that conventional treatments such assurgery, radiotherapy and chemotherapy are not very effective (Line et al.,1989).In the late 1960’s Saccomanno proposed the use of sputum cytology asa technique to enhance the detection of early lung cancer. Sputum cytologyevaluates cytomorphologic changes in exfoliated bronchial epithelial cells.It was suggested that sputum cytology would provide the earliest possibleindication of lung cancer (Hjerpe, 1988, Saccommano et al., 1974). In the1970’s, in an attempt to improve the detection of early lung cancers, theNational Cancer Institute of United States sponsored three prospective,randomized trials of lung cancer screening with chest radiographs andsputum cytology among 31,360 men (Cooperative Early Lung Cancer DetectionProgram) to evaluate the usefulness of chest x-ray in combination withsputum cytology in detecting early lung cancer. The end point for thesetrials was mortality rate from lung cancer. The men screened were allcigarette smokers who smoked at least one pack or more of cigarettes a day,45 years of age or over and asymptomatic at the time of entry into thestudy (Berlin et al., 1984). Results of these trials suggest more earlystage lung cancer can be detected in the asymptomatic smokers by periodicscreening. However the overall mortality from lung cancer was not loweredprobably due to a lead-time bias (Fontana et al., 1991, Mulshine et al.,131989). The study also showed that the yield of early lung cancer by sputumcytology was low; the prevalence of carcinoma in situ or microinvasivecancer was only 0.1%. The reason for this is that sputum cytology mainlydetects centrally located squamous cell cancers that readily exfoliatecells into the bronchial secretion. However, it is less sensitive for thedetection of adenocarcinomas and large cell carcinomas which tend todevelop more peripherally in smaller airways where they obstruct thebronchial lumen quite early. Such cells are not exfoliated into the lumenuntil they have invaded the larger central airways (MacMahon et al., 1988,Fontana et al., 1986, Frost et al., 1983).2.1.2 Biochemical, Immunological and Molecular MarkersAttempts to find biochemical, immunological, and molecular markersthat can enhance the sensitivity of sputum cytology to detect lung cancerat its earliest stage are still at their infancy (Uei and Noguchi, 1990,Gazdar et al., 1990, Humphrey, 1989, Gail, 1988, Inoue et al., 1988).Immunocytochemical studies of antigens expressed on the cell surface ofpre-malignant cells by monoclonal antibodies directed to a glycolipidantigen of small cell carcinoma and a protein antigen of non-small celllung cancers (Tockman et al., 1988) showed encouraging results. Amongthose subjects with positive immunostaining of their sputum cells, 91%eventually developed lung cancer within 2 years. In contrast, 88% of thosesubjects who had negative immunostaining remained free of cancer. Afollow-up study had been proposed to validate this immunostaining approachand to screen patients with Stage I resected lung cancer who represent agroup with a high risk for second primary cancers (Mulshine et al., 1991,14Muishine et al., 1989). However, even if these and other techniques canbe proven to be of value in early detection, localization of these smallmalignant or pre-malignant lesions will remain a major problem. None ofthese techniques provide any information regarding the location of theabnormal cells in the upper or lower respiratory tract. To localize thesource of the abnormal cells and to clearly delineate the size and extentof these lesions, bronchoscopic examination is generally required.2.1.3 White Light BronchoscopyConventional white light bronchoscopy is currently the only methodthat can localize x-ray occult lesions within the tracheobronchial tree(Marsh et al., 1989). The current approach is visual examination of thetracheobronchial tree under local anesthesia. Since occult lung cancersmay be present in more than one site, a careful search of all the bronchialsegments is essential (Marsh et aL., 1983). Bite biopsies, bronchialbrushing, and washings are performed on any suspicious or abnormal areas toconfirm the presence or absence of cancer by cytology or histopathology.In about 25% - 30% of cases, a repeat bronchoscopic examination undergeneral anesthesia with multiple brushings or blind biopsies may berequired before the abnormal cells seen in sputum cytology examination canbe localized (Balchum et al., 1982). A study by the Mayo Clinic showedthat carcinoma in situ was visible bronchoscopically in less than 30% ofthe cases (Woolner, 1983). A recent follow-up study by the Lung TumourStudy Group in patients with completely resected T1NOMO lung cancerssuggested that conventional bronchoscopy prior to surgery likely underestimated the local extent of the primary cancer and missed areas of15significant dysplasia or carcinoma in situ in 10-25% of the patients(Thomas et al., 1990). The reason for this is that early cancers are onlya few cell layers thick (0.2 mm to 1 mm) and on average less than 0.8 cm insurface diameter (Woolner et al., 1984). These small relatively flatlesions may not produce enough changes on the bronchial surface to allowdetection by conventional white-light bronchoscopy. In some cases of insitu carcinoma, subtle niucosal abnormalities such as an increase inredness, granularity or mucosal thickening can be recognized as signs ofabnormality. Unfortunately, these changes can also be associated withinflammatory airways diseases making it very difficult for correctdiagnosis (Oho and Ainemiya, 1984).2.2 Diagnostic Applications of Fluorescent Tumour MarkersBecause of the difficulty in detecting and localizing in situcarcinomas, there has been great interest in the use of fluorescent drugsto enhance the diagnostic accuracy of conventional bronchoscopy. The useof fluorescent drugs to “mark” or “tag” cancerous cells has been availablefor many years. Dyes and drugs such as acridine orange, tetracycline,berberine sulfate, toluidine blue, fluorescein, eosin, and porphyrins havebeen shown to selectively or preferentially accumulate in malignant tuinours(Dougerty, 1974, Tomson et al., 1974, Wilbanks and Richart, 1970, Rail,1957, Moore, 1947). The diagnostic potential of such drugs which arepreferentially accumulated in tumor tissues is obvious. In principle, itshould be possible to detect and delineate the tumour by visualization ofthe fluorescence emitted by these tumour markers under proper excitations.162.2.1 Fluorescent Timtour Markers: HematoporphyrinAlthough a number of fluorescent tumour markers are potentiallyuseful for localization of tumours, much of the clinical and laboratorystudies have focused on the use of porphyrins. The history of fluorescencedetection of cancer using certain porphyrin compounds dates back to atleast 1924 when Policad observed that certain malignant tumours in animalsand man emitted a reddish fluorescence due to the accumulation ofendogenous porphyrins (Policad, 1924). In 1942, Auler and Banger observedfluorescence in implanted tumours in rats but not in normal tissuesfollowing the systemic injection of hematoporphyrin (Hp) (Auler and Banger,1942). In 1948, investigations by Figge, Weiland and Manganiello not onlyconfirmed this observation in a wide range of induced or transplantedtumours in mice, but also found fluorescence in embryonic, lymphatic, andtraumatized regenerating tissues as well (Figge et al., 1948). When Hp wasinjected into animals, a brilliant red-orange fluorescence was produced byultraviolet excitation of the drug localized in these tissues. Hp is madefrom hemoglobin but contains no iron. It was first prepared in 1867 andwas then called cruentene. It was later renamed as hematoporphyrin (Hp) ataround 1871 and was prepared by treating hemoglobin with sulfuric acidfollowed by extraction with alcohol. In the 1950’s, several unsuccessfulclinical testings were made to demonstrate HP fluorescence in head and neckcancers, the biliary tree and lymph nodes in humans, (Rassmussen-Taxdal etal., 1955). The clinical procedures required a very large intravenousdosage (300 to 1000 mg per patient) of Hp to emit sufficient redfluorescence for the tumour to be detectable. At this dosage there is17considerable increase in photo-toxic reactions and at the same time adecrease in the degree of specificity.The early studies that were performed on animals and humans used acommercial Hp preparation which is a crude mixture of many porphyrins.Schwartz and co-workers in 1955 showed that it was a minor impure fractionin Hp that had the tumour localizing properties (Schwartz et al., 1955).So, in 1960, Lipson under the direction of Schwartz prepared a derivativeof hematoporphyrin by an acetic-sulfuric acid treatment of hematoporphyrin(Lipson et al., 1961). This product was termed hematoporphyrin derivative(Hpd). Subsequent studies indicated that Hpd has superior and morereproducible tumour localization properties than Hp or other simplemonomeric porphryins in a variety of tumours of the bronchus, esophagus andcervix (Kyriazis et al., 1973, Gray et al., 1967). A clinical study of 226cancer patients was conducted in 1968 by Gregorie and co-workers who showedthat Hpd was localized in 84% of adenocarcinomas and 77% of squamous cellcarcinomas (Gregorie et al., 1968). In 1971, Leonard and Beck foundpositive tumour fluorescence in 100% of 40 patients with tumours of oralcavity, pharynx and larynx following Hpd injection (Leonard and Beck,1971). Hpd also exhibits photodynamic properties and was first used in1966 for photodynamic treatment of a metastatic chest wall breast cancer(Dougherty, 1989).2.2.2 Hpd And Its Active ComponentsHpd is a complex mixture of many compounds including hematoporphyrin(Hp), hydroxyethylvinyl-deuteroporphyrin, (Hvd), protoporphyrin and avariety of dimers and oligomeric species bound covalently by ester or ether18linkages (Dougherty 1987, Byrne et al., 1987). Hp, Hvd and protoporphyrinare not preferentially retained in vivo in tumours (Dougherty, 1987). In1984, Dougherty suggested that Hp molecules linked by ether/ester bondsthrough the 2 and/or 4 hydroxyethyl groups, known as dihematoporphyrinether/ester (DHE) is the tumour localizing component of Hpd (Dougherty,1984). A partially purified preparation of Hpd containing 80-85% of thistumour localizing fraction known as Profimer sodium (Photofrin) is now madeavailable through Quadralogics Technologies Inc. (Vancouver, B.C.) forphase III clinical trials of photodynamic therapy in patients with lung,esophagus or bladder cancers.2.2.3 Photo-Physical PropertiesHpd and Photofrin have characteristic absorption spectra in the ultraviolet and visible regions, with its major absorption in the soret band ataround 405 rim. This corresponds to transitions from the ground singletstate to the second excited singlet state. Four minor excitation peakscalled the Q band are observed in the visible region above 500 rim.Spontaneous decay of the singlet state to the lowest vibrational stateproduces the characteristics dual-peak red fluorescence at 630 run and 690rim. It has been established that the absorption, excitation and emissionspectra of Hpd and Photofrin are effected by the type of solvent, pH andother factors. In its aggregated form, the soret peaks of the absorptionspectra are blue shifted and their emission spectra are shifted towards thelonger wavelength (red shifted) (Rotomskis et al., 1989, 1984, Brookfield,1984, Moan and Sommer, 1981). The fluorescence intensity is decreased in19oligomers, by at least a factor of four compared to the dimers (Smith,1985, Moan and Sommer, 1981, Andreoni and Cubeddu, 1984).Besides spontaneous decay to the ground state that results influorescence, the excited Hpd or Photofrin molecules can alternativelyundergo intersystem crossings to the more stable triplet state. Tripletstate molecules can transfer their acquired energy to ground state oxygento produce singlet oxygen and other oxygen radicals resulting in damage ordeath of cells (Chapman et al., 1991, Mitchell et al., 1990, Thomas et al.,1987, Henderson and Miller, 1986, Moan and Sonmier, 1985, Henderson andDougherty, 1984, Weishaupt et al., 1976). Although the photodynainic effecton tumours is beneficial, Hpd or Photofrin also photosensitizes the skin.Patients have to remain out of exposure to any strong light for four weeksor more to avoid skin burns (Dougherty, 1989, Wooten et al., 1988, Razum etal., 1987). Based on theoretical calculations, very low doses of Photofrin(< 0.34 mg/kg) may be devoid of skin photosensitivity (Potter, 1990).However, usefulness of low dose Photofrin or Hpd in the detection of earlylung cancer has not been established.2.3 Tumour Detection Using Hpd or PhotofrinTo facilitate the detection and localization of early lung cancer,several approaches have been developed using fiberoptic bronchoscopy incombination with Hpd or Photofrin as the fluorescent marker (Kato andCortese, 1985, Profio et al., 1984, Cortese et al., 1982, Cortese andKinsey, 1981, Profio et al., 1979, Lipson et al., 1961). Excitation of Hpdor Photofrin is normally performed using lasers or mercury arc light20sources. Violet-blue light of wavelength 400 run - 415 rim is transmitted byfiberoptics to the tissue surfaces via the fiberoptic bronchoscope toexcite the drug which then fluoresces with its characteristic salmon redcolour. The emitted fluorescence is collected by the imaging bundle of thebronchoscope or by a collecting fiber inserted into one of the channels ofthe bronchoscope. The collected fluorescent light is transmitted tosensitive light detectors and displayed as a visual image on a videomonitor, or as an audio or digital signal.2.3.1 Mechanism for DetectionFluorescence detection using Hpd or Photofrin as the fluorescentagent is based on the principle that (i) Hpd/Photofrin in tissues emitsfluorescence with peaks at 630 rim and 690 rim when excited by light at 390run - 415 rim, and (ii) the concentration of Hpd/Photofrin in malignanttissue is higher than the adjacent normal non-malignant tissues from 3hours to at least 72 hours after intravenous injection. Tumours can bedetected by differences in Photofrin concentration seen as a more intensespecific red fluorescence compared to the surrounding normal tissue.Although the red fluorescence from large tumours can be readily visualized,the fluorescence from small early lesions is usually not visible withconventional fiberoptic bronchoscopy because the fluorescence quantum yieldof the tumour-localizing components of Photofrin is very low. Profio et alhad calculated it to be less than 2% (Profio, 1984). Background noise canadversely affect the detectability of such low fluorescent light. Toovercome this problem, sensitive light detectors such as image intensifiersor photomultiplier tubes have been used to amplify the fluorescence.21However, even with highly sensitive light detectors, a very highconcentration of the drug (2-3 mg/kg Lv) is needed before fluorescence canbe detected (Kato and Cortese, 1985, Profio and Balchuin, 1985, Profio,1984, Hayata et al., 1982). At these doses, Hpd and Photofrin result inskin photosensitivity as discussed in chapter 2.2.3 above. (Dougherty,1989, Wooten et al., 1988, Razum et al., 1987). While this may betolerable for treatment of lung cancer, it would be clinically unacceptableto use Hpd or Photofrin as a diagnostic agent.2.3.2 Reviews of Fluorescence Endoscopic Detection SystemsKinsey and Cortese of the Mayo clinic developed the first non-imagingdevice in 1978 (Kinsey et al., 1978, Kinsey and Cortese, 1980). Thisdetection system used a double lumen bronchoscope. The emittedfluorescence from the bronchial surface was transmitted to thephotomultiplier tube via a fiberoptic guide which was placed in one of thelumens of a double channel bronchoscope. A rotating chopper wheel was usedto alternate between the fluorescence and white light reflected modes at 30Hz. Excitation of Photofrin was by 405 rim light from a filtered 200 Watt,mercury arc lamp. The alternate pulses of 405 mi light and the unfilteredwhite light are transmitted to the bronchial surfaces via the illuminationguide of the bronchoscope. The frequency at which the chopper wheelrotated allowed the reflected light image to be visible in real time. Alock-in amplifier generated a frequency-modulated audio signal. The audiopitch produced was proportional to the intensity of the red fluorescence,and thus alerted the bronchoscopist to areas of the tracheobronchial treethat were suspicious for cancer. The system allowed for simultaneous22visual white light examination and the search for Photofrin fluorescence inthe fluorescence mode. The major drawback of this device is that thefluorescence intensity varies depending on the geometry, distance, angleand intensity of the excitation light.The earliest fluorescence imaging system was reported in 1979 byProfio and Doiron at UCSB (Profio et al., 1979). In this system, theemitted fluorescence was imaged with an image intensifier which wasattached to the eyepiece of the bronchoscope. A 690 run band-pass filterwas placed in front of the intensifier to allow transmittance ofHpd/Photofrin red fluorescence while excluding reflected excitation lightand background autofluorescence. The fluorescence intensity at this bandwidth was amplified 30,000 times by the intensifier. Tumour containingHpd/Photofrin was seen as a brighter spot. Colour information was lostwith the intensifier and hence the image was displayed in gray scale.Excitation of Photofrin fluorescence was by 405 run - 413 run light from aKrypton-ion laser. The excitation light was conducted via a 400 micronquartz fibre inserted through the biopsy channel of the bronchoscope. Aflip-flop mirror configuration allowed rapid switching between white lightand the violet light illumination (Profio et al., 1982). The system had avery high sensitivity with 97% of tuniours showing positive fluorescence,but the specificity was only 50% (Balchuni et al., 1990). The requirementof a subjective human judgement of the brightness contrast was by far themajor drawback of the image intensifier system. Although lesions withbright spots were readily detectable, precise delineation of the size andextent of the tumour in areas with diffuse fluorescence, blurred margins orwith low contrast from background was very difficult. Furthermore, the23human eye is more sensitive for colour vision than to discern differencesin brightness contrast. In addition, false positive signals occurred whenthe bronchoscope tip was closer to the bronchial surface or cartilaginousareas (see chapter 1, section 1.3 for a description of the bronchial wall).Again, this system does not correct for the dependence of the fluorescenceintensity on the tissue geometry, the angle and distance of the excitationlight source as well as the pickup light guide from the bronchial surface.The system developed at the Tokyo Medical College in Japan had acombination of imaging and non-imaging detectors. It consisted of a pulseexcimer laser and a Spectro-Image Analyzer (Hirano et al., 1989).Photofrin was excited with the excimer laser at 405 run and the emittedfluorescence was transmitted through the optical fibre of the bronchoscopeand then separated by a half reflecting mirror such that the signal wassimultaneously picked up by an image intensified SIT camera and aspectrograph (polychromator and intensified photodiode array, with anoptical multi-channel analyzer). The fluorescence spectrum was displayedon the video monitor to distinguish the Photofrin fluorescence and tissueautofluorescence. Subtraction of the tissue autofluorescecne spectrum fromthe Photofrin fluorescence can then be obtained. A xenon lamp was used forreflected white light illumination to produce a colour video image. Thislight transmission was turned on and off at 30 Hz by an optical chopperdriven by a motor. Thus white light and fluorescence images as well asPhotofrin spectral characteristics can be observed simultaneously in realtime. Although the system had many advantages in terms of specificity andsimultaneous white-light fluorescence image display, the margins of thetumour cannot be clearly defined on the displayed pseudo-images.24Furthermore, like the other systems, the fluorescence intensity issignificantly influenced by the position of the bronchoscope from thebronchial surfaces.A different approach using background subtraction was made byBauingartner and co-workers in Germany (Baumgartner et al., 1987).Fluorescence from Photofrin and bronchial or bladder tissues were excitedalternately at two wavelengths, one at 405 nm and the other at 470 run. AKrypton-ion laser with a specially designed resonator dichroic mirrorproduced the two wavelengths for illumination. The 405 nm violet lightexcited both Photofrin and tissue autofluorescence, while the 470 run bluelight gave very similar tissue autofluorescence but very little of thePhotofrin fluorescence. Thus, if the fluorescent signals or images fromthese two wavelengths were subtracted, the net result was mainly the redPhotofrin fluorescence. This was accomplished in near real time by imageprocessing techniques. Although this was a sensitive system and somecontrast enhancement was provided with the subtraction, the displayedsubtracted images were still dependent on the distance of the bronchoscopefrom the tissue surface.Based on a similar principle of the German group, but using a muchcheaper Helium-neon laser a fluorescence photometer was developed by Potterand Mang (Potter and Mang, 1984). Fluorescence was excited at 612 run and632.8 run using two 5 mW Helium-neon laser. This system was designed forthe detection of non-palpable metastatic breast cancers in the skin and themicro-metastases to regional lymph nodes (Mang et al., 1991). Delineationof surgical margins during thoracotomy and detection of intraoperativemetastases during mediastinoscopy were potential applications. However,25the drawback was again the angular and distance dependence of theexcitation light and the receiving light guide from the tissue surface.To address and solve some of these problems of distance and angledependence, a ratio fluorometer probe was devised by Profio and co-workers(Profio et al., 1984). The concept of ratioing is important because it cancorrect or cancel the variations of the excitation light power, angle,distance and surface geometry. In the original report by Profio and coworkers, the fluorescence signal at 690 run was ratioed against the signalof the reflected violet (405 rim - 410 run) excitation power. Althoughratioing canceled the dependence on distance, angle and excitationintensity, they found that the ratio was influenced by variations in theviolet reflectivity as well as the fluorescence yield of the drug (Profioet al., 1984). An improvement was later accomplished by taking the ratioof the Photofrin fluorescence intensity at 690 rim to a greenautofluorescence at 560 rim. This approach assumed that normal andmalignant tissues emit similar autofluorescence at the green region whenexcited by 405 rim and hence the spectra were normalised at 560 rim (Profioet al., 1986). Because more Photofrin was retained by tumour tissuecompared to the adjacent normal bronchial tissues, the tumour can bedetected by an elevated red-green ratio. Since the red-green ratio is adimensionless function, it cancels the dependence of angle and distancebetween the bronchoscope tip and the bronchial surface. The excitation offluorescence was procured by illumination with a Krypton-ion laser at 405run. A two channel bronchoscope was used, one channel to conduct theillumination quartz fibre while the other channel was used for thecollecting fibre. The distal end of the illumination fibre was fitted with26a microlens to spread out the light. The emitted fluorescence was pickedup by the collecting fibres and was split by a dichroic mirror. The twobeams were then passed through a red and green band-pass filters at 690 runand 560 run respectively. The normalized ratio of red to green fluorescencewas then measured and displayed as a digital read-out as well as an audiosignal. Using this device a specificity of 100% was reported, but thesensitivity was only 70% (Balchuxn et al., 1987). The low sensitivity couldbe due to the high threshold ratio that was chosen as indicative of tumour.In that study, a red-green ratio of 4 to 5 times higher than an apparentlynormal area was considered to be abnormal. The disadvantage of the ratiofluorometer probe is that it is a non-imaging device. The exact site andsize of the tumour cannot be delineated accurately. Furthermore as a onepixel system, small cancers can be readily missed because of fieldaveraging effect.The group at the Swiss Federal Institute of Technology in Lausanne(EPFL) also employed the ratioing concept to develop a device to imageearly lung cancers (Wagnieres et al., 1990). A dichroic mirror was used tosplit the fluorescence light into two regions which were then passedthrough two band-pass filters, one in the 520 rim to 580 rim range and theother between 600 run and 720 rim. The red and green light were thencaptured by separate image intensified CCDs. The two images (red andgreen) were ratioed. Autofluorescence subtraction was also performed forcontrast enhancement. The processed pseudo-image was then displayed on avideo monitor. This device allows time sharing between reflected whitelight and fluorescence modes by using an optical chopper. This system isstill not yet capable of real time pixel by pixel ratio imaging. There is27also significant light loss and optical distortion resulting in poorquality of the pseudo-image.Another approach was made by the group in Lund, Sweden (Andersson etal., l987a, Montán et al., 1985). They developed a two-dimensional multi-spectral imaging system with background autofluorescence subtraction. ACassegranian telescope was used to collect the fluorescence light in amulti-mirror configuration. A spherical mirror was divided into identicalquadrants and interference filters were arranged in front of each of thequadrants so that four fluorescence wavelength bands were imaged onto anintensified CCD matrix detector. The four sectors had adjustable screwsthat allow for positioning the individuals images onto the detectorsurface. Subtraction or ratioing of the four images can be carried out byan image processing board and a pseudo-image (false colour coding) wasobtained that can delineate the tumour site with a relatively highcontrast. The image acquisition and processing takes about 5 seconds andthis may be a problem for lung examination because of respiratory motionartifacts. Another problem is related to the excitation light source usinga nitrogen laser (337 run). The safety of illuminating the bronchus withultra-violet light has not been established.2.4 Previous work using of Tissue Autofluorescence In DiagnosisIt has been reported that even without introducing any foreignfluorescent agents, tissues exhibit characteristics intrinsicautofluoresence. Differences in autofluorescence between normal andcertain abnormal tissues have been demonstrated in excised human tissues28and in animal models of tumour (Alfano et al., 1984, Tang et al., l989a,Tang et al., l989b, Caplan, 1967, Herley, 1944, Figge et al., 1944, Sutroand Burman, 1933, Wood, 1908 (in Anderson and Parrish, 1982)). However,differences in tissue autofluorescence is generally considered too weak tobe quantified or are insensitive to current detection devices especiallyfor small lesions (Lakowicz, 1983). This belief led many investigators toassume that an exogenous fluorescent dye is essential for the detection ofcancers. Furthermore, because the autofluorescence intensity betweennormal and cancerous tissues were thought to be similar in the green regionof the visible spectrum, background fluorescence was used for ratioing orsubtraction to enhance the contrast between tumour and normal tissues whena fluorescent dye such as Hpd/Photofrin is used. The ideas are soingrained that fluorescence detection has been generally equated with theuse of fluorescent drugs (Mang et al., 1991, Wagnieres et al., 1990,Baumgartner et al., 1987, Andersson et al., 1987a, Profio et al., 1986,Montán et al., 1985, Kato et al., 1984).In a limited fashion, autofluorescence has been used for thedetection of dental decay (Alfano and Yao, 1981), cancer in rats (Anderssonet al., 1987b, Alfano et al., 1984), as a marker for photoaging (Leffell etal., 1988), to characterize brain tumours (Montán and Strombland, 1987),and to discriminate atherosclerotic plague from normal arterial wall(Andersson-Engels et al., 1991, Lucas et al., 1990, Richards-Kortum et al.,1989a, Gmitro et al., 1988, Hoyt et al., 1988, Clarke et al., 1988, Sartoriet al., 1987). Very recently, laser-induced autofluorescence is been usedin the diagnosis of human colon adenomas and carcinomas (Schomacker et al.,1992, Rava et al., 1991, Richards-Kortum et al., 1991, Kapadia et al.,291990, Cothren et al., 1990), as well as oral tuniours (Yang et al., 1987).Oral tuniours are unique in that they contain endogenous porphyrin and hencefluoresces red without any exogenous fluorescent drug (Yang et al., 1987,Harris and Werkhaven, 1987).There has been very limited investigation on the diagnosis of lungcancer using tissue autofluorescence. The conclusion from a study of 17patients with bronchial tumours investigated by the group in Lund, Swedensuggested that the use of autofluorescence alone to detect cancer of thebronchus is probably not possible (Andersson-Engels, 1989).There is much speculation as to why cells or tissues autofluoresce(Benson et al., 1979, Herly, 1944). Naturally occurring fluorophors in thecells such as pyrimidines and flavin nucleotides co-enzymes, carotene andporphyrin are responsible for the cell fluorescence at the visiblewavelengths (Benson et al., 1979, Fellner, 1976, Blankenhorn, 1958). Themost intensely fluorescing component in the cell was found to be in themitochondria. This is attributed to the proteins bound to respiratory coenzymes in mitochondria (Aubin, 1979, Chance and Schoener, 1966). Flavinis known to fluoresce and exhibit spectral changes when it is transformedfrom the oxidized to reduced state. It has been demonstrated that thespectra maxima of flavins vary from 520 mu to 535 mu depending on the redoxstate (Aubin, 1979, Benson, 1979, Chisla et al., 1975). Riboflavins areknown to fluoresce in the visible spectrum (about 510 run to 530 run) and area part of the co-enzyme flavin adenine dinucleotide (FAD) which isresponsible for the oxidation-reduction in the mitochondria (Chance andSchoener, 1966). Nicotinamide adenine dinucleotide (NADH) has also beenreported to be a major tissue fluorophore involved in the redox state of30the mitochondria. NADH is known to fluoresce strongly at 470 rim whenexcited at 340 rim (Chance et al., 1962, Welsh et al., 1977). Lipofuscin,collagen and elastin in animal cells have very strong autofluorescence inthe green-yellow spectral regions (references in Aubin, 1979). Naturallyoccurring porphyrins are found in urine, stool, and erythrocytes. Theprincipal fluorescence spectrum of porphyrin is found to be between 597 rimto 634 rim (Harris and Werkhaven, 1987). Fluorescence emission maxima forkeratin and melanin is observed in the visible region at 525 rim to 540 rim(references in Alfano et al., 1984). The ultra-violet fluorescence ofcells is from tryptophan-containing proteins in the respiratory enzymes ofthe mitochondria and from nucleic acids (Udenfriend, 1962). Humaneosinophils emit autofluorescence in the green-red regions of the spectrumand this characteristic was used in the isolation and analysis of thesecells using flow microfluorometry (Weil and Chused, 1981). Macrophagesoften contain ingested fluorescing particles which can produce“autofluorescence” when excited at the appropriate wavelength. Inconnective tissue, unidentified fluorophors in collagen and elastin havestrong autofluorescence in the blue-green spectral regions (Neuman andLogan, 1950, Banga and Bihari-Varga, 1974).31CHAPTER 3: LIGHT INTERACTION WITH TISSUES3.1 Photo-BioloRical ResponseThe spectrum of electromagnetic radiation ranges from radio-waves togamma rays. Light is the part of the electromagnetic radiation spectrum wesee, and ranges in wavelength from 400 run to 700 run apparently changingcolours from violet to red.The interaction of electromagnetic radiation with matter involves theexchange of energy. A molecule is said to be excited when one or more ofits electrons has been displaced into a higher orbital. These moleculesare often unstable and may decay by radiative and/or non-radiativeprocesses. The energy lost during these processes can be dissipated asheat, luminescence, or as a chemical reaction. More specifically, theimmediate photo-biological response of light interaction with biologicaltissues can be divided into: (i) photo-thermal response; (ii) photo-physical response; and (iii) photo-chemical response. Photo-thermalresponse is the result of the transformation of absorbed light energy toheat, leading to coagulation or destruction of the target tissue. Photo-chemical interactions result when light is absorbed by either endogenous orexogenous chromophores introduced to initiate chemical reactions such asproduction of reactive singlet oxygen species in photodynamic therapy. Anexample of a photo-physical light-tissue interaction is fluorescence.Applications of photo-thermal response and photo-chemical response aremainly therapeutic, while photo-physical effects are used both fordiagnostic and therapeutic purposes. Therapeutic applications of light isbased on the effect of light on tissue, whereas diagnostic applications32exploit detectable changes resulting from the effect of tissue on light(Parrish and Wilson, 1991). Comprehensive reviews of these effects oftissue interactions with non-ionizing radiation and establishedapplications of these effects in medicine and biology are described inWilson and Jacques, 1990, Welch et al., 1989, Boulnois, 1989, and Andersonand Parish, 1982.Tissues can be defined by their optical properties (discussed below),thermal properties (heat capacity and heat diffusivity), mechanicalproperties (viscosity, elasticity and tensile strength), chemicalcomposition, architecture or structure (physical arrangements of organellesand cells), and physiological state. The effects of light-tissueinteractions are determined by these properties, as well as the conditionsof the irradiation and the time after irradiation at which the evaluationis undertaken (Thomsen, 1991, Boulnois, 1989).3.2 Optical Properties of Turbid Media: TissueTissue is a scattering and absorptive medium with a higher refractiveindex than air. As an electromagnetic wave enters a medium with refractiveindex n, the wave amplitude decreases (absorption) and its phase velocitybecomes c/n (dispersion). The refractive index is not only a property ofthe medium however but also depends on the frequency of the electromagneticwave. The refractive index for light travelling in soft tissue isdetermined to be in the range 1.38 - 1.41 and for adipose tissue to be 1.45(Bolin et al., 1989 and references therein). Light can reflect at thetissue surface, scatter due to the spatial variations in the reflective33index or dielectric properties within the tissue, and be absorbed bynatural chromophores in the tissue. The amount of light that is scatteredand absorbed in the tissues is also wavelength dependent. Wavelengths inthe ultra-violet and infra-red regions are highly absorbed in tissues. Thepenetration depth, S (i.e. the depth at which the light intensity isattenuated by a factor of l/e due to absorption and scattering), is lessthan 20 urn at these wavelengths. Absorption in the ultra-violet region isdue to tissue proteins, while water absorption dominates at wavelengthgreater than 1300 run. In the wavelength region between 450 rim to 600 rim,where the absorption is due primarily to heme and melanin the absorptionand scattering coefficients are of the same proportion and S is about 0.5mm to 2.5 mm. To a lesser extent, bilirubin, flavins, carotenoids, thecytochrome pigments of the respiratory chains in the niitochondria, andwater also absorb light in this region. Between 600 rim to 1300 rim,scattering is the dominant form of interaction. This region is the mosttransparent for tissue and light is able to penetrate relatively deeply,and S being about 2.0 mm to 15.0 mui (Svaasand, 1990, Wilson et al., 1990,Parrish, 1981). Visible light (400 run - 700 rim) is absorbed very slightlyby unpigmented cells. This is the reason why cells appear transparent ortranslucent (from scattering of light by particles in the cells) underbright field microscopy. Most of what is absorbed by the cells is due tothe presence of cytochrome and flavoproteins and other macromolecules andproteins as described in chapter 2.4.Since this thesis deals mainly with bronchial tissues, a schematicrepresentation of the optical processes that may occur at the differentbronchial tissue layers during the propagation of light is illustrated in34figure 1. The structure and histology of the bronchial wall is given inChapter 1, section 1.3. The respiratory tract is a dynamic organ, and assuch, the optical properties must be dealt with as a dynamic process.Because of its complex structure, precise modeling of the path of opticalradiation within the tissue is very difficult. The many different physicalstructures and chromophores within the different layers of the bronchialwall influence its optical properties. However, within any given layer,the following basic optical processes may occur. Usually about 4% of anormally incident light beam is reflected from the air-tissue interface andboundaries of the layers due to the mismatch in the refracted index(Fresnel reflections or specular surface reflections). The rest of thelight then penetrates into the tissue encountering multiple scattering(back and forward scattering) and absorption processes which may lead tophoto-chemical reaction or loss of energy as heat, fluorescence orphosphorescence. For very thin or transparent tissue, a portion of theincident light will be transmitted through the tissue.The angle 8 at which light travels in tissue is given by Snell’s Law:sin 8 =1/2(sin 8)Light travelling from tissue to air (i.e. from a higher refractiveindex material to a material with lower refractive index) will beinternally reflected if 0 exceeds the critical angle:sin = sin 9O°(1/2), where < 17235Figure 1 Schematic representation of the propagation of lightbronchial tissue.SUBMUCOSAL LAYERinincident lightFluorescencescatteringandabsorptionspecular reflection (—4Z - —7%)MUCOSAL LAYERCARTILAGEtransmitted36Therefore, not all the back scattered light in tissues is transmittedthrough the tissue-air interface. Diffuse light scattered at an anglegreater than 48° with respect to the interface normal is totally internallyreflected back into the tissue.The intrinsic optical parameters of the tissue, coefficient ofabsorption (Pa) scattering (ps), and anistropy (g) determine thedistribution of light within the tissue. The phase and polarizationparameters are ignored in this treatment of photon propagations since theseparameters are very quickly randomized in tissue. The absorption andscattering coefficients are the probabilities per unit path length withinthe tissue that a photon will be absorbed or scattered from a singlecentre, respectively.A review of some of the very limited published work on thepropagation of light in different biological tissues (Wilson and Jacques,1990) suggests that: (i) the movement of light in tissue is influenced bythe intrinsic optical parameters which may vary between different tissuetypes, or within a single tissue, (ii) these optical parameters arestrongly dependent on the wavelength in the visible range, and (iii)interfaces, boundaries conditions, and geometry of incident light willaffect the distribution of light.Absorption coefficientThe absorption of light in tissues occurs when the resonant frequencyof the molecules matches the wavelength of the incident light. Where thereis a transfer of energy from a photon to a chromophore, the photon ceasesto exist. Usually the absorbed energy is dissipated as heat, but the37energy from a small fraction of absorbed radiation may lead to photo-chemical reactions or luminescence. The average absorption of a localregion of tissue is optically characterized by the absorption coefficient,The mean free path between absorption events is given by“a•Whenscattering is also present, the effective path length of radiation withinthe tissue or medium is increased, generally increasing the likelihood ofabsorption.Scattering coefficientIn tissue, scattering is an elastic interaction between opticalradiation and the tissue in which only the direction of photon propagationis changed. Scattering can also be inelastic which results in a change ofwavelength (this is however not common in tissues). The scattering oflight is caused by heterogeneities in the refractive index of tissues.Physical heterogeneities within a cell reflect the presence of cellorganelles such as cell nuclei, mitochondria, the extracellular matrix andcollagen fibrils, all of which influence light scattering (Jacques, 1990).The effect of scattering within a tissue is dependent on both the degree ofscattering and the spatial distribution of the scattering light. There aretwo types of scattering: Rayleigh scattering and Mie scattering. Rayleighscattering predominates where the size of the molecules or particles isless than 1/10 of the wavelength. The scattering in this case is veryweak, isotropic, polarized, and varies with the fourth power of thewavelength. Mie scattering occurs where the particle size greatly exceedsthe wavelength. It is highly forward directed and is independent ofwavelength. The average scattering within a tissue is optically38characterized by its scattering coefficient p. The mean free path betweenscattering events is l/p.AnisotropyWhen a scattering event occurs, the trajectory of the photon will bedeflected by an angle 6. In general, light scattering in tissues isforward directed (anisotropic), i.e., the photon deflection angle is smalland so the photon will continue in the same general path with very smalldeviations. Experimental evidence indicating that light scattering intissues is forward directed in tissues was reported for human stratumcorneum and epidermis at ultra-violet and visible wavelengths, for humandermis at 633 nui, for blood, and for liver at wavelength between 350 run to2200 nm (Jacques et al., 1987, Flock et al., 1987). In anisotropicscattering, many scattering events are required before the photon loses itssense of history and achieves a random walk. Anisotropic scattering occurswhen the scattering particle is about the same size or greater than thewavelength of the incident light. This type of scattering is approximatelymodeled by Mie theory which deals with the scattering by spheres of sizessimilar to the incident wavelength. The angular distribution of scatteringis defined by a single scattering phase function, p(6), which describes theprobability that the photon will be deflected by a given angle 6. Thephase function can be represented by a series expansion as a sum ofLegendre polynomials:p(O) = ZWp(cosO)n039The Henyey-Greenstein function,p(6) =1,’4ir(l-g2)/ l+-2gcos8)3”which describes light scattering in galaxies (Henyey and Greenstein, 1941)is used to approximate light scattering in tissues (Jacques et al., 1987).Anisotropy g, describes the average angular dependence of scattering and gis defined as the expectation value or the average cosine of the scatteringangle <cosO>:<cosO> = g= Sp(O)cos(O)2lrsinodo.Boundary conditionsWhile the spatial distribution of light within tissue is determinedby the absorption and scattering cross sections, the mismatch of therefractive index in tissues and the interface of the air/water/glass/orother materials characterizes the surface boundary conditions. Two opticalsurface effects involving the boundary are: (i) specular surfacereflectance as light enters the tissue, and (ii) total internal reflectanceas light attempts to escape the tissue.The specular surface reflectance, at a mismatched boundary isgiven by:= [(n2-1)/(n+ 1240where n1 is the refractive index of tissue and n2 is the refractive indexof the adjacent medium. Specular reflection does not provide anyinformation about tissue absorption and scattering properties but tissuesurface roughness and refractive index values can be obtained.Total internal reflection occurs at a mismatched boundary, n2 < n1,where photons attempt to escape the tissue at angles that are greater thanthe critical angle. The total amount of internally reflected light, R. iscalculated using Fresnel’s law of reflectance for randomly polarized light(Hecht, 1987).The back-scattered light striking the interface is assumed to bedistributed diffusely. For tissue/air interface, only about 50% of theback-scattered light is transmitted, i.e. about 50% is internallyreflected. A discussion on the effect of surface roughness on internalreflection and the influence of tissue refractive index in internalreflection is found in Jacques, 1990, Wilson and Jacques, 1990.3.3 Light Propagation in Tissue: TheoryThe exact, detailed mathematical modelling or calculation of lightpropagation in tissue is very complex. Tissue is highly scattering andinhomogeneous with many local microscopic spatial variations in therefractive index. In addition, if the angular distribution of thescattered light is not isotropic, the intensity can be higher in certaindirections than in others. A mathematical computation of the propagationof light in tissue can be made by analytical theory or transport theory.Analytical theory is based on the fundamental equations governing field41quantities. This approach is often based on Twersky’s work (Twersky, 1970,Twersky, 1962). However, an exact solution of Maxwell’s electromagneticwave equations is almost an impossible task. Detailed informationregarding the scattering and absorption processes associated withmicroscopic spatial variations in the dielectric properties of tissue mustbe considered in the calculation. In principle, starting with Maxwell’sequations is the most fundamental approach, since the statistical nature ofthe tissue as well as all diffraction and interference effects are takeninto account in the calculations. However, this approach is mathematicallycomplex and, in practice, approximations must be made. Its usefulness istherefore limited.Transport theory deals directly with the transport of photon energywhich may be locally absorbed or elastically scattered through turbidmedia. The wave properties of light (diffraction and interference effects)are generally neglected. Transport theory has been shown experimentally tobe applicable in many situations. The basic equation in transport theoryis the radiative transfer equation. The transport equation and severalapproximate numerical and analytical models that have been developed todescribe the propagation of light in tissue will be discussed brieflybelow.Radiative transfer equationThe distribution of light in biological medium can be adequatelydescribed by the radiative transfer equation (Chandrasekhar, 1960). Thisequation is equivalent to the Boltzman equation used in neutron transporttheory and the kinetic theory of gases. Whereas neutron transport is42concerned with particle fluence rate (particle density multiplied by thespeed), photon transport is expressed in the energy fluence rate (spaceirradiance). The application of Boltzman equation to photon transport ofx-ray and gamnia ray is discussed in texts on radiation shielding (Chiltonet al., 1984, Profio, 1979, Goldstein, 1959). The transport of opticalphotons may be treated in the same fashion as x-ray photons with thedifferences in the interaction cross sections. The Boltzman equationexpresses the balance between gains and losses. The generation of Ramanscattering is considered negligible, so there is no change in thewavelength on scattering. The scattering and absorbing centres are assumedto be uniformly distributed in the medium. The equation of transfer is:=- Pb(r,f2)+j’it p(O)Ø(r,c2’)df2’+S(r,12)where (r,f2), in W/m2/sr, is the angular energy flux density at position rand angle 2 (power crossing unit area normal to the direction vector ).In optics, this quantity is known as radiance. The energy fluence rate,also called the space irradiance is equal to the radiance integrated over4ir solid angle. The normalized angular scattering function, P(9) is thescattering phase function discussed previously, with f2.f’= cosO, anddescribes the probability of scattering from angle 2’ into angle . Theintegration is over all 4ir steradians of solid angle in a spherical coordinate system, azimuthal symmetry is assumed. The total interactioncoefficient is equal to + 1a S(rj) is the volume source intensityin W/m3/sr.43The left hand term of the above equation represents the change in theflux density due to photon transport. The first term on the right handside represents the losses in (rjl) due to attenuation by scattering andabsorption, the second term is its gain due to the scattering of photonsback into the original direction of motion. The third term on the right isthe contribution from other sources.The Boltzman equation is solved subject to specified boundaryconditions. The angular flux density is zero for any rays entering thetissue, and a specular reflection condition is imposed at the centre of asphere or axis cylinder (Profio, 1989, Welch et al., 1987, Profio andDoiron, 1986). Except for very simplified cases, the Boltzman equationcannot be solved analytically. Two- and three-dimensional solutionsrequire a serious restriction of the angular distribution of the scatteredlight (Case and Zweifel, 1969). Computer methods have been developed tosolve this equation numerically, however, numerical solutions are stillcostly, especially for two-and three-dimensional problems (Welch et al.,1987).Monte Carlo methodAn alternative to solving the equation of transfer is to simulatelight transport in a scattering medium with the Monte Carlo method (Keijzeret al., 1989a, Prahl et al., 1989, Wilson and Adam, 1987, Flock et al.,1989). This is basically a computer simulation of the tracks of individualphotons. Each photon out of a large sample is followed in its random walkuntil it is absorbed. The total light distribution is estimated from theresulting distribution of absorbed photons. The Monte Carlo method has44been applied to transport of optical photons in three-dimensional geometryproblems since this method can calculate light transport withoutapproximating the tissue geometry or angular distribution of light. Theadvantages of Monte Carlo simulations are their exactness and flexibility(the exact scattering properties can be simulated after sufficient photontrajectories are computed as well as treatment of internal reflection atmismatched boundaries, non-uniform irradiance, and non-perpendicularirradiance). However, this method is subject to statistical errors and isalso time consuming. At the present time reasonable answers can only beobtained after many hours of computing time.Approximate mathematical modelsBecause the radiative transport equation cannot be solved exactlyexcept for a few special cases, approximate solutions to the radiativetransfer equation have been developed to study the propagation of photonsin various media. The inhomogeneous structure of tissues allows one toneglect spatial coherence and interference mechanisms of light, hencepermitting substantial simplifications to be made.In conditions where absorption dominates, Beer’s law is a simplemodel that yields reasonable results. Beer’s law describes the exponentialattenuation of light as it passes through tissues. However, mostbiological tissues are highly scattering, and hence Beer’s law is notapplicable at many wavelengths. A model of two diffuse light fluxes normalto the tissue surface travelling in the forward and backward directions wasproposed (Kubelka, 1954, Kubelka, 1948) to describe light propagating intissues. In this model, a forward flux travelling through a very small45path length ôz, is attenuated due to absorption and scattering and isincreased by the scattering contribution of the backward flux. A similarrelation holds for the backward flux, with the exception that it travels inan opposite direction of dz. Thus the distribution of light as a functionof a single variable, namely the depth in the medium can be easilycalculated. A newly suggested seven-flux model has been proposed toprovide for forward and backward diffuse fluxes propagating in the x, y,and z directions as well as a collimated light flux in the direction ofbeam incidence (Welch et al., 1987). The Kubelka-Munk theory has beenapplied to the quantitative characterization of the normal vessel wall andatherosclerotic plaque (Vogel et al., 1991), blood (van Gemert and Henning,1981), and skin (Wan et al., 1981). The Kubelka-Munk theory is restrictedto uniform beam incidence and to cases in which radial intensity variationis negligible. For uniform diffuse incident light, this model provides asimple relationship between measured reflectance and transmission. TheKulbeka-Munk’s absorption and scattering coefficients of thin tissuesamples are:Skm =(3/4)(l_g)p_(’Iji;-ln[T(t)] as)twhere T0 is the incident collimated beam, t is the sample thickness, A1and Skm are the Kulbeka-Munk’s absorption and scattering coefficientsrespectively. Hence, using these relations, the optical parameters can beeasily evaluated.46Diffusion theory is a simple approximation to the general radiativetransport equation (Ishimaru, 1989). It is derived from the transportequation with the assumption that light is scattered almost uniformly afterencountering numerous scattering events. This means that the angularenergy flux density (r,fl) is only weakly dependent on direction 2. Thediffusion approximation is valid for optically dense and highly scatteringmedia, and it describes the intensity of light as a sum of three termsnamely; an isotropic term, a diffusely forward directed term, and a forwarddirected S function. The latter accounts for the very forward-directedlight scattering that is common in tissues. Applying the diffusionequation to the radiative transfer equation provides a second-order partialdifferential equation that can be solved for appropriate boundaryconditions. Since the model requires diffuse light within the tissue, thetheory tends to break down near boundaries and localized internal sources(Welch et al., 1987). In general, reasonable accuracies have been obtainedfor Nd:YAG laser light which is highly scattered in most tissues such asbrain, bladder and the dermis of skin (Welch et al., 1987, Cheong et al.,1990 and references therein). A diffusion approximation algorithm to solvethe 1-dimensional transport equation deriving the optical properties of ratliver between 350 run and 2200 rim has been reported (Parsa et al., 1989,Jacques and Prahi, 1987).3.4 Measurements of Tissue Optical PropertiesMonte Carlo and Boltzman transport theory calculations and some ofthe approximate mathematical models mentioned above require the knowledge47of intrinsic optical parameters: absorption coefficient, scatteringcoefficient, and scattering phase function. It is not possible tocalculate all these coefficients from the tissue composition and structure;they must be measured. The total attenuation coefficient which is the suniof the scattering and absorption coefficients, can be measured by thetransmission of a narrow, collimated beam through a thin sample. A tissuesample is usually about 100 urn, one mean free path. To minimize scatteringthe sample is placed between two optical flats separated by calibratedspacers slightly thinner than the sample. A collimated beam is transmittedthrough the sample and then to a distant detector with a narrow collectionsolid angle. The is usually >>p, hence the measurement is usually areliable direct measurement of p.The scattering phase function can be directly measured using agoniometric apparatus. The laser beam is scattered by the thin tissuesample held by two optical flats, and a collimated detector is rotatedaround the sample to measure the scattered light as a function of angle 0(Jacques et al., 1987). However, it is difficult to measure the absorptioncoefficient independent of the scattering coefficient, especially when theabsorption is small compared to the scattering. When it is possible towork with large thick tissue samples, integrating sphere measurements canbe performed to measure the reflectance and transmission of diffuse lightas a function of wavelength to evaluatea’ p5(l-g).Details of these and other techniques which are currently availableto evaluate the optical properties are discussed by Jacques, 1990, Cheonget al. , 1990, Wilson and Jacques 1990, Wilson et al., 1990, Svaasand, 1990,Marchesini et al., 1989).483.5 Photo-Physical Process Following Light Absorption*Only radiation that is absorbed will result in a photo-chemicalchange in a molecule or can be effective in producing luminescence(including fluorescence and phosphorescence). According to the GrotthusDraper Law, molecules must have a chromophore in order to absorb light.The fact that each molecule that emits luminescence or is chemicallyaltered by radiation absorbs one quantum of radiation, is known as theEnistein-Stark quantum law. As long as the dose (i.e the product of theintensity and exposure time) is the same, the photo-chemical effect will bethe same. The total number of absorbed photons determines the extent of aphoto-physical process. The energy of the photon and the rate which thesephotons are absorbed have no influence on the photo-physical process. Agiven exposure dose at a given wavelength produces a given degree ofresponse independent of irradiance (Reciprocity law). The quantum yield offluorescence is independent of exciting radiation energy: The intensity offluorescence depends on the number of exciting photons. The change inwavelength (loss of energy) of luminescence emission spectra relative tothose of the exciting radiation energy, is known as the Stoke’s shift. Theluminescence spectrum is invariant with changes in the excitation radiationenergy.When light interacts with matter, it may be either scattered(diffracted light) or absorbed. The kinetic scheme of the behaviour ofmolecules following optical excitation is illustrated by the Jablonski’senergy diagram shown in figure 2. The ground, first singlet, second*Refereences used in preparation of section 3.5 were Lakowicz, 1983, Pesceet al., 1971, Udenfriend, 1969, Udenfriend, 1962.49singlet, and third singlet electronic states are depicted by S0, S1, S2,and S3 respectively. Associated with each of these electronic states are anumber of vibrational energy levels, depicted by V0, V, V2, V3, etc.Between each of these vibrational quantum numbers are a number of quantizedrotational energies. The transitions within the vibrational and rotationalenergy levels to the next lower electronic state are vertical, this governsthe position and profile of the absorption band. This is what is known asthe Franck-Condon principle which states that all electronic transitionsare vertical, occurring without any change in the position of the nuclei.In a singlet excited state, the electrons in the higher-energyorbital have the opposite spin orientation as the second electron in thelower orbital. These two electrons are said to be paired. In a tripletstate these electrons are unpaired, that is their spins have the sameorientation.An incoming photon excites the molecule from the ground state to anyvibrational level of an excited state (Sn). For clarity, figure 2, showsonly one photon absorption for each of the electronic states. Irrespectiveof the vibrational level reached by the initial absorption, the moleculeloses excess vibrational energy within lO13s to occupy the V0 level. Thisradiationless transition is called “vibrational relaxation”; it isrepresented by the vertical wavy lines pointing downward in figure 2. Thisprocess is then followed by a rapid, radiationless internal conversion tothe highest vibrational level of the next lowest energy singlet state,illustrated by the horizontal wavy lines in figure 2. This conversionprocess involves no change of energy. The overall radiationless decayprocess is very fast, within about lOs, the molecule is at the lowest50excited singlet state (S1). Because of this rapid relaxation, thepopulation yield of the lowest energy singlet state following excitation toany level is essentially unity and therefore the emission spectraldistribution and the fluorescence lifetime are usually independent of theexact excitation wavelength. Thermal equilibrium is established within theS vibrational levels. From this level the molecule can subsequently decaythrough radiative fluorescence emission, intersystem crossing or internalconversion to the lower orbital ground state.3.5 .1 FluorescenceFluorescence is one of the several possible pathways of the deexcitation process by which an excited state electron at the S1, V0 levelgives up the energy and return to the ground state as shown in figure 2.Fluorescence lifetime is typically in the order of 10-8 to l0 seconds,representing the average period of time a fluorophore remains in theexcited states prior to returning to the ground state. Again, the FranckCondon principle is invoked, and this time the vibrational levels of theground state govern the profile of the emission band. The wavelength ofthe emission fluorescence spectra is longer than the wavelength of theexcitation light (referred to as Stoke’s shift). The reason is that partof the energy difference between the ground state, V0 level and the upperV level of the S1 state are accounted for by the vibrational relaxationbefore and after fluorescence emission. Generally, the fluorescenceemission spectrum appears to be a mirror image of the absorption spectrum.This is due to the same transitions being involved in both absorption andemission, and the similarities among the vibrational energy levels of S51and S1. Fluorescence does not require an electron to change its spinorientation. Several factors such as solvent effects, solvent relaxation,quenching, and a variety of excited state reactions can influence thefluorescence emission. These factors are discussed in details in the textPrinciple of Fluorescence Spectroscopy (Lakowicz, 1983).Radiationless decay to the ground state and the intersystem crossingto the first triplet state are strongly competing radiationless processes.Fluorescence is delayed (about 10-6 seconds) if the excited electron movesto a forbidden triplet state before moving back to the lowest singlet stateand emitting light. If a electron in the T1, V0 state absorbs thermalenergy to reach the level of the S1, V0 state and one of the unpairedelectrons then reverses its spin orientation, this can result in anintersystem crossing in the reverse direction to the singlet state followedby fluorescence. This indirect fluorescence is delayed but has the sameemission profile as that of the fluorescence from the direct pathway.A transition from a singlet to a triplet state (three energy levelsdue to an unpaired electron in spin) is described as intersystem crossing.The triplet state is always at a lower energy level than the correspondingsinglet state. Radiative transitions between singlet and triplet state are“forbidden” by laws in quantum mechanics, thus the absorption of a photonto give a S0 - -> T1 transition is not allowed. Emission from the firsttriplet state (T1) is called phosphorescence and is generally shifted tolonger wavelength relative to fluorescence. Transition from a tripletstate to the ground state requires one of the paired electrons to changeits spin orientation. Phosphorescence has much longer lifetime of theexcited state, typically ranging from milliseconds to seconds.‘—————-—C)——————CD00.lIIII0<——IIIIIIIIC)04%,ICDCDNi-0 ft00p.‘fth-0CDq•1•.•x—:IIIICD00Ci)0CDI-•oIIHI1IIoftCD.CDCDCl)ç.HCDrf\JLft.CD•_••CDCD.•—•CDCDCDCDftCl)ffI\f\lSJ\J\f\f”.J\J\fftftci)HI-IICD ci)rtHCl•Jft,tj•(I)(0CDCD)fto1 CDCDft<Ci)l)CDCi)fti—iCDCD<I—i•0ft CDCDHci) ft)(N H CDEnergyLevelsdCDH)C-ol)0Ii)i-•C)j_i-I—ift000 H)I-’.Ci)(N(N0- oD(I)ci) 0Cl)HHdCD0-—.ftHH(NH0CDCl)H00-CDH)Cl)ftCDCD Ci)Hft(NC))HftriftCl)CDCi)0<00H)ftft0CDH0CDCDd0 Cl00HCDCDH0ftC))l- C)C))Ci)I-i0C))oC))CtC))•-‘c.‘•d0i-0-i—0I—.C)ftH)CDCD0Ci)CDCi)HCl)U,533.5.2 Characteristics of FluorescenceFluorescence is a result of the absorption of light, and it can occuronly following absorption. It involves the emission of light and anoutside source of energy is required. If light intensity 10 is directedthrough a medium some absorption of the incident radiation will occur. Thelight that is not absorbed is transmitted and can be denoted I. Thus (I-represent the radiation energy absorbed by the medium. As a resultof absorption fluorescence may occur (as discussed above). The intensityof fluorescence is related to the absorption by:F = Q(I-where F is the fluorescence intensity; 1 is the intensity beforeabsorption; I is the intensity after absorption; Q is the quantumefficiency.From Lambert’s law, the fluorescence intensity can be rewritten as:F QI(l - e),where A is the absorbance. The fluorescence intensity is directly relatedto the intensity of the excitation light and directly proportional toabsorption under condition of low absorbance. Fluorescence is proportionalto the number of absorbing molecules. The ratio of the number of photonsemitted (energy) to the number of photons absorbed (energy) is calledquantum efficiency where quantum efficiency is a property of thefluorophores.54Some of the factors that may influence the intensity of fluorescenceare solvent effects and relaxation, dynamic and static quenching, highoptical density and inner filter effect. Details of these factors arediscussed in the text, PrincIple of Fluorescence Spectroscopy (Lakowicz,1983).Some of the main attributes of fluorescence as a tool in diagnosis isits specificity and sensitivity compared to other “non-invasive” methods.Fluorophores absorb and emit light at characteristic wavelengths andtherefore can be selectively excited and detected in a complex compositionsuch as tissue. In fluorescence diagnosis, the characteristics of thefluorophores within the tissue as well as the optical properties(absorption and scattering) must be considered. Fluorescence signals arecomplicated by interactions of scattering with fluorescence, absorption andreabsorption of other molecules in the tissue where these effects must beconsidered. In principle, fluorescence measurements can be made on thefluorescence intensity of the excitation spectra, emission spectra orfluorescence intensity as a function of delay time after pulsed excitationat a fixed excitation and emission wavelengths.55Chapter 4: EXPERIMENTAL DESIGN AND RESULTS4.1. Thesis ObjectiveThe overall goal of this thesis was to study the mechanism of earlylung cancer detection with and without fluorescent tumour localizing drugs.The major objectives were:1) To test the hypothesis that detection and localization of early lungcancer can be achieved using low dose Photofrin without skinphotosensitivity.2) To test the hypothesis that laser-induced autofluorescence issignificantly different between normal, pre-cancerous and cancerousbronchial tissues.3) To determine the optimal emission wavelength(s) to differentiatebetween normal, pre-cancerous and cancerous tissues if differences inlaser-induced autofluorescence exist as in postulate 2.4) To determine conceptually (if postulates 2 and 3 are correct) how thespectral differences can be exploited for detection of early lungcancer and dysplasia using imaging and non-imaging methods.4.2 Detection of Early Lung Cancer Using Non-Skin Photosensitizing Dose ofPhotofrin4.2.1 RationaleAs discussed in chapter 2, detection of early lung cancer presents adiagnostic challenge even for an experienced bronchoscopist. Fluorescencetumour markers such as Hpd and Photofrin have been used to facilitate the56detection of these early lesions. However, the dose at which detection hasbeen achieved using methods as discussed in chapter 2.3.2 was very high (2-3 mg/kg intravenously). At these doses, Hpd and Photofrin result in skinphotosensitivity which lasts for four weeks or more. Although this is nota major problem when using the drug for treatment (PDT), it is clinicallyunsuitable for diagnostic purposes.To determine if skin photosensitivity can be avoided, ratiofluorometry was performed in patients with known or suspected lung cancerusing a much lower dose of Photofrin (0.25 mg/kg). The choice of this drugdose was based on the theoretical estimation that a Photofrin dose of lessthan 0.34 mg/kg is below the threshold of photodynamic action and hence maynot cause skin photosensitivity (Potter, 1990).Ratio Fluorometer ProbeThe schematic diagram for the ratio fluorometer probe is shown infigure 3. The principle of ratio fluorometry has been discussed in chapter2, section 2.3.2. In its original design for fluorescence bronchoscopy, atwo channel bronchoscope was used, one channel to conduct the illuminationfibre, while the second channel was used for the collecting fibre. Thishowever, restricted the range of examination because of the largebronchoscope size and loss of angulation with the two fibres. Amodification was then made to this design. Using a beam-splitter one ofthe two fibres was eliminated by using the imaging guide of the bronchscopeto collect the fluorescent light from the bronchial surface. Thismodification allowed the use of a bronchoscope smaller than the doublechannel bronchoscope and extended the range of the examination.57Figure 3 Schematic diagram of the ratio fluorometer probe.Dichroic RedMirror FilterRed DigitalMeterFluorescenceLensGreenFluorescencePhoto multiplierRFCFAudioOutputDC Auto—Amplifiers rangingCircuitAnalogDivider(Ratio)—HV58Patient SelectionTwenty patients with known or suspected lung cancer were selected forthis study majority of cases. Most of these patients had positive sputumcytology and radiologically occult lung cancers or marked atypia in thesputum cytology. Ratio fluorometry was performed 24 hours afterintravenous injection of 0.25 mg/kg Photofrin.Fluorescence BronchoscopyFollowing white light bronchoscopy examination, fluorescencebronchoscopy was carried out using the modified ratio fluorometer probe.405 run light from a Krypton-ion laser at 15 mW/cm2 irradiance, wastransmitted to the tip of the fiberoptic bronchoscope (Olympus BF1O) via a850 um quartz fiber with a diverging microlens (Laser Therapeutics Inc.,Buellton, CA). The divergence angle of the microlens fibre was 45° with aillumination spot size of approximately 0.53 cut2. This fibre was insertedinto the biopsy channel of the bronchoscope. The emitted fluorescence wascollected and transmitted to the ratio fluorometer by the imaging guide ofthe fiberoptic bronchoscope. Fluorescence intensities at 520 ± 10 nut and690 ± 10 nut, as well as their ratio (690 nm/520 nut) were displayed on adigital read-out. The red-green ratio value is independent of intensity,distance, and angle between the excitation light at the bronchial tip andthe bronchial surface. Before scanning the tracheobronchial tree, the redgreen ratios of at least two apparently normal areas as determined bywhite-light bronchoscopy were first measured. Five readings were taken ateach area which were then biopsied for pathologic confirmation.Measurements of the ratios were then made throughout the bronchial tree. A59preliminary clinical-pathologic study showed that the highest coefficientof variations of the red-green ratios from normal areas was approximately20%. Therefore a red-green value of 1.5 times above the mean value of thenormal sites was considered to be potentially significant. These areaswere later biopsied for histopathological confirmation.Skin Photosensitivity TestA test for skin photosensitivity was performed before and within 4hours after Photofrin administration. Testings was performed on the lowerback of the patients. A 500 W Sylvania projector lamp (Sylvania,Drummondville, Quebec) with a filter (Schott KG-4, Schott Optical GlassInc., Duryea, PA) to remove infra-red light was used as the light source.Areas of 1.5 cm x 2.5 cm of the patient’s back were exposed to light dosageof 15 - 30 J/cm2 in 5 J/cm2 increments. This dosage was chosen based on aprevious study of patients who had received 2 mg/kg of Photofrin. Whentheir skin could tolerate 15 J/cm2 of test light, they could be exposed tooutside light without any skin reaction (Lam et al., 1990a). To reduce theprobability of a false negative skin test due to the variability ofindividual skin responses and difference in climatic conditions betweenVancouver and other places, the highest test dose was increased to 30J/cm2. The power output of the light source was measured by a radiometer(IL1700, International light, Newburyport, MA). Skin reaction was read at24 hours after light exposure and graded as 1+, 2+, and 3+. A 1+ reactionis equivalent to the minimum erthema dose (MED) and was the light dosewhich induced a mild reddening of the skin without edema, blistering ornecrosis. A 2+ reaction referred to moderate redness with edema but no60blistering or necrosis. The 3+ reaction referred to redness, edema, andblistering (Lam et al., 1990a). Since, normal skin without Photofrin willnot show any reation to natural light, no skin photosensitivity test wereperformed prior to Photofrin administration.A total of 12 carcinoma in situ lesions were detected in these 20patients who were given 0.25 mg/kg of Photofrin. There were no falsenegative results (negative predictive value 100%). Of the 12 carcinoma insitu lesions, 7 were from patients with radiologically occult sputumcytology positive lung cancer, 4 from patients with residual microscopiccancer in the bronchial resection margin after surgery and one was from asynchronous In situ carcinoma in a patient with a large invasive tumour.Figure 4 shows the red-green ratios of pathologically confirmed areas fromthese 20 patients. The median red-green ratio of carcinoma in situ lesionswas 2.8 times above normal. Areas of dysplasia were also detectable. Fiveof the biopsies showed severe dysplasia and 3 showed mild dysplasia. Theselesions have a median red-green ratios of 1.8 times above normal. Therewas only one false positive biopsy from an area with inflammation.No skin photosensitivity was detected in all 20 patients both ontesting with the simulated light source and upon subsequent exposure tonatural (sun) light. The lack of complication in 20 consecutive patientsruled out the probability of skin photosensitivity reaction higher that 15%(Hanley and Lippman-Hand, 1983). Furthermore, this result is in agreementwith the theoretical estimation that a Photofrin dose of < 0.34 mg/kg isbelow the threshold of photodynaniic action and hence may not cause skinphotosensitivity (Potter, 1990).61Figure 4 Red-green ratios of pathologically confirmed areas from 20patients who had received 0.25 mg/kg Photofrin intravenously.(-) represents the median red-green ratios of carcinoma in situlesion from the normal background. (cis = carcinoma in situ).A total of 12 carcinoma in situ lesions were detected in these20 patients, with a median red-green ratios of 2.8. Fivebiopsies showed severe dysplasia and 3 showed mild dysplasia,with a median red-green ratios of 1.8 times above normalbackground. The normals were from bronchoscopically apparentnormal areas with the red-green ratios arbitrary set toapproximately 1. The red-green signal ratios in the verticalaxis represent the red-green ratios of the abnormal sitesversus the control normal sites.Red-Green Ratios of Pathological Confirmed Areas(With 0.25 mg/kg Photofrin)10:80-4rM54.V y. +++++1 —I 111111111 I I 111111 111111 11111111 I II I 111111 11111 111111111110 I I ICis Dysplasia Inflammation624.2.2 Mechanism for Detection by Ratio FluorometryAlthough dysplasia and early lung cancer can be detected usingPhotofrin with no apparent skin photosensitivity, the mechanism of cancerdetection based on this principle was not fully understood. The mechanismof tumour detection by ratio fluorometry was therefore investigated.A total of 52 patients were studied before and after administrationof low and high dose photofrin using ratio fluorometry. Fluorescencebronchoscopy was carried as described previously. In addition to the red-green ratios, the red fluorescence at 690 run and green fluorescence at 520run were also recorded. Care was taken that the tip of the bronchoscope waskept about 1 cm from the bronchial surface and the measurements wererecorded during the same phase of the respiratory cycle.The individual red (690 run) and green (520 run) fluorescenceintensities of the tumour area expressed as a ratio of the normalbackground before and after administration of 0.25 mg/kg and 2 mg/kg ofPhotofrin are shown in figures 5 and 6, respectively. In figure 5, a valueabove 1.0 is indicative of a higher red fluorescence (and hence morePhotofrin) in the tumour area compared to normal, while for figure 6, avalue less than 1.0 is indicative of a lower green fluorescence in thetumour area. A total of 28 tumours were studied without Photofrin, 11tumours studied with 0.25 mg/kg Photofrin and 7 tumours with 2 mg/kgPhotofrin. Without any photofrin, both large tumours and carcinoma in situlesions had decreased red fluorescence. With 0.25 mg/kg Photofrin, thecarcinoma in situ lesions have an increase in red fluorescence, on averageabout 1.7 times higher than the non-tumour area. Most of the large tumoursdid not show a significant increase in red fluorescence.63Figure 5 Red (690 rim ± 10 rim) fluorescence intensity of tumours as aratio of the normal background without and with (0.25 mg/kg and2 mg/kg) Photofrin. A total of 28 tuiuours were studied withoutPhotofrin, 11 tumours studied with 0.25 mg/kg Photofrin and 7tu.inours with 2 mg/kg Photofrin. With 0.25 mg/kg, the carcinomain situ lesions have an average of 1.7 times higher redfluorescence than the non-tumour area. Most of the largetumour do not show a significant increase in red fluorescence.Red Intensity of Tumour-Normal Ratios as aFunction of Photofrin Dose4-0 visible tumour• carcinoma in situii- :0.25 2.0Photofrin (mg/kg)64Figure 6 Green (520 run ± 10 run) autofluorescence of the tumour as aratio of the normal background without and with (0.25 mg/kg and2 mg/kg) Photofrin. With and without Photofrin, the mean greenautofluorescence of tumour is only about 30% of the normalsurrounding tissue.Green Intensity of Tumour-Normal Ratios as aFunction of Photofrin Dose1.00 visible tumour0.9• carcinoma in situ_0.80.70.6• 8.E 0.50•0..fl4 0• 0.0.30 00.2 0000.0 I I0.0 0.25 2.0Photofrin (mg/mg)65Even at 2 mg/kg Photofrin, more than half of the large tumour did notretain more Photofrin than the adjacent 11normal” tissue. With and withoutPhotofrin, the green autofluorescence is much lower in tuniours compared tonormal tissues. The mean green fluorescence of tumours is only about 30%of the normal surrounding tissue. These results suggest that the elevatedred-green ratios in tumours compared to normal tissues were mainly due to adifference in the green autofluorescence. In the original study by Profioand co workers (Profio et al., 1984), the green autofluorescence wasassumed to be similar for normal and tumour tissue. The results in thisstudy show that this assumption was incorrect. A significant difference inautofluorescence intensity in the green region of the emission spectrumexists between normal and malignant bronchial tissue.4.3 Fluorescence SpectroscopyTo confirm that differences in tissue autofluorescence exist betweennormal, pre-cancerous and cancerous tissues, fluorescence spectroscopy wasperformed in vivo. Potentially abnormal areas were first localized usingratio fluorometry. Following the spectroscopic measurements (discussedbelow), the areas were biopsied for pathologic diagnosis. The Heliumcadmium laser was studied as a more practical alternative to the Kryptonion laser. In addition to being much cheaper, the 442 rim wavelength lightdoes not induce red fluorescence in the fiberoptic bundles of thebronchoscope and hence a special illumination light guide is not necessary.For the 405 rim light, it was delivered to the tip of the fiberoptic66bronchoscope using a 600 urn-core-diameter quartz fibre with a divergingniicrolens (Laser Therapeutics, Buellton, CA).The emitted fluorescence was collected by means of a flat-cut,polished 400 urn core-diameter quartz fibre 1 cm distal to the excitationlight source. Care was taken to ensure the collecting fibre is in thecentre of the illuminated area. Fluorescence measurements were made withthe tip of the collecting fibre just in contact with the tissue surface.The emitted autofluorescence was focussed onto the entrance slit ofthe spectrograph (Jarrell-Ash Monospec 27, with a 300 g/rnin gratings, blazedat 500 nm, and a 0.29 rim/element resolution) with a f/l.5 collimating lensand an f/3.14 focussing lens to match the f/3.8 spectrograph. Thespectrograph was fitted with a 2 mm-width slit. A low-fluorescence, longpass filter (cutoff wavelength at 470 rim, Oriel, Strattford, CT) was placedbetween the lenses to block off the scattered and reflected excitationlight. The spectrograph was adjusted to cover the 450 rim to 750 rim range.The dispersed spectrum was recorded with an intensified 1024-elementsilicon linear diode array detector (EG & G PARC, Model 1455R 700HQ). Thedetector was interfaced to an Optical Multi-Channel Analyzer (OMA III,Model 1461/2 detector interface/controller with a 14-bit analog digitalconverter). The OMA controller was interfaced to a host computer (IBM-PC80386) which allows data storage and spectral display/processing. Aschematic of the spectroanalyzer is shown in figure 7.Background subtraction was performed on each tissue spectrum duringthe acquisition mode before transmitting the data to the host computer tocorrect for the dark noise of the detector. Spectral calibration wasperformed by using the spectrum of a low-power, low-pressure mercury-argon67lamp (Oriel, Strattford, CT, Model 6035). Non-uniformity of the detectorfor the spectrum range of interest was checked and the detailed aspectregarding correction and calibration of the intensity are discussed inappendix 7.3. The detector response was found to be uniform within thespectral range of interest. The fluorescence intensities in all thespectra are in arbitrary units. Fluorescence spectra were always obtainedby using the same gain of the diode array detector.Figure 8 shows a typical autofluorescence spectra of a normalbronchus and carcinoma in situ lesion from the same patient using anexcitation wavelength at 405 nm. The in vivo autofluorescence spectrainduced by 442 run light, representative for a normal bronchus, a severelydysplastic lesion, and a carcinoma in situ lesion from a patient are shownin figure 9. The shape of the emission spectrum induced by the 405 nmlight was identical to that with the 442 run light. Both emission spectrashow an overall decrease in fluorescence intensity mainly in the greenregion of the visible spectrum (500 nm to 580 run). For the normalbronchus, the spectra have a peak at about 520 run, decreasing to a minimumat about 580 run and rising to a smaller peak between 600 run to 610 rim.Statistical AnalysisA total of 284 spectra from histopathogically confirmed areas wereconverted into ASCII data to be used as a learning set for discriminantfunction analysis. Each fluorescence spectrum consisted of 1025 intensityvalues spanning the wavelength range of 450 rim to 750 nm (i.e the ASCIIdata consisted of 1025 pairs of data points, wavelength and intensityvalues).68Figure 7 Schematic diagram of the spectroanalyzer system used to collectin vivo autofluorescence spectra.COMPUTER SYSTE/MON TORSPECROORAPH(2mm)f/3.14Long Pass FHtertf/1.5LIGHT SOU RCESKr LASERHe—Cd LASERLOMA1DETECTOR CONTROLLER69Figure 8 In vivo autofluorescence spectra for normal bronchus andcarcinoma in situ lesion excited by using a Krypton-ion laserat 405 run (a.u. = arbitrary units). There is a significantdecrease in the green autofluorescence for the carcinoma insitu lesion compared to normal. These differences are muchsmaller in the red region of the fluorescence spectra.Fluorescence Intensity as a Function of Wavelength(Excitation at 405 nm)1615141312I)C.)UC)U43210420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)Normal70Figure 9 In vivo autofluorescence spectra for normal bronchus, severelydysplastic and carcinoma in situ lesions excited by using aHeliuni-cadmiuni laser at 442 nm (a.u. = arbitrary units). Thereis a significant decrease in the green autofluorescence for the1615141312z1oUUV43210carcinoma in situ lesion compared to normal.Fluorescence Intensity as a Function of Wavelength(Excited at 442 nm)420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)Normal71The first 100 and last 100 pairs of data were trimmed off since inthese regions, there were no signal from these pixels. The data wereclassified according to the histopathology into seven categories as shownin Table I and organized into data sets, each consisting of only onespectrum class. The data were then normalized and re-sampled so that theoutput data set consisted of one intensity value per rim from 461 run to 660run, resulting in 200 intensity values for each spectrum class.These data were normalized in a variety of ways as follows: (i) bysetting the average intensities between 530 run to 570 run (green region) tobe the same for all spectra; (ii) by setting the average intensitiesbetween 600 nm to 660 run (red region) to be the same for all spectra; and(iii) by setting the average intensities between 500 run to 660 rim to be thesame for all spectra and the procedure was set such that if the averageintensity values at 470 run was less than a set value, that particular casewould be ignored in the calculations.A commercially available program BMDP (biomedical Data Processing,BMDP Statistical Software, Los Angeles, California) was used to generateand test discriminant functions from these learning sets. A stepwiseanalysis of the data was performed using the 7M discriminant analysisroutine of this software package. This routine performs a stepwisediscriminant analysis between two or more groups. The purpose of thestepwise discriminant procedure is to select those features that providethe best discrimination, while disregarding those that make little or nocontribution to the classification process. Features are selected on thebasis of their ability to separate the groups in feature space and arechosen in a stepwise fashion. Details of discriminant function analysis,72the selection procedures and the assumptions this analysis makes about thedistribution, size and forms of the data are described in the Appendix,7.1.Figure 10, and figure 11 are the graphs of the mean curves for eachclass normalized in the red and green regions respectively. The normalized(500 mu to 660 rim) average intensity as a function of wavelength for eachclass is shown in figure 12. The dip or valley —575 rim to —585 rim for allthe curves can probably be attributed to hemoglobin absorption. The unnormalized mean curves for each class were also calculated and this isshown in figure 13. A non-parametric statistic, the Mann-Whitney rank sumwas first applied. The p-values for each class at 10 run interval wereplotted in figures 14 and 15. This gives a measure of the similaritybetween the intensity distribution of the normal and the specified class ata particular wavelength. A p-value of less than 0.05 is considered to besignificant, and a value of less than 0.005 is considered to be verysignificant. The p-values for the unnormalized data were also examined butnot plotted. Significant differences were observed between pathologygrades whether normalized or un-normalized data were used. Thesedifferences are due to the average intensity differences between theclasses. To check the stability of the peaks and valleys of the curvesacross classes, the spectra data were normalized at 600 rim - 660 rim,normalized as described above (normalization procedure (iii)). All thecurves peak at the same locations as shown in figure 16.A stepwise disciminant analysis was performed on these data sets,each consisting of the features values calculated for all spectra. TableII gives a description of the features (individual features will generally73be referred to by their abbreviated description, as listed in the table).The features used for the computation of the linear classificationfunctions were chosen in a stepwise fashion. The software package allowsthe user to force the program to use all the features in the dicriminantfunction, or let the program select the most discriminating features to beused in the discriminant function. The user can also select a subset ofthe features to be used to form the discriminant function. In thisanalysis, BMDP was programmed such that during forward stepping all (ormost, some not included due to failure of tolerance test) the features wereentered into the dicriminant function (DF), while during backward stepping,only features that significantly contributed to the discriminantperformance were allowed to remain in the discriminant function. The Fstatistic was used to determine if a feature contribute significantly tothe discriminant function performance. The significance level was set to0.5% (0.005). The program provided a jackknifed classification table(i.e., every case is classified by using a discriminant function generatedfrom the remaining n-l cases). This gave the classification accuracy ofthe DF generated when the data were subdivided into learning and test sets(Lachenbruch, 1975). The jackknife procedure is described in Appendix 7.1.The jackknife classification accuracy is a more correct representation ofhow the discriminant function would perform on other test sets(Lachenbruch, 1975). Empirical studies suggested that the error of theclassification accuracy of the discriminant function on new data set shouldbe within ± 20% of the jackknife classification error (Lachenbruch, 1975).The goal of this analysis was to select the feature(s) that willprovide the best discrimination for detection of pre-cancerous and74cancerous lesions. Table III summarized the results obtained following thevarious discriminant analyses. Included in this table is the jackknifeclassification result (described above) and a histogram classificationresult in which the cut-off for misclassifications (the number of falsepositives) is set to be approximately 5, and the corresponding number offalse negatives and true negatives were then selected from the histogram.The actual feature and classification data used to generate and test thediscriminant functions for this table were from a subset of the 284 spectracollected. The data were classified as normal or abnormal. Abnormal setsconsisted of the different grades of dysplasia, carcinoma in situ andinvasive tumour. Metaplasia and inflammation were not reported in thistable. One requirment of the discriminant function analysis is that alarge data set with approximately 10 to 20 data points (spectra) perfeature be used in the analysis. If this criterion is not met then thedicriminant function that was generated in the learning set will not givethe same performance on other test sets of data. As shown in Table III,not more than four features combinations were used at any one time as inputfeatures to generate the discriminant function. There is no simple methodto determine which combination of features are the most discriminating,other than testing all possible combinations. The data reported in TableIII demonstrated that the optimal emission wavelengths to differentiatebetween normal, pre-cancerous and cancerous tissues are in the green regionbetween 525 nm and 570 rim and in the red region between 600 rim to 660 rim.Increasing the band widths of both the red and green regions was shown toimprove the discrimination.75Table I Classification of a database of 284 spectra according to thehistopathologic grades and organized into data sets eachconsisting of only one spectrum class.Class Pathology N1 Normal 832 Metaplasia 353 Mild dysplasia 204 Moderate dysplasia 95 Severe dysplasia 346 Carcinoma in situ 237 Tumour 8076Figure 10 The mean curves for each class (classified as in Table I)normalized in the red regions (600 run - 660 run). The averageintensities between 600 run to 660 nni are set to be the same forall spectra. (a.u. = arbitrary units).Average Intensity as a Function of WavelengthNormalized at 600 nm - 660 nm1413121110I!321I I I I490 510 530 550 570 590 610 630 650 670 690Wavelength (nm)• normal• metaplasiao mild• moderate0 severeA CisV tumourV77Figure 11 The mean curves for each class (classified as in Table I)normalized in the green regions (530 rim - 570 run). The averageintensities between 530 rim to 570 run are set to be the same forall spectra. (a.u. = arbitrary units).12- • normal1 1 • metaplasia11 omjld10 • moderatesevere—‘9 AcjSV tumour;I.2-1-0—490Average Intensity as a Function of WavelengthNormalized at 530 nm - 570 nmI I I I I510 530 550 570 590 610 630 650 670 690Wavelength (nm)78Figure 12 The average intensity as a function of wavelength for eachclass normalized at 500 run to 660 run (a.u. = arbitrary units).3.503.152.80_.2.100.700.350.00Normalized Average Intensity as a Function of wavelength(Normalized between 500 nm - 660 nm)490 510 530 550 570 590Wavelength (nm)O normal• metaplasia• mild dysplasiaO moderate dysplasiaA severe dysplasia“ carcinoma in situA tumour610 630 650 67079Figure 13 The unnormalized average curves for each class (a.u. =arbitrary units).Average Intensity as a Function of WavelengthUnnormalized13 • normal• metaplasia• omild12• moderate0 severe10 AcisV tumour•-4oI310I I I I I490 510 530 550 570 590 610 630 650 670 690Wavelength (urn)80Figure 14 The p-values for each class normalized in the red region (600rim - 660 rim) at 10 rim interval. The vertical axis is in logscale.P-Value Normalized at 600 nm - 660 nmI10010-110-21 0-10-410-5• metaplasiao mild• moderate° severeA CiSV tumouri/I/ A..0I, i AA•7\ N //hA\\./ \ \\/ \490 510 530 550 570 590 610 630 650Wavelength (nm)670 69081Figure 15 The p-values for each class normalized in the green region (53010010-110-21 0-310-41 0-rim - 570 rim) at 10 rim interval. The vertical axis is in logscale.P-Value Normalized at 530 nm - 570 nm\• metaplasiao mild• moderate° severeA CistumourIA\\ \..\I I.•I!“A\ A -A\490 510 530 550 570 590 610 630 650 670 690Wavelength (nm)82Figure 1612420480The average intensity for each curve normalized in the redregion (600 run - 660 run). The average intensities between 600rim to 660 run are set to be the same for all spectra and if theaverage intensity values at 470 rim was less than a set vaule,that particular case would be ignored in the calculations. Themoderate dysplasia category was ignored in these calculation.The peaks and valleys of the curves are stable across theclasses. All curves peak at the same location.arbitrary units).(a.u. =Autofluorescence Spectra of In Vivo Bronchial TissueNormalized Over 600 nm - 660 nm• normal• metaplasiao mildD severeA cis‘ tumour0I.NI0z500 520 540 560 580 600 620 640 660Wavelength (nm)83Table II Description of the features used in the discriminant functionanalysisFeatures DescriptionIntensity at:gi 500 rim — 520 rimg2 500 rim — 550 rimg3 500nm—570nmrl 630 rim — 660 rimr2 610 rim — 660 rimr3 600 rim — 660 rimgl 501 nm — 520 rimg2+ 521 rim — 550 rimg3 551 rim — 570 rimyellow 571 nm — 600 rim+rl 601 nm — 620 nm+r2 621 rim — 640 rim+r3 641 rim — 660 rimNormalization I Normalization constantlog log transformation of intensity valuesHue hue transformation of intensity valuesModified Hue Modified hue transformation of intensityvalues84Table III Summary of the discriminant function analysisData Input Featuresfeatures used Classification Results. ++ . ++Jackknife Histogramunnormalized gl—3, rl—3 g3 92.5% (12) 94% (15)normalized gl—3,rl—3 gl—3, ri 77.5% (56) 73.1% (67)(600—660nm)normalized gl—3, rl—3 gl—3 94.8% (13) 94.8% (13)(600—660nm) normalized// normalized#unnormalized gl,rl gi 92.8% (18) 92.4% (19)unnormalized g2, r2 g2 95.2% (12) 94.4% (14)unnormalized g3 r3 95.2% (12) 95.6% (11)normalized every Snm feature* 73.5% (66) 75.9% (60)(500—600nm)unnormalized 535, 560 74.3% (64) N/A585, 605unnormalized —2Onm 510, 560 95.6% (11) 96% (10)unnormalized g1, r1 g1 92.8% (18) N/Aunnormalized g2, r2 g2 95.6% (11) N/Aunnormalized g3, r3 g3 95.2% (12) N/Aunnormalized Hue 1—3k Hue 2 70.7% (73) 94% (15)unnormalized modified Hue 3 71.5% (71) 92.7% (16)Hue+ 1-3* 515@ + green 1 = 501—520 nm525 + green 2 = 521—550 nm530 + green 3 = 551-570 run575@+ yellow = 571—600 nm590 + red 1 = 601—620620 + red 2 = 621—640 run635@ + red 3 = 641—660 run645@largest F to remove value85Table III Summary of the discriminant function analysis (continued)Data Input Featuresfeatures used Classification ResultsJackknife Histogramunnormalized log (gl—3) log (gl—3) 91.2% (22) 91.2% (22)log (rl—3) log ri)unnormalized log (gi) both 90.8% (23) 92.4% (19)log (ri)unnormalized log (g2) both 90.8% (23) 92.8% (18)log (r2)unnormalized log (g3) both 90.8% (23) 93.2% (17)log (r3)unnormalized log (g2) g3,log(g3) 94.8% (13) 96.0% (10)g3, log(r2)r3, ratio3unnormalized log(g2),g3 g3 95.2% (12) 94.0% (14)ratio 3 ratio3Hue (1-3) Hueunnormalized Hue (1—3) Hue 1 69.9% (75) 95.2% (12)unnormalized modified Hue 1 69.9% (75) 92.8% (18)Hueunnornialized #modified Hue 1 69.5% (76) 63.4% (30)Huenormalized# log (gl-3) log (gi) 69.5% (76) N/A++ ( ) denotes number of missclassified (false positive).also see explanation in text.864.4 Detection of Dysplasia and in situ Carcinoma by Ratio FluorometryUsing Tissue Autofluorescence Alone4.4.1 Non-Imaging MethodsTo test the hypothesis that early lung cancer can be detected basedon differences in tissue autofluorescence alone without Photofrin, eighty-two volunteers who had previous exposure to asbestos, diesel fumes and/ortobacco smoke in their work-place were studied using ratio fluorometry. Ofthe 82 individuals, 25 were non-smokers, 40 were ex-smokers, and 17 werecurrent smokers with mean age values of 52, 55, and 49 years, respectively.Fluorescence bronchoscopy was carried out using a ratio fluorometeras described in the previous section. In this study, 442 nm wavelengthfrom a Helium-cadmium laser was used as the excitation light. All abnormalareas were biopsied to confirm the presence or absence of dysplasia orcancer. In addition, one or more random biopsies were taken fromapparently normal areas for pathological examination. Before andimmediately following bronchoscopy, a sputum sample was taken forcytological examination.From the 82 subjects a total of 237 biopsies were taken and analyzed.The red-green ratios from pathologically normal areas, pre-cancerous, andcarcinoma in situ are shown in figure 17. The progressive increase in theratios correlates with the pathologic grades. The red-green ratios fromareas with moderate dysplasia, severe (marked) dysplasia and the carcinomain situ lesions were significantly higher than the normal areas (p < 0.05).The elevation in the red-green ratio is attributed to the lower green87autofluorescence in the pre-cancerous and cancerous lesions. Theproportion of subjects classified by smoking status and bronchoscopyoutcome is shown in figure 18. There was a progressive increase in theproportion of subjects with nietaplasia or mild dysplasia among current andex-sniokers. One or more sites of moderate or severe (marked) dysplasiawere found in 12.5% of the ex-smokers and current smokers but none in thenon-smokers. Carcinoma in situ was found in two of the ex-smokers. Preand post-bronchoscopy sputum cytology failed to detect these small precancerous or carcinoma in situ lesions.4.4.2 Fluorescence ImagingAlthough ratio fluorometry is an important concept, it has manylimitations. One of its major drawbacks is that a “reference” site(presumably normal by visual examination) is needed in the same subject tocalibrate the ratio fluorometer probe in vivo. If the control sites happento be abnormal, the ability to detect carcinoma in situ or severe dysplasiamay be diminished. As a non-imaging device, it does not provideinformation regarding the extent and size of the lesions. In addition,being a single pixel probe, very small early lesions can be missed due to afield averaging effect.The idea of using the spectroscopy data in section 4.3 to develop animaging device was therefore tested. The main principle of such an imagingsystem is to induce tissue autofluorescence at the blue region of thespectrum (for example at 442 run) and to collect the weak fluorescence by animage intensified charge couple device (ICCD).88Figure 17 Red-green ratios from pathological confirmed normal areas,dysplasia and carcinoma in situ lesions of 82 volunteers. (cis= carcinoma in situ, * indicates that the red-green ratios fromareas with moderate dysplasia, severe dysplasia and thecarcinoma in situ lesions were significantly higher than thenormal areas (p < 0.05). Red fluorescence is measured at 690rim ± 10 rim and green fluorescence at 520 rim ± rim).Red-Green Ratios from Pathological Confirmed AreasUsing Fluorescence Bronchoscopy (Ratio Fluorometer)*0V0I;:T**1.0 I I I INormal Metaplasia Mild . Moderate Severe . CisDysplasia Dysplasia Dysplasia89Figure 18 The proportion of subjects classified by smoking status andfluorescence bronchoscopy (ratio fluorometry) outcome. One ormore sites of moderate or severe dsyplasia were found in 12.5%of the ex-smoker and current smoker but none in non-smokervolunteers. Carcinoma in situ lesions was found in 2 of theex-smokers.10090807060503020100Smoking Status and Prevalance of “Lesions”Using Fluorescence Bronchoscopy (Ratio Fluorometer)Normal Metaplasia Mild Moderate SevereDysplasia Dysplasia DysplasiaNon-SmokerV//Zzi Ex-SmokerC-SmokerCis90The spectral information can then be utilized by placing appropriateband-pass filters in front of the ICCD camera. A clinical pre-prototypelung imaging fluorescence endoscope (LIFE) based on these principles oftissue autofluorescence is shown in figure 19 (Palcic et al., 1991, Palcjcand Lam, 1991, Hung et al., 1990, Lam et al., 1990b).Figures 20a and 21a are examples of a normal bronchus and carcinomain situ lesions, respectively, detected using the LIFE device. Theabnormality could not have been detected using conventional bronchoscopy asshown by the white-light bronchoscopy images (figures 20b and 2lb).The performance of the fluorescence imaging system was compared withwhite light bronchoscopy to determine if the sensitivity of detectingdysplasia or carcinoma in situ can be improved using fluorescence imagingin conjunction with conventional white-light bronchoscopy. The sensitivityand specificity of the white-light bronchoscopy versus fluorescence imagingwere compared using the bronchial biopsy pathology as the “gold standard”.A total of 94 subjects were examined. There were 53 patients withknown or suspected lung cancer and 41 volunteers not known to have lungcancer prior to entrance into the study. Details of their age and smokinghistory are shown in Table IV.Conventional fiberoptic bronchoscopy under white-light (xenon) lampillumination was first carried out as described in the previous section.Fluorescence bronchoscopy using the LIFE device was then carried out.Briefly, laser light at 442 rim (Helium-cadmium laser) was delivered to thebronchial surface via the illumination bundle of an Olympus BF2ODfiberoptic bronchoscope. The emitted fluorescence collected by the imagingbundle of the bronchoscope was spectrally divided by two filters of91different wavelengths, one at 480 rim to 520 rim (green region) and the otherin the red region ( 630 rim). The respective images were acquiredsequentially by the ICCD and digitized by an imaging board residing in acomputer. Using a non-linear discriminant function combination of the redand green intensity values (gray levels), a pseudo-image was formed whichwould allow the delineation of abnormal areas when displayed on a highresolution RGB analog monitor.All abnormal areas found using white-light and/or fluorescencebronchoscopy were biopsied for pathologic confirmation. In addition, oneor more random biopsies were taken from the visually apparent (white lightand fluorescence) normal areas for pathological examination. In thefluorescence images, the exact area of each biopsy was manually delineatedon the RGB monitor and the average red and green intensities (gray levels)over the area were calculated. These average values were used in a nonlinear discriminant function analysis to create a value which can be usedto separate the normal from abnormal tissues sites.A total of 328 biopsies were collected from the 94 subjects. 173 ofthese biopsies were found to be normal, 14 were found to show milddysplasia, 33 were found to show moderate dysplasia, 15 were found to showsevere dysplasia, 29 were carcinoma in situ and 64 were found to showinvasive tumour. The prevalence of dysplasia and carcinoma in situ usingfluorescence imaging is shown in figure 22. In 15% of the lung cancerpatients, synchronous carcinoma in situ was found in addition to the largeinvasive cancer; 8% of these patients had moderate dysplasia and 6% hadsevere dysplasia. For the current smokers in this study, 40% had moderatedysplasia and 12% had severe dysplasia. Of the ex-smokers in this study,9225% had moderate dysplasia, 6% had severe dysplasia and 13% had carcinomain situ.The non-linear Discriminant Function values of the red and greenfluorescence intensities in normal tissues, mild, moderate, or severedysplasia lesions, carcinoma in situ or invasive cancer is shown in figure23. The difference in the tissue autofluorescence discriminant functionvalues were found to be statistically significant between normal tissuesand both dysplasia or cancerous lesion (p < 0.005). However, dysplasialesions cannot be differentiated from carcinoma in situ because of thesignificant overlap. But since moderate and severe dysplasia are precancerous lesions, clinically it would be important to be able to localizethese lesions as well.The sensitivity of the fluorescence imaging in detecting moderate orsevere dysplasia and carcinoma in situ was found to be at least 50% betterthan conventional white light bronchocopy (sensitivity of 71.5% forfluorescence imaging compared to 48.4% for white-light bronchoscopy). Thespecificity of 94% was the same for both methods.93Figure 19 Schematic diagram of the clinical prototype lung imagingfluorescence endoscope (LIFE).ACQUISITIONHe—Cd LASER iiCONTROLL IDIGITIZATION AND MACE PROCESSING DISPLAYrIIII III IL. JL J94Figure 20 (a) Example of a fluorescence image of a normal bronchus, wherethe green colour has been arbitrarily used to depict areas ofnormal tissue. (b) Whiteliht of the same bronchus.F : JL.— —Figure 2195(a) Fluorescence image of a bronchus showing an area ofabnormal fluorescence arbitrarily depicted as the reddish areafor which the corresponding biopsy showed the prescence ofcarcinoma in situ. (b) White1i of the same bronchus96Table IV Characteristics of the subjectsN YearsLung Cancer patients 53 63 ± 10 years*Current smoker 17 52 ± 10 years*Ex_smoker 16 57 ± 10 years*Non_smoker 8 50 ± 10 years* ex—smoker and current smoker have smoked more than a pack of cigarettesper day for > 20 years. Ex—smoker gave up smoking for more than 5 years.97Figure 22 The prevalence of dysplasia and carcinoma in situ detectedusing fluorescence bronchoscopy (imaging) as verified by biopsypathology.50454035302520151050Prevalance of Lesions UsingFluorescence Bronchoscopy (Imaging)//z Lung Cancer PatientsNon-SmokerxX1 Ex-SmokerC-SmokerMildDysplasiaCarcinomain situDysplasia Dysplasia98Figure 23 A group box of the non-linear Discriminant Function (DF) valuescalculated from the red and green intensity values (gray scale)for the pathology grades. The centre line of the boxesrepresents the median DF of the biopsy grade. The size of thebox indicates the fraction of DF values above and below theinterquartile values. The bars indicate the range of themeasured values for the biopsy site. The small circlesindicate the location of the outliners.BIOPSY GRADE VS DF VALUE2.0UZ 0.50 00.0L I I I-2.0 I I I INORMAL MILD MOD SEVERE CIS TUMOURBIOPSY PATHOLOGY99Chapter 5: DISCUSSION AND CONCLUSIONThe important findings in this thesis work are as follows:i) Dysplasia and carcinoma in situ can be detected using a low non-skinphotosensitizing dose of Photofrin.ii) The major mechanism of tumour detection using ratio fluorometry andPhotofrin was due to a significant reduction in tissueautofluorescence. To a much lesser extent preferential retention ofPhotofrin in tumours also play a role.iii) In vivo fluorescence spectroscopy showed significant differences inthe emission spectra between normal, pre-cancerous and canceroustissues when the bronchial surface was illuminated by a violet (405run) or a blue (442 run) laser light.iv) Spectral differences between normal, pre-cancerous and canceroustissues were characterized by an overall decrease in autofluorescenceintensity between 500 run to 580 run.v) Dysplasia and carcinoma in situ can be detected using laser-inducedautofluorescence alone without the use of exogenous fluorescent drugssuch as Photofrin.vi) Detailed discriminant function analysis of fluorescence spectra“points to” the best methods to detect dysplasia and carcinoma insitu using autofluorescence. This was confirmed by two separatemethods - ratio fluorometry and fluorescence imaging.1005.1 Tumour Detection Usin8 Non-Skin Photosensitizing Dose of PhotofrinOver the last several decades a great deal of effort has been spenton developing fluorescent tumour localizing drugs to facilitate thedetection and localization of early lung cancers. A number of fluorescentmarkers have been shown to accumulate preferentially in malignant tissues(Dougherty, 1984, Rail, 1957, Moore, 1947). To date, most of the clinicaland laboratory studies have been concentrated on the use of porphyrins, inparticular Hematoporphyrin derivative (Hpd) and its partially purifiedpreparation, Photofrin (Profimer sodium) (Dougherty, 1989, Dougherty etal., 1984, Lipson et al., 1961).Several non-imaging and imaging bronchoscopic devices have beendeveloped to detect early lung cancer using Hpd or Photofrin. Althoughclinical experiences with these devices suggested that small radiologicallyoccult cancers can be detected (Balchum et al., 1990, Kato et al., 1990,Kato and Cortese, 1985, Cortese et al., 1979), wider clinical applicationsof Photofrin as a fluorescent tumour marker have been impeded by theassociated skin photosensitivity. At the standard dose of 3 mg/kg Hpd or 2mg/kg of Photofrin, the same drug dose used in photodynamic therapy skinphotosensitivity can last for four weeks or more (Wooten et al., 1988,Razuin et al., 1987). This is a major drawback to use Hpd or Photofrinfluorescence as a diagnostic tool for the detection of early lung cancers.Theoretical calculations suggested that non-therapeutic low doses ofphotofrin (< 0.34 mg/kg) may be free of skin photosensitivity side effects(Potter, 1990). Preliminary studies in animals showed that 0.5 mg/kg ofPhotofrin (1/4 of the therapeutic dose) may be effective in the detectionof occult, non-palpable micrometastases in the regional lymph nodes of rats101(McGinnis et al., 1990) and in chemically induced urinary carcinoma of dogs(Bau.mgartner et al., 1987). Using the ratio fluorometer endoscopicdetection system, this present study showed, for the first time thatcarcinoma in situ and dysplasia can be detected with very low doses ofPhotofrin (0.25 mg/kg) with no apparent skin photosensitivity both on skintesting with an artificial light source and subsequent exposure to outsidelight (Lam et al., 1990a).While skin photosensitivity may be avoided by the use of very lowdose Photofrin, the contribution of low-dose Photofrin in cancer detectionwas small. The major mechanism of cancer detection by ratio fluorometrywas mainly due to the differences in the green tissue autofluorescencebetween normal and malignant tissues. The increase in red fluorescence dueto Photofrin was relatively small. This was confirmed by in vivospectroscopy as shown in figure 24. The slight increase in redfluorescence in tumour tissue following administration of photofrin is notvery large compared to the huge differences in green autofluorescence.5.2 Tissue Autofluorescence in DiagnosisThe idea that tissues autofluoresce is not new. In 1908, Wood notedvisible fluorescence in the skin when excited with a ultra-violet light(reference in Anderson and Parrish, 1982). Since then, “Wood’s lamp” havebeen used to diagnose erythrasma, tinea capitis, and visualizing subtlechanges in epidermal melanin pigmentation (Caplan, 1967). The fluorescenceof the skin observed under “Wood’s lamp” was reported to be largely due tothe dermis autofluorescence (reference in Anderson and Parrish, 1982). The102fluorophore(s) responsible for this autofluorescence are unknown. Humanskin fluorescence induced by a 325 nm laser light was reported to be due toelastin and collagen which are major components of human dermis that arealtered by age and photo-exposure (Leffell et al., 1988). Greenautofluorescence has also been observed in human epidermal cells (Feilner,1976). In 1933, Sutro and Burman reported that cellular tumours whichextended into the muscular layer of the stomach or rectum fluoresce from apurple to a deep brown and that a medullary carcinoma of the thyroidappeared purple against white in a zone of fibrous tissue (Sutro andBurnian, 1933) This “colour” difference seen by filtered ultra-violetirradiation had been used as an aid to surgeons and pathologists indetecting small areas of disease (Sutro and Burman, 1933). An attempt wasmade in 1944 to differentiate between benign and malignant lesions ofbreast tissue by application of these differences in colour phenomena(Herley, 1944). When there is epithelial degeneration, a decrease in bluecolour was observed and in carcinoma, where the cancer cells are tightlypacked, the tissue appears purple with varying intensity, passing into tan,orange, and brown (Herley, 1944). Strong red fluorescence was observed inthe cancerous tissues of the genitalia of women (Figge et al., 1944).Policard in 1929, had noted a brilliant red fluorescence in the centre of arat sacronia (Policard, 1929). This red fluorescence is characteristic ofhematoporphyrin. The endogenous porphyrin in tumours have been attributedto the presence of bacteria in the ulcerated and necrotic tissue (Harrisand Werkhaven, 1987). In recent years, Alfano and co-workers had reportedsubtle spectroscopic difference between normal and cancerous lung andbreast tissues in excised animal and human specimens (Tang et al., l989a,103Tang et al., 1989b, Alfano et al., 1988, Alfano et al., 1987). Using threedifferent excitation sources, 457.9 run, 488.0 run, and 514.5 run, theyreported that normal lung and tumour tissues exhibit differentautofluorescence spectra. Maximal peaks were observed at 496 run, 509 run,and 531 run for normal tissues, excited by 457 run, 488.0 rim, and 514.5 runlaser light respectively; a subsidiary maximum was observed at 606 rim. Forthe tumour tissues, maxima were observed at 503 run, 515 rim, and 537 rim,respectively (Alfano et al., 1988). However, in another report, using457.9 rim excitation source, the normal lung tissue was observed to exhibita maximum at 511 rim with a subsidiary maxima at 555 rim and 600 rim; whereasthe tumour tissue had a maximum at 514 run (Tang et al., 1989a). In thesame report, using 514.5 run, the spectrum profile of the normal lung tissuewas reported to have a maximal peak at 563 rim and 603 rim whereas thecancerous tissue was reported to have a maximal at 560 rim (Tang et al.,1989a). In an earlier report, the spectrum profiles of a normal human lungand tumour tissue excited at 488.0 rim were reported to have maximal peaksat 512 rim and 520 run, respectively, with the normal lung tissue exhibitingtwo subsidiary maxima located at 554 rim and 600 rim (Alfano et al., 1987).Thus, changes in the spectral shape appear to be strongly dependent onexperimental conditions. Significant differences in autofluorescenceintensity between normal and malignant tissues were not observed in thesestudies.The difference in spectra shape and intensity between Alfano’sstudies and this study may be due to the tumour size and possibledifference between lung parenchymal tissue and bronchial tissue. Inaddition, Alfano and co-workers examined lung tissues in vitro several104hours after excision (Tang et al., 1989a, Tang et al., 1989b, Alfano etal., 1988, Alfano et al., 1987). Freezing and thawing of bronchial tissuesat variable time after excision reduced the differences in fluorescenceintensity between malignant and normal bronchial tissues as well asaltering the spectral shape (Figures 25 and 26, Appendix 7.2). Withfreezing, there is a slight accentuation of the emission peak at about 600nni, and with thawing there is a decrease in intensity at 1 and 3 hour. Incontrast, the fluorescence intensity of tumour tissue from the same patientincreased with thawing. An alteration in the spectral shape of deep-frozentumours was observed by other investigators in the characterization ofbrain tissues. They suggested that disruption of cell membranes at thawingmay induce changes in the membrane-bound and cytoplasmic components whichmay then be reflected in the change in spectrum shape (Montán andStromblad, 1987). It may be resonable to conclude that the largedifferences in tissue autofluorescence between normal and malignant lesionsare unique for the in vivo situation.No increase in the red fluorescence in the bronchial tumour tissue(in vivo or in excised human tissue) was observed to suggest the presenceof endogenous porphyrin. This is in keeping with a recent report by Harrisand Werkhaven (Harris and Werhaven, 1987). However, in animal studiesusing the hamster cheek pouch model (Lam et al., l992a, Kluftinger et al.,1992), strong red fluorescence at around 640 run was observed, in dysplasticlesions, carcinoma in situ and small non-ulcerated invasive tumours.Again, a significant decrease in green autofluoresence was observed in thepre-cancerous and cancerous tissues compared to normal (figure 27). Theincrease in the red fluorescence for the carcinoma in situ lesion seen in105figure 27 is attributed to the prescence of endogeneous porphyrin. Detailsof the hamster cheek pouch model are described in appendix 7.2.Differences in tissue autofluorescence is generally considered tooweak to be quantified or are insensitive to current detection devicesespecially for small lesions. This belief had led many investigators toassume that an exogenous fluorescent tumour localizing drug is necessaryfor the detection of cancers. As discussed earlier, Hpd or Photofrin hadbeen most extensively used in the detection of early lung cancer. Becausethe fluorescence yield of Photofrin or Hpd is often small for thin earlylung cancer (less than 2%), background noise can adversely affect thedetectability of such low fluorescent light. In order to enhance thecontrast, one approach is to use the background fluorescence(autofluorescence) as the denominator for ratioing or subtraction whenusing Hpd or Photofrin. The idea had been so deep-rooted that fluorescencedetection has been generally equated with the use of fluorescent drugs.Thus almost all the devices that had been developed had concentrated onenhancing the fluorescence detection in the 600 nm to 690 nm spectralregion (Photofrin or Hpd maximum fluorescence peaks). The autofluorescencein the green region of the spectrum had been assumed to be similar fornormal and malignant tissues. The red Hpd or Photofrin fluorescenceintensity is then ratioed against the green autofluorescence normalizedaround 560 nm (Wagniere et al., 1990, Profio et al., 1984). Alternatively,the autofluorescence intensity was subtracted with a net result in the redfluorescence of Photofrin (Andersson-Engels et al., 1991, Mang et al.,1991, Hirano et al., 1989, Baumgartner et al., 1987, Montán et al., 1985,Profio et al., 1985).106Figure 24 In vivo fluorescence spectra from a normal area and an areawith carcinoma in situ lesion without and 24 hour afterintravenous injection of 1 mg/kg Photofrin. The slightincrease in the red fluorescence is very small compared to thesignificant difference in the green autofluorescence betweenthe normal and tumour area. (a.u. = arbitrary units)4‘—, 30)2C.)0)C.)a)0I0450 480 510 540 570 600 630 660 690 720Wavelength (urn)107Figure 25 Effects of freezing and thawing on the autofluorescence of anormal bronchus fragment. With time, the autofluorencesdecreases. (a.u. arbitrary units).16151413121o4)04)C.)4)4543210Fluorescence Intensity as a Function of Wavelength(Excitation: 442 nm)420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)frozenhourshour108Figure 26 Effect of thawing and freezing on the autofluorescence of anexcised tumour from the same patient as shown in figure 25.With time the autofluorescence intensity of increases. Thereis a significant decrease in autofluorescence in the greenregion compared to the normal fragment (figure 25). With time,this difference is less pronounced. (a.u. = arbitrary units).Fluorescence Intensity as a Function of Wavelength(Excitation: 442 nm)1615141312C.)UC.)U43210420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)hourshourfrozen109Figure 27 Autofluorescence spectra from a normal site and an area withcarcinoma in situ of a Syrian hamster cheek pouch. A Heliumcadmium laser at 442 nm was used for excitation. (a.u. =arbitrary units). A significant decrease in the greenautofluorescence was observed in the cancerous tissues comparedto normal. The increase in red fluorescence intensity (peak ataround 640 nm) for cancerous tissues is due to the prescence ofendogenous porphyrin.Fluorescence Intensity as a Function of Wavelength(Excited at 442 nm)65Normal1 Carcinoma in situ0420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)110A study of 17 patients with bronchial tumours investigated by thegroup in Lund, Sweden indicated that the use of autofluoresence alone todelineate tumour regions of the bronchus is probably not possible(Andersson-Engels, 1989). Although a difference between theautofluorescence emission of a normal and tumour sites were observed byAnthony and co-worker, the conclusion from this study was that Photofrinwill likely be needed to enhance the detection (Anthony et al., 1989).In contrast to these studies, we have shown here that the differencesin autofluorescence between normal and abnormal tissue is very significantand that these differences are sufficient to develop a fluorescence imagingapparatus contrary to previous belief.5.3 Mechanism for Difference in Tissue AutofluorescenceThe basis for the difference in tissue autofluorescence betweennormal, pre-cancerous and cancerous tissues is not known. An understandingof the underlying principles or cause of the difference in the observedfluorescence will likely enhance its application for detection of earlylung cancers.The fluorescence emission spectra of many fluorophors can beinfluenced by a variety of chemical and physical processes such as solventeffects, solvent relaxations, and quenching (Lakowicz, 1983). Fluorescencequenching refers to any process which decreases the fluorescence intensityof a particular fluorophore. Complex formation (static quenching), energytransfer and collision between fluorophore and quencher (dynamic quenching)can result in quenching. For dynamic quenching, i.e collisional quenching,111the quencher is diffused and in contact with the fluorophores while it isin its excited state, which upon contact subsequently returns to the groundstate without emissions of a photon. In static quenching, the fluorophoreand quencher form a complex which is non-fluorescent. Examples ofquenchers include molecular oxygen, purines and pyrimidines, hydrogenperoxide, and methionine (references in Lakowicz, 1983, Pesce et al., 1971,Udenfriend, 1962). In addition, optical properties of the sample (tissue)such as turbidity and inner filter effect may result in the quenching ofthe fluorescence signals (Lakowicz, 1983, Udenfriend, 1962). Thedistribution of the excitation light in the tissue, the fluorescencequantum yield, the path length of the fluorescence in tissue (the pathlength is determined by the excitation/collection geometry, the scatteringproperties of the tissue), and the tissue absorption coefficient for thefluorescence have also been suggested to influence the fluorescenceintensity detected at a tissue surface (Jacques, 1990, Keijzer et al.,1989b, Richards-Kortum et al., 1989a, Richards-Kortum et al., 1989b). Theoverall intensity and spectral distribution of the measured fluorescencedepends on the excitation/collection geometries of the detector system. Ingeneral, fluorescent light will undergo both scattering and reabsorptionwhen returning to the surface (Jacques, 1990, Keijzer et al., 1989b,Richards-Kortuin et al., 1989a).Several hypotheses have been proposed to explain the observeddifferences in autofluorescence:i) Proliferations and transformations of cells which result in anincrease in cell layers;ii) Different scattering properties cause by cytological changes;112iii) Changes in the macromolecular composition of the extracellular matrix(such as a change in the ion concentration, e.g. calcium ions whichact as a quencher (reference in Lakowicz, 1983), basement membraneand connective tissues (such as collagen and elastin);iv) Differences in the redox potential which involve co-enzymes and cofactors. Different levels of these co-enzymes and co-factors such asNADH (reduced form nicotinamide adenine dinicleotide), NAD+ (oxidizedform of nicotinamide adenine dinucleotide), riboflavin, FMN (flavinmononucleotide) and FAD (flavin adenine dinucleotide) will result inchanges in the fluorescent intensity;v) Pathological tissues may contain specific chemical compounds(quenchers) that may alter or mask the normal autofluorescence;vi) Decreased amounts of fluorophores in the tumour cells compared tonormal epithelial cells;vii) High concentration of absorbing species (not resulting influorescence) in tumour cells, which may attenuate the excitationlight reaching the fluorophores;The decrease in fluorescence intensity in cancerous tissues isprobably due to a combination of factors. In the literature reports can befound suggesting that there are differences in the level of riboflavin andnicotinamide dinucleotide between normal and tumour tissues (Rivlin, 1973,Pollack et al., 1942). Decreased levels of riboflavin had been found incancer patients (Rivlin, 1973). The changes of the oxidation-reductionequilibrium for NADHNAD+ is known to be associated with pre-cancerous andcancerous state. Following murine sarcoma virus transformation in rat113kidney fibroblast, the absolute concentration of NAD+ and NADH was found tobe decreased 2 to 3 times (reference in Richard-Kortuni et al., 1991). NAD+is non-fluorescent while NADH is highly fluorescent (Lakowickz, 1983,Udenfriend, 1962). The fluorescent group for NADH is the reducednicotinamide ring and its fluorescence is partially quenched by collisionswith the adenine moiety (Lakowickz, 1983, Udenfriend, 1962). When excitedat 337 rim, the emission spectrum of NADH shows a maximum at around 470 run(Andersson-Engels, 1991), Based on this observation it had been suggestedby other investigators that the decreased autofluorescence in malignanttissue may be due to a lesser content of NADH (Schomacker et al., 1992,Rava et al., 1991, Andersson-Engels et al., 1991, Cothren et al., 1990).In addition, hypoxic areas in tumours have been demonstrated to correlatewith decreased tissue autofluorescence signal at 450 rim following nearultra-violet excitation (Welsh et al., 1977, Gosalvez et al., 1972a,Gosalvez et al., l972b).Riboflavin, flavin mononucleotide (FMN), and flavin adeninedinucleotide (FAD) absorb light in the visible range at about 450 rim andemit fluorescence at around 515 rim (Ghisla et al., 1975). The oxidizedform of flavin has an absorption band centered at 450 rim and a strongfluorescence emission maximum at around 520 nm. In contrast, free, reducedflavins are non-fluorescent. Flavin fluorescence is quenched both bystatic and dynamic processes. Like NADH, flavin can be dynamicallyquenched by adenine. FAD also forms stacked complexes in which the flavinfluorescence is quenched by adenine (static quenching) (Lakowicz, 1983,Udenfriend, 1962). The concentration of riboflavin in tumour tissue is lowin comparison to that of normal tissues (Pollack et al., 1942), which is114consistent with the view that cancer tissues have a deficient aerobicoxidation system (Rivlin, 1977). The equilibrium of the redox state mayindirectly account for the differences in the autofluorescence in the greenregion of the spectrum. Pyridoxal 5’-phosphate, structural proteinscollagen and elastin are also known to fluoresce with maxima at around 520rim when excited by visible light in the range of 410 run to 450 rim. Therehave been reports of decreased levels of serum pyridoxal 5’-phosphate incancer patients (reference in Richard-Kortum et al., 1991). The level ofthese compounds has been implicated for the basis of the observeddifferences in autofluorescence in bladder, colon, and gastrointestinaltissues (Rava et al., 1991, Richard-Kortum et al., 1991, Cothren et al.,1990). The composition of the in vivo interstitial environment of tumourtissue is significantly different from those of most normal tissues, Thetumour interstitial compartment is characterized by a larger interstitialspace, high collagen concentration, low proteoglycan and hyaluronateconcentrations compared to normal tissues (Jam, 1987). These factors mayalso influence the autofluorescence properties.Several investigators have suggested that the fluorescence signalsdetected at the tissue surface are complicated by the interplay ofscattering with fluorescence and absorption, in particular, reabsorption byoxyhemoglobin (Jacques, 1990, Keijzer et al., 1989b, Richards-Kortum etal., 1989a). The excitation and collection geometries of the detectorsystem had also been implicated to influence the spectra shape and overallintensity of the fluorescence spectra. However, when the specificgeometries and parameters relevant to fluorescence bronchoscopy (asdescribed in this work) are taken into account, these effects are expected115to be of minimal significance. All the emitted fluorescence was collectedwithin the illuminated area. In terms of blood reabsorption, carcinoma Insitu and pre-cancerous lesions are not known to have increased bloodvessels. Furthermore, an increase in blood (hemoglobin) reabsorption wouldhave resulted in changes in spectral peaks or valleys in theautofluorescence spectra. However, this was not observed in both thespectra from freshly excised human tissues and in vivo. A huge differencewas observed in the autofluorescence intensity between normal and abnormaltissues.5.3.1 Preliminary DataIt is outside of the scope of the thesis to elucidate the exactmechanism(s) of reduced tissue autofluorescence in malignant tissues.Nevertheless, since this phenomenon is so intriguing, attempts were made totest some of the above hypotheses. For example, the fluorescence intensityspectra for FAD, FMN, riboflavin were measured and are shown in figure 28.Collagen IV and basement membrane were also measured and their spectra areshown in figures 29 and 30 respectively. Essentially most, if not all, ofthese spectra show similarities to the in vivo autofluorescence spectra. Afluorescence image of basement membrane material also shows thecharacteristics strong green autofluorscence using the LIFE device (figure30). It would be simple to suggest that observed differences inautofluorescence are due to these fluorophores. However, so many of theseobservations are indirect and the relative amounts of these fluorophoresare yet to be established. Extraction procedures to determine the relativeamounts of these fluorophores in normal and tumour tissue is very difficult116and requires difference skills than that of a biophysicist. In addition,much of the spectroscopic properties of these fluorophores are affected bythe local conditions, for example, concentration, pH, and temperature.This, too, must be carefully worked out before arriving to meaningfulconclusions.The results in figure 31 suggest that the major source of theautofluorescence comes from the sub-epithelial layer. Details of theseexperimental procedures are described in Appendix 7.2. An intriguingphenomenon was observed following removal of the epithelium (only one celllayer), which caused a 40% increase in fluorescence intensity. Afterremoving the submucosa there was a further increase in fluorescenceintensity by another 50% (Hung et al., 1991). Based on these results thedecrease in green autofluorescence in the pre-cancerous and canceroustissues may be interpreted as potentially due to i) a decrease influorophores in the sub-epithelial layers of the tumour tissue, or ii)increase in non fluorescent absorbing species in the mucoscal layers of theabnormal tissues, iii) the excitation light is not reaching the subepithelial layer due to a different in optical properties of the precancerous and cancerous tissues.Preliminary data, using a fluorescent automated cell sorter (FAGS)under excitation wavelength at 457.9 run on the following cell cultures werestudied: A549 human lung adenocarcinoma cells, CGL-210 human normal lungdiploid cells, HTB 5B human lung carcinoma cells and normal skin epithelialcells from a normal volunteer. The results suggested a decrease inautofluorescence in the tumour cells relative to the normal cells. Normalcells have about 30% more green fluorescence than the tumour cell lines117while the red autofluorescence for tumour cells was significantly higherthan the normal cells. However the slight difference in intensity observedhere still could not account for the huge differences in autofluoresceneintensity seen in vivo.Optical properties measurements using an integrating sphere were alsomade on human bronchial fragments and on the hamster cheek pouch. Detailsof this experiment are described in appendix 7.2. The data are presentedin Table V. Tumour tissues were found to be less scattering but moreabsorbing than the normal tissues.Further investigations to confirm these observations are merited. Itis clear that to elucidate the mechanisms for the decrease in the greenautofluorescence will be a major undertaking involving other disciplinessuch as biochemistry, molecular biology in addition to physics.5.3.2 Differences in Autofluorescence in Other TissueThe decrease in fluorescence intensity especially in the green regionof the spectrum is not unique for dysplasia and cancerous lesions of thebronchial tissues; it has also been observed in other epithelial tunioursand dysplasia when spectroscopic measurements were made in vivo.Gastrointestinal, bladder and colonic tissues have been shown to exhibit alower autofluorescence in tumour (Schomacker et al., 1992, Richard-Kortumet al., 1991, Andersson-Engels et al., 1991, Rava et al., 1991). Thehamster cheek pouch model of oral squamous cell carcinoma also exhibitedcharacteristic decreased green autofluorescence as well as endogenousporphyrin (Lam et al., 1992a, Kluftiriger et al., 1992).118Figure 28 Fluorescence spectra of FAD, FMN and riboflavin excited using a442 run Helium-cadmium laser. The maximum peak for FAD and FMNis at 542 nm, while for riboflavin it is red shifted to 556 run.(a.u. = arbitrary units).Fluorescence Intensity as a Function of Wavelength(Excitation: 442 nm)16151413121oUU43210420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)119Figure 29 Fluorescence spectrum of collagen IV excited at 442 rim using aHelium-cadmium laser. (a.u. = arbitrary units).Fluorescence Intensity as a Function of Wavelength(Excitation: 442 nm)1615141312zc!1OC.)420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)Collagen IV120Figure 30 Fluorescence spectrum (au = arbitrary units) and fluorescenceimage of basement membrane A Heliuxncadxnium laser at 442 nmwas used for the excitation.1817161514,1 3l2c)H4321Fluorescence Intensity as a Function of Wavelength(Excitation: 442 nm)Basement membrane0450 480 510 540 570 600 630 660 690 720 750Wavelength (nm)121Figure 31 Autofluorescence spectra of a freshly excised human bronchus.With the epithelium removed (one cell layer) there is anincrease of fluorescence intensity and with removal of the subepithelial layer, the cartilage shows a further increase inintensity compared to the intact bronchus. (a.u. = arbitraryunits).1817161514,1 312C)C)I!43210Fluorescence Intensity as a Function of Wavelength(Excited at 442 nm)420 450 480 510 540 570 600 630 660 690 720Wavelength (nm)cartilageepithelium stripped offintact bronchus122Table V Summary of the optical properties of freshly excised humanbronchial fragments and Syrian hamster cheek pouches.Absorption ScatteringTissue Coefficient Coefficient(mm) (mm)BRONCHIAL FRAGMENTS(Human)normal (6) 0.056 ± 0.014 4.38 ± 1.91tumour (6) 0.38 ± 0.032 0.83 ± 0.30HAMSTERnormal (50) 1.25 ± 0.44 3.38 ± 1.03inflammation (18) 1.60 ± 1.03 5.62 ± 0.99hyperplasia (3) 3.32 ± 0.74 4.68 ± 1.46dysplasia (2) 2.58 ± 2.95 4.46 ± 1.68papilloma (1) 0.067 1.64ulceration (3) 2.50 ± 1.11 5.43 ± 2.441235.4 Development of Lung Imaging Fluorescence EndoscopeAlthough the mechanism responsible for the dramatic decrease influorescence intensity of pre-nialignant and malignant tissues is not yetelucidated, the fact that these differences exist can be exploited todesign and develop a fluorescence imaging device for detection of dysplasiaand carcinoma in situ lesion during standard bronchoscopic examination(Palcic et al., 1991).Stepwise discrimination function analysis of 284 spectra to generateand test discriminant function, demonstrated that the optimal wavelengthsat which the differences can best be exploited is in the green regionbetween 525 rim to 570 rim and in the red region between 600 rim to 660 rim.Increasing the band widths of both the green and red regions was shown toimprove the discrimination. These studies suggest that several algorithmscould be applied for discriminating normal tissue from malignant tissues.Some of these algorithms (not part of the thesis) have now been applied toa pre-prototype lung fluorescence imaging endoscopic system (LIFE) to imageearly lung cancer under bronchoscopic examinations. The system design wasconceptually developed by the scientific staff at the Cancer ImagingSection of the British Columbia Cancer Agency (BCCA) (Palcic et al., 1991,Lam et al., l990b, Hung et al., 1990) and is not part of this thesis. Atthe present time, the system is developed and tested in a joint projectbetween Cancer Imaging, BCCA and Xillix Technologies Corporation,Vancouver, British Columbia who is responsible for commercialization ofthis device.Analyses on a database of 328 biopsy confirmed sites from 53 patientsand 41 volunteers, suggests that fluorescence imaging can detect dysplasia124and carcinoma in situ with a sensitivity of 72.5% and a specificity of 94%compared with white light bronchoscopy (sensitivity of only 48.4%,specificity of 94%) (Lam et aZ., 1992b).As discussed in chapter 2, it is difficult to detect and preciselylocalize pre-cancerous and in situ carcinoma lesions by conventionalbronchoscopy. These early lesions are often a few cell layers thick (0.2mm to 1 mm) and on average less than 0.8 cm in surface diameter (Woolner etal., 1984). These lesions may not produce enough changes on the bronchialsurface to allow detection by white light bronchoscopy. In some cases ofin situ carcinoma, subtle changes exist consisting of an increase inredness, granularity or slight thickening of the mucosa. Unfortunately,these changes can be associated with inflammatory airways diseases makingit very difficult for diagnosis. In the study by the Mayo Clinic, in situcarcinoma were visible bronchoscopically in less than 30% of the cases. A50% improvement in sensitivity using the fluorescence imaging systemcompared to white light bronchoscopy suggested that fluorescence imaging,therefore, may be an important adjunct to conventional white-lightexamination to improve the ability to diagnose and stage lung cancer moreaccurately.5.5 Directions for Future StudiesThe data collected from this thesis demonstrate that, using laserinduced autofluorescence, sufficient differences exist between normal andearly bronchial lesions that could clearly delineate pre-cancerous,cancerous lesions from the normal tissues. Although detection of pre125cancerous and cancerous lesions can be achieved using a non-skinphotosensitizing dose of photofrin, the major mechanism of tumour detectionusing ratio fluorometry and Photofrin was due to the significant decreasein autofluorecence in pre-cancerous and cancerous tissues. To a muchlesser extent, preferential retention of Photofrin in tumours also plays arole. Although the addition of this drug did not prove to be clinicallyuseful, other fluorescent tumour localizing drugs that do not cause skinphotosensitivity may be useful. The sensitivity of detection using suchexogenous fluorescent drugs must be compared to that using tissueauto fluorescence alone.While the data collected in this thesis cannot address definitivelywhich mechanism(s) is responsible for the dramatic decrease in fluorescenceintensity of pre-malignant and malignant tissues, some experimentalevidence suggests that the strong fluorescence signal from the normalbronchial tissues derives from the sub-epithelial layer. Collagen, elastinand flavoproteins may play an important role. Further work to elucidateand understand the underlying principles for the decrease inautofluorescence is merited.The improvement seen in using the fluorescence imaging system todelineate and precisely localized the extent of pre-inalignant lesions inindividual at risk of developing lung cancer has opened up newpossibilities in the study of the natural history of lung cancer as well asproviding means to establish more precise end-points in investigations ofthe efficacy of chemoprevention drugs. To be able to identify patientswith pre-cancerous and early lung lesions, has made it possible todetermine if niorphoinetric measurements of ostensibly normal epithelial126cells, in the region of pre-malignant or malignant lesions of lung cancerpatients, can reveal the presence of these lesions by malignancy associatedchanges (MACs) (Nieburgs, 1968). If MACs can be shown to exist in lungcancer patients as in the case of cancer of the uterine cervix (Bibbo etal., 1989, Burger et al., 1981, Wied et al., 1980), colon (Montag et al.,1991) and other tissues, this may have potential applications in improvingthe accuracy of sputum cytology examination to identify high riskpopulations for lung cancer. Fluorescence bronchoscopy can then be used tolocalize the source of abnormal sputum cells. The ability to detect andtreat pre-cancerous lesions and early lung cancer will improve thetraditionally poor prognosis of lung cancer.127Chapter 6: REFERENCESAlfano, R.R., Yao, S.S. Human Teeth With and Without Dental Caries Studiedby Visible Luminescent Spectroscopy. Journal of Dental Research 60:120-122, 1981.Alfano, R.R., Tata, D.B., Cordero, J., Tomashefsky, P., Longo, F.W., andAlfano, M.A. Laser Induced Fluorescence Spectroscopy from NativeCancerous and Normal Tissue. IEEE Journal of Quantum Electronics 20(12): 1507-1511, 1984.Alfano, R.R., Tang, G.C., Pradhan, A., Lam, W., Choy, D.S.J., and Opher, E.Fluorescence Spectra From Cancerous and Normal Human Breast and LungTissues. IEEE Journal of Quantum Electronics 23(10): 1806-1811, 1987.Alfano, R.R., Tang, G.C., Pradhan, A., Bleich, M, Choy, D.S.J., and Opher,E. Steady State and Time-Resolved Laser Fluorescence from Normal andTumour Lung and Breast Tissues. Journal of Tumour Marker Oncology3(2): 165-174, 1988.American Cancer Society. Cancer Facts and Figures - 1991. American CancerSociety, Atlanta, Georgia, 1991.Anderson, R.R., and Parrish, J.A. Optics Properties of Human Skin. In TheScience of Photomedicine, Regan, J.D. and Parrish, J.A. (eds), PlenumPress, New York and London, pp 147-194, 1982.Andersson, P.S., Montán, S., and Svanberg, S. Multispectral System forMedical Fluorescence Imaging. IEEE Journal of Quantum Electronics33(10): 1798-1805, 1987a.Andersson, P.S., Kjellén, Montán, S., Svanberg, K., and Svanberg, S.Autofluorescence of Various Rodent Tissues and Human Skin TumourSamples. Lasers in Medical Science 2:41-49, 1987b.Andersson-Engels, S. Laser-Induced Fluorescence for Medical Diagnosis.Lund Reports on Atomic Physics LRAP-108, Paper XIII, 1989.Andersson-Engels, S., Baert, L., D’Hallewin, M.A., Johansson, J., Stenrani,U., Svanberg, K., and Svanberg, S. Fluorescence Characteristics ofAtherosclerotic Plaque and Malignant Tumors. SPIE 1426 Optical Methodsfor Tumor Treatment and Early Diagnosis: Mechanisms and Techniques: 31-43, 1991.Andreoni, A., and Cubeddu, R. Photophysical Properties of Photofrin II inDifferent Solvents. Chem Phys Letters 108: 141-144, 1984.Anthony, D.J., Profio, A.E., and Balchum, O.J. Fluorescence Spectra inLung With Porphyrin Injection. Photochemistyr and Photobiology 49(5):583-586, 1989.128Aubin, J.E. Autofluorescence of Viable Cultured Mammalian Cells. Journalof Histochemistry and Cytochemsitry 27(1): 36-43, 1979.Auerbach, 0., Saccomanno, G., Kuschner, M., Brown, R.D., and Garfinkel, L.Histologic Findings in Tracheobronchial Tree of Uranium Miners and Non-Miners with Lung Cancer. Cancer 42: 483-489, 1978.Auler, H., and Banzer, G. Untersuchungen uber die Rolle der Porophine beiGeschwulstranken Menschen und Tieren. Ztschr. F. Krebsforsch 53: 65-68, 1942.Bailor J. Early Lung Cancer Cooperative Study Group: Early Lung CancerDetection. Summary and Conclusions. American Review of RespiratoryDisease 130: 565-570, 1984.Balchum, 0.J. , Doiron, D.R., Profio, A.E., and Huth, G.C. FluorescenceBronchoscopy for Localizing Early Bronchial Cancer and Carcinoma inSitu. Recent Results in Cancer Research 82: 97-120, 1982.Balchum, 0.J., Profio, A.E., and Razum, N. Ratioing Fluorometer Probe forLocalizing Carcinoma in situ in Bronchi of the Lung. Photochemistryand Photobiology 46(5): 887-891, 1987.Balchum, 0.J., Profio, A.E., and Lydic, 0. Detection and Treatment ofEarly Lung Cancer. International Photodynamic Association 3rd BiennialMeetings Abstracts: Vlll/7, pp 10, 1990.Banga, I., and Bihari-Varga, M. Investigations of Free and Elastin-BoundFluorescent Substances Present in the Atherosclerotic Lipid andCalcium-Plaques. Connective Tissue Research 2: 237-241, 1974.Baumgartner, R., Fisslinger, H., Jocham, D., Lenz, H., Ruprecht, L., Stepp,H., and Unsold, E. A Fluorescence Imaging Device For EndoscopicDetection of Early Stage Cancer - Instrumental and ExperimentalStudies. Photochemistry and Photobiology 46(5): 759-763, 1987.Becci, P., McDowell, E.M., and trump, B.F. The Respiratory Epithelium.VI. Histogenesis of Lung Tumors Induced by Benzo[a]pyrene-Ferric Oxidein Hamster. Journal of National Cancer Institute 61(2): 607-618, 1978.Benson, R.C., Meyer, R.A., Zaruba, M.E., and McKhann, G.M. CellularAutofluorescence - Is It Due to Flavins?. Journal of Histochemistryand Cytochemistry 27(1): 44-48. 1979.Berlin, N.I., Buncher, C.R., Fontana, R.S., Frost, J.K., and Melanied, M.R.The National Cancer Institute Cooperative Early Lung Cancer DetectionProgram. Early Lung Cancer Detection: Introduction. American Reviewof Respiratory Disease 130: 545-549, 1984.Bibbo, M., Montag, A.G., Lerma-Puertas, E., Dytch, H.E., Leelakusolvong,S., Bartels, P.H. Karyometric Marker Features in Tissue Adjacent to129Invasive Cervical Carcinomas. Analytical and Quantitative Cytology andHistology 11(4): 281-285, 1989.Blankenhorn, D.H., and Braustein, H. Cartenoids in Man. III. TheMicroscopic Pattern in Fluorescence in Atheromas and Its Relation toTheir Growth. Journal of Clinical Investigation 37: 160-165, 1958.Bohr, V.A., Phillips, D.H., and Hanawalt, P.C. Heterogenous DNA Damage andRepair in the Mammalian Cenome. Cancer Research 47: 6426-6436, 1987.Bolin, F.P., Preuss, L.E., Taylor, R.C., and Ference, R.J. RefractiveIndex of some Mammalian Tissues Using a Fiber Optic Cladding Method.Applied Optics 28(12): 2297-2300, 1989.Boulnois, J-L. Photophysical Process in Laser-Tissue Interactions. In:Primer on Laser Angioplasty, Ginsburg, R., and White, J.C. (eds).,Futura Publishing, New York : 29-89, 1989.Brookfield, R.L. Time-Resolved Fluorescence Spectroscopy ofHematoporphyrin. In Primary Photoprocesses in Biology and Mmedicine,Bensasson, R.V., Jon, G., Land, E.J., and Truscott, T.G. (eds), PlenumPress, New York, pp 329-333, 1984.Brown, C.C., and Kessler, L.G. Projection of Lung Cancer Mortality in theUnited States: 1985-2025. Journal of the National Cancer Institute 80:43-51, 1988.Brown, S.B., Shillcock, M., and Jones, P. Equilibrium and Kinetic Studiesof the Aggregation of Porphyrins in Aqueous Solution. Biochem. J. 153:279-285, 1976.Burger, G., Jutting, U., and Rodenacker, K. Changes in Benign CellsPopulations in Cases of Cervical cancer and Its Precursors. Analyticaland Quantitative Cytology 3(4): 261-271, 1981.Byrne, C.J., Marshallsay, L.V., Ward, A.D. The Structure of the ActiveMaterials in Hematoporphyrin Derivative. Photochemistry andPhotobiology 46(5): 575-580, 1987.Caplan, R.M. Medical Uses of Wood’s Lamp. JAMA 202 (11): 123-126, 1967.Case, K.M. and Zweifel, P.F. Linear Transport Theory, Addison-Wesley,Reading, Massachusetts, 1969.Chance, B., Cohen, P., Jobsis, F., and Schoener, B. IntracellularOxidation-Reduction States In Vivo. Science 137: 499-508, 1962.Chance, B., and Schoener, B. Fluorometric Studies of flavin Components ofthe respiratory Chain. In Flavins and Flavoproteins, Slater, E.C.(ed), Elsevier Science Publishers, Amsterdam, 1966.130Chandrasekhar, S. Radiative Transfer, Dover Publications Inc., New York,1960.Chapman, J.D., Stobbe, C.C., Arnfield, M.R., Santus, R., Lee, J., andMcPhee, M.S. Oxygen Dependency of Tumor Cell Killing in vitro byLight-Activated Photofrin II. Radiation Research 126: 73-79, 1991.Cheong, W-F, Prahi, S.A., and Welch, A.J. A Review of the OpticalProperties of Biological Tissues. IEEE Journal of Quantum Electronics26 (12): 2166-2185, 1990.Chilton, A.B., Shultis, J.K., and Faw, R.E. Principles of RadiationShielding, Pentice-Hall, Englewood Cliffs, pp 310-365, 1984.Clarke, R.H., Isner, J.M., Gauthier, T., Nakagawa, K., Cerio, F., Hanlon,E., Gaffney, E., Rouse, E., and DeJesus, S. SpectroscopicCharacterization of Cardiovascular Tissue. Lasers in Surgery andMedicine 8:45-89, 1988.Cortese, D.A. and Kinsey, 3.11. Hematoporphyrin-Derivative Fluorescence forLung Cancer Localization. Seminars in Respiratory Medicine 3(1), 37-41, 1981.Cortese, D.A., Kinsey, J.H., Woolner, L.B., Sanderson, D.R., and Fontana,R.S. Hematoporphyrin Derivative in the Detection and Localization ofRadiologically Occult Lung Cancer. American Review of RespiratoryDiseases 126: 1087-1088, 1982.Cothren, R.M., Rchards-Kortum, R., Sivak, M.V., Fitzmaurice, M., Rava,R.P., Boyce, G.A., Doxtader, M., Blackman, R., Ivanc. T.B., Hayes,G.B., Feld, M.S., and Petras, R.E. Gastrointestinal Tissue Diagnosisby Laser-Induced Fluorescence spectroscopy at Endoscopy.Gastrointestinal Endoscopy 36(2): 105-111, 1990.Devesa, S.S., Blot, W.J., Fraumeni, J.F. Declining Lung Cancer Rates AmongYoung Men and Women in the United States: A Cohort Analysis. Journalof the National Cancer Institute 81(20): 1568-1571.Doll. R. and Peto, R. The Causes of cancer. In Quantitative Estimates ofAvoidable Risks of Cancer in The United States Today, Oxford UniversityPress, Oxford, 1982.Dougherty, T.J. Activated Dyes as Antitumor Agents. Journal of theNational Cancer Institute 52(4): 1333-1336, 1974.Dougherty, T.J., Potter, W.R., and Weishaupt, K.R. The Structure of theActive Component of Hematoporphyrin Derivative. In ProphyrinLocalization and Treatment of Tumors, Doiron, D.R., and Gomer, G.J.(eds), pp 301-314, 1984.Dougherty, T.J. Studies on the Structure of Porphyrins Contained inPhotofrin II. Photochemistry and Photobiology 46(5): 569, 1987.131Dougherty, T.J. Photodynamic Therapy - New Approaches. Seminars inSurgical Oncology 5: 6-16, 1989.Dupont, R.L. Marijunana Smoking: A National Epidemic. American LungAssociation Bulletin 66: 2-7, 1980.Fearon, E.R., and Vogelstein, B. A genetic Model for ColorectalTumorigenesis. Cell 61: 759-767, 1990.Fellner, M.J. Green Autofluorescence in Human Epidermal Cells. Archivesof Dermatology 112: 667-670, 1976.Figge, F.H.J., Jones, E.G., and Wolfe, G.F. Fluorescence Studies onCancer. III The Extraction and Identification of Porphyrins From theRed-Fluorescent Exudates on the Genitalia Women. Cancer Research 4(8):483-486, 1944.Figge, F.H.J., Weiland, G.S., and Manganiello, L.0.J. Cancer Detection andTherapy, Affinity of Neoplastic, Embryonic, and Haematized Tissues forPorphyrin and Metalo Porphyrins. Proc Soc Exp Biol Med 68: 640-641,1948.Flock, S.T., Wilson, B.C., and Patterson, M.S. Total AttentuationCoefficients and Scattering Phase Functions of Tissues and PhantomMaterials at 633 nm. Medical Physics 14: 835-841, 1987.Forrest, J., and Lee, R.M.K.W. The Bronchial Wall: Intergrated Form andFunction. In The Lung Scientific Foundations, Volume 1, Crystal, R.G.,and West, J.B. (Editor-in-Chief), Raven Press, New York, pp 729-739,1991.Fontana, R.S., Sanderson, D.R., Woolner, L.B., Taylor, W.F., Miller, W.E.,Muhm, J.R., Bernatz, P.E., Payne, W.S., Pairolero, P.C., andBergstralh, E.J. Screening for Lung Cancer A Critique of the Mayo LungProject. Cancer 67: 1155-1164, 1991.Fontana, R.S., Sanderson, D.R., Woolner, L.B., Taylor, W.F., Miller, W.E.,and Muhm, J.R. Lung Cancer Screening: The Mayo Program. Journal ofOccupational Medicine 28(8):746-750, 1986.Frost, J.K., Erozan, Y.S., Gupta, P.K., and Carter, D. Cytoplathology. InAtlas of Early Lung Cancer National Cancer Institute, NationalInstitutes of Health, U.S. Department of Health and Human Services,Igaku-Shoin, Tokyo, pp 39-76, 1983.Gail, M.H. Multiple Markers For Lung Cancer Diagnosis. Journal of theNational Institute 80: 97-101, 1988.Gadzar, A.F., Muishine, J.L., and Kramer, B.S. Biological, Molecular, andClinical Markers for Diagnosis and Typing of Lung Cancer. In132Immunodiagnosis of Cancer, Herberman, R.B.,, and Mercer, D.W. (eds),Marcek Dekker, New York, pp 453-468, 1990.Ghisla, S., Massey, V., Lhoste, J-M., and Mayhew, S.G. FluorescenceCharacteristics of Reduced Flavins and Flavoproteins. In Reactivity ofFlavins, Yagi K. (Eds), University of Tokyo Press, Japan, pp 15-24,1975.Gmitro, A.F., Cutruzzola, F.W., Stetz, M.L., and Deckelbaum, L.I.Measurement Depth of Laser-Induced tissue Fluorescence With Applicationto Laser Angioplasty. Applied Optics 27(9): 1844-1849, 1988.Goldstein, H. Fundamental Aspects of Reactor Shielding, Addison-Wesley,Reading, Massachusetts, pp 159-166, 1959.Gosalvez, M., Thurman, R.G., Chance, B., and Reinhold, H. RegionalVariation in the Oxygenation of Mouse Mammary Tumours in vivodemonstared by Fluorescence of Pyridine Nucleotide. British Journal ofRadiology 45: 510-514, 1972a.Gosalvez, M., Thurman, R.G., Chance, B, and Reirthold, H.S. Indication ofHypoxic Areas in Tumours From in vivo NADH Fluorescence. EuropeanJournal of Cancer 8: 267-269, l972b.Gralla, R.J., and Einhorn, L.H. Treatment and Prevention of Small CellLung Cancer and Non-Small Cell Lung Cancer. Series number 150, RoyalSociety of Medicine Limited, New York, 1989.Gray, M.J., Lipson, R., van s. Maeck, J., Parker, L., and Romeyn, D. Useof Hematoporphyrin Derivative in Detection and Management of CerivalCancer. Am. J. Obst. and Gyrtec. 99(6): 766-771, 1967.Greenberg, S.D., Spjut, H.J., Estrada, R.G., Hunter, N.R., and Grenia, C.Morphometric Markers for the Evaluation of Pre-neoplastics Lesions inthe Lung Diagnostic Evaluation by High-Resolution Image Analysis ofAtypical Cells in Sputum Specimens. Analytical and QuantitiativeCytology and Histology 9(1): 49-54, 1987.Gregorie, H.B. Jr., Horger, E.D., Ward, J.L., Green, J.R., Richards, T.,Robertson, H.C., Jr., and Stevenson, T.B. Hematoporphyrin-DerivativeFluorescence in Malignant Neoplasms. Annals of Surgery 167(6): 820-828, 1968.Gryzbowski, S., Cop, P. Early Diagnosis of Carcinoma of Lung:Simultaneous Screening with Chest X-ray and Sputum Cytology. Cancer25: 113-120, 1970.Hanley, J.A., and Lippman-Hand, A. If nothing goes wrong is everything allright? Interpreting zero numerators. JAMA 249: 1743-1745, 1983.133Harris, C.C. Chemical and Physical Carcinogenesis: Advances andProspective for the 1990s. Cancer Research (Supplement) 51: 5023s-5044s, 1991.Harris, D.M., and Werkhaven, J. Endogenous Porphyrin Fluorescence inTumors. Lasers in Surgery and Medicine 7:467-462, 1987.Hayata, Y., Kato, H., Konaka, C., Ono, J., Matsushima, Y., Yoneyama, K.,Nishimiya, K. Fiberoptic Bronchoscopic Laser Photoradiation for TumourLocalization in Lung Cancer. Chest 82(1): 10-14, 1982.Hecht, E. Optics, Second Edition, Addison-Wesley Publishing, Reading,Massachusetts, 1987.Henderson, B.E., Ross, R.K., and Pike, M.C. Toward the Primary Preventionof Cancer. Science 254: 1131-1138, 1991.Henderson, B.W., and Dougherty, T.J. Studies on the Mechanisms of TumorDestruction by Photoradiation Therapy (PRT). In Porphyrin Localizationand Treatment of Tumours, Doiron, D.R., and Gomer, C.J. (eds)., Alan R.Liss, Inc. , New York, pp 601-612, 1984.Henderson, B.W., and Miller, A.C. Effects of scavengers of Reactive Oxygenand Radical Species on Cell Survival Following Photodynamic Treatmentin Vitro: Comparison to Ionizing Radiation. Radiation Research 108:196-205, 1986.Henyey, L.G., and Greenstein, J.L. Diffuse Radiation in the Galaxy.Astrophysical Journal 93: 70-83, 1941.Herly, L. Studies in Selective Differentiation of Tissues by Means ofFiltered Ultraviolet Light. Cancer Research 4(4): 227-231, 1944.Hirano, T., Ishizuka, M., Suzuki, K., Ishida, K., Suzuki, S., Miyaki, S.,Honma, A., Suzuki, M., Aizwa, K., Kato, H., and Hayata, Y.Photodynamic Cancer Diagnosis and Treatment System Consisting of PulseLasers and an endoscopic Spectro-Image Analyzer. Lasers in the LifeSciences 3(1): 1-18, 1989.Hjerpe, A. Early Detection of Lung cancer Using Sputum Cytoloyy. In NewFrontiers in Cytology Modern Aspects of Research and Practice,Coerttler, K., Feichter, G.E., and Witte, S. (eds), Springer-Varlag,Berlin, Heidelberg, pp 434-439, 1988.Hoyt, C.C., Richards-Kortum, R.R., Costello, B., Sacks, B.A., Kittrell, C.,Ratliff, N.B., Kramer, J.R., and Feld, M.S. Remote BiomedicalSpectroscopic Imaging of Human Artery Wall. Lasers in Surgery andMedicine 8: 1-9, 1988.Humphrey, P.A. The Role of Tumor Markers in the Early Detection of Cancer.Seminars in Surgical Oncology 5: 186-193, 1989.134Hung, J., Lam, S., Jaggi, B., Pon, A., and Palcic, B. Localization of LungCancer by Tissue Autofluorescence. Proceedings of the American Collegeof Chest Physicians 56th Annual Scientific Assembly, p, 108s, October,1990.Hung, J., Lam, S., Miller, R., and Palcic, B. Laser InducedAutofluorescence of Normal and Malignant Tissues. Lasers in Surgeryand Medicine 3(supplement): 57, 1991.IARC (WHO). Tomatis, L. (Editor-in-Chief), Aitio, A., Day, N.E.,Heseltine, E., Kaldor, J., Miller, A.B., Parkin, D.M., and Riboli, E.(Co-editors). Tobacco, In Cancer: Causes, Occurence and Control, WHOIARC Scientific Publications N° 100, Lyon, 1990.Ihde, D.C., and Minna, J.D. Non-Small Cell Lung Cancer Part I: Biology,Diagnosis, and Staging. Current Problems in Cancer, Volume XV (2),Haskell, C.M. (editor-in-chief), 1991.Innoue, S., Ito, M., Mizuno, S., Isobe, H., Miyamoto, H., Kawakami, Y., andOhnuma, T. Simultaneous Flow Cytometric Detection of Nuclear DNA andTumor-Associated Antigens in Lung Cancer. Analytical and quantitativeCytology and Histology 10(4): 243-250, 1988.Ishimaru A. Diffusion of Light in Turbid Material. Applied Optics 28(12):2210-2215, 1989.Jacques, S.L., and Prahl, S.C. Modelling Optical and Thermal Distributionsin tissue During Laser Irradiation. Laser in Surgery and Medicine 6:494-503, 1987.Jacques, S.L., Alter, C.A., and Prahl, S.A. Angular Dependence of HeNeLaser Light Scattering by Human Dermis. Lasers Life Science 1: 309-333, 1987.Jacques, S.L. Optical Engineering in Medicine: Design of Devices andDosimetry, SPIE SC78 Tutorial, 1990.Jam, R.K. Transport of Molecules in the Tumor Interstiitium: A Review.Cancer Research 47: 3039-3051, 1987.Jett, J.R., Cortese, D.A., and Fontana, R.S. Lung Cancer: Current Conceptsand Prospects. CA: A Cancer Journal for Clinicians 33(2): 74-86, 1983.Kapadia, C.R., Cutruzzola, F.W., O’Brien, K.M., Stetz, M.L., Enriquez, R.,and Deckelbaum, L.I. Laser-Induced Fluorescence Spectroscopy of HumanColonic Mucosa: Detection of Adenomatous Transformation.Gastroenterology 99: 150-157, 1990.Kato, H., Konaka, C., Hayata, Y., Ono, J., limura, I., Matsushima, Y.,Tahara, M., Lei, J., Nasiell, M., and Auer, G. Lung CancerHistogenesis Following in vivo Bronchial Injections of 20-135methyicholanthrene in Dogs. Recent Results in Cancer Research 82,Springer, Berlin, pp 69-86, 1982.Kato, H., Aizawa, K., Ono, J., Konaka, C., Kawate, N., Yoneyama, K.,Kinoshita, K., Nishimiya, K., Sakai, H., Noguchi, M., Tomono, T.,Kawasaki, S., Tokuda, Y., and Hayata, Y. Clinical Measurement ofTumour Fluorescence Using a New Diagnostic System With HematoporphyrinDerivative, Laser Photoradiation, and a spectroscope. Lasers inSurgery and Medicine 4: 49-58, 1984.Kato, H., and Cortese, D.A. Early Detection of Lung Cancer by means ofHeniatoporphyrin Derivative Fluorescence and Laser Photoradiation.Clinics in Chest Medicine 6(2): 237-253, 1985.Keijzer, M., Jacques, S.L., Prahl, S.A., and Welch, A.J. LightDistributions in Artery Tissue: Monte Carlo Simulations for Finite-Diameter Laser Beams. Laser in Surgery and Medicine 9: 148-154, 1989a.Keijzer, M., Richards-Kortuin, R.R., Jacques, S.L., Feld, M.S. FluorescenceSpectroscopy of Turbid Media: Autofluorescence of Human Aorta. AppliedOptics 28(2O):4286-4292, l989b.Kinsey, J.H., Cortese, D.A., and Sanderson, D.R. Detection ofHematoporphyrin Fluorescence During Fiberoptic Bronchoscopy to LocalizeEarly Bronchogenic Carcinoma. Mayo Clinic Proceddings 53: 594-600,1978.Kinsey, J.H., and Cortese, D.A. Endoscopic System for Simultaneous VisualExamination and Electronic Detection of Fluorescence. Review ofScientific Instruments 51(10): 1403-1406, 1980.Kiuftinger, A.M., Davis, N.L., Quenville, N.F., Lam, S., Hung, J., Palcic,B. Detection of Squamous Cell Cancer and Pre-Cancerous Lesions byImaging of Tissue Autofluorescence in the Hamster Cheek Pouch.Surgical Oncology: In press, 1992.Kubelka, P. New Contributions to the Optics of Intensely Light ScatteringMaterials Part II. J. Opt. Soc. Am. 44: 330-335, 1954.Kubelka, P. New Contributions to the Optics of Intensely Light ScatteringMaterials Part I. J. Opt. Soc. Am. 38: 448-457, 1948.Kubik, A., Parkin, D.M., Khlat, M., Erban, J., Polak, J., and Adamec, M.Lack of Benefit From Semi-Annual Screening for Cancer of the Lung:Follow-Up Report of a Randomized Controlled Trial on a Population ofHigh-Risks Males in Czechoslovakia. International Journal of Cancer45: 26-33, 1990.Kyriazis, G.A., Balm, H., and Lipson, R.L. Heniatoporhyrmn-DerivativeFluorescence Test Colposcopy and Colpophotography in the Diagnosis ofAtypical metaplasia, Dysplasia, and Carcinoma in situ of the CervixUteri. Am. J. Obstet. Gynecol. 117(3): 375-380, 1973.136Lachenbruch, P.A. Discriminant Analysis, Hafner Press, New York, 1975.Lakowicz, J.R. Principles of Fluorescence Spectroscopy, Plenum Press, NewYork, 1983.Lam, S., Palcic, B., McLean, D., Hung, J., Korbelik, M., Profio, A.E.Detection of Low Dose Photofrin II . Chest 97: 333-337, 1990a.Lam, S., Palcic, B., Jaggi, B., Hung, J., Pon, A., and Profio, A.E. LungImaging Fluorescence Endoscope (LIFE): Applications of a ClinicalPrototype. Proceedings of the 3rd Biennial Meeting InternationalPhotodynnamic Association, p 31, l990b.Lam, S., Hung, J., and Palcic, B. Mechanism of Detection of Early LungCancer by Ratio Fluorometry. Lasers in Life Sciences, 1991.Lam, S., Kiuftinger, A.M., Hung, J., Davis, N.L., Quenville, N.F., andPalcic, B. Detection of Oral Squamous Cell Cancer and Pre-CancerousLesions by Fluorescence Imaging in a Hamster Cheek Pouch Model. SPIE,1992a.Lam, S., MacAulay, C., Hung, J., LeRiche, J., Prof io, A.E., and Palcic, B.Detection of Early Lung Cancer Using a Lung Imaging FluorescenceEndoscope (LIFE) Device. Journal of Thoracic Surgery: In Press, l992b.Lam, S., Hung, J.Y.C., Kennedy, S.M., LeRiche, J.C., Vedal, S., Nelems. B.,MacAulay, C.E., Palcic, B. Detection of Dysplasia and Carcinoma insitu by Ratio Flurometry. American Review of Respiratory Disease(Submitted), 1992c.Law, M.R. Genetic Predispostion to Lung Cancer. British Journal of Cancer61: 195-206, 1990.Leffell, D., Stetz, M.L., Milstone, L.M., and Deckelbaum, L.I. In VivoFluorescence of Human Skin, A Potential Marker of Photoaging. Archivesof Dermatology 124: 1514-1518, 1988.Leonard, J.R., and Beck, W.L. Hematoporhyrin Fluorescence: An Aid inDiagnosis of Malignant Neoplasms. Laryngoscope 81: 365-372, 1971.Lerner, J.M. Optics of Spectroscopy, Volume I and Volume II. SPIE OELASE’lO Tutorial SC71, 1990.Line, B.R., Armstrong, M.R., Hendrick, W.J., and Killanm, D.A.Radiographic Studies of Lung Cancer, In Thoracic Oncology, Roth, J.A.,Ruckdeschel, J.C., and Weisenburger, T.H. (eds), W.E. Saunders Company,Philadelphia, pp 108-109, 1989.Lipson, R.L., Baldes, E.J., and Olsen, E.M. The Use of a Derivative ofHematoporphyrin in Tumor Detection. Journal of National CancerInstitute 26: 1-12, 1961.137Link, C.J., Burt, R.K., and Bohr, V.A. Review: Gene-Specific Repair ofDNA Damage Induced by UV Irradiation and Cancer Chemotherapeutics.Cancer Cells 3(11): 427-436, 1991.Loeb, L.A., Ernster, V.L., Warner, K.E., Abbotts, J., and Laszlo, J.Smoking and Lung Cancer: An Overview. Cancer Research 44: 5940-5958,1984.Loeb, L.A. Mutator Phenotype May be required for Multistage Carcinogensis.Cancer Research 51: 3075-3079, 1991.Lucas, A., Radosavljevic, M.J., Lu, E., and Gaffney, E.J. Characterizationof Human Coronary Artery Atherosclerotic Plaque Fluorescence Emission.Canadian Journal of Cardiology 6(6): 219-228, 1990.McBride, M., Ma, B., and Band, P. Projected Cancer Frequency in BritishColumbia and Regions, 1986, 1991, 1996, 2001. Monograph No. 2,Division of Epidemiology, Biometry and Occupational Oncology, BCCA,1989.McDowell, E.M., and Trump, B.F. Histogenesis of preneoplastics andneoplastic lesions in Tracheobronchial Epitheliuni. In Survey andSynthesis of Pathology Research, Volume 2, Cruse, J.M. (ed), S. Karger,Basel, pp 235-279, 1983.McGinnis, C., and Mang, T.S. In Vivo Fluorescence Photometry For theDetection of Cellular Retention of Photofrin II. InternationalPhotodynamic Association 3rd Biennial Meetings Abstracts: XV/2, pp 18,1990.MacMahon, H., Ell, S.R., and Ferguson, M.K. Diagnostics Methods in LungCancer. In Lung Cancer: A Comprehensive Treatise, Bitran, J.D.,Golomb, H.M., Little, A.G., and Weichselbaum, R.R. (eds), Grune andStratton, Inc., Orlando, Florida, pp 87-112, 1988Mang, T.S., McGinnis, C., Crean, D.H., Khan, S., and Liebow, C.Fluorescence Detection of Tumors: Studies on the Early Diagnosis ofMicroscopic Lesions in Pre-Clinical and Clinical Studies. SPIE OpticalMethods for tumor Treatment and Early Diagnosis: Mechanisms andtechniques 1426: 97-110, 1991.Marchesini, R., Bertoni, A., Andreola, S., Melloni, E., and Sichirollo,A.E. Extinction and Absorption Coefficients and Scattering Phasefunctions of Human tissues in vitro. Applied Optics 28(12): 2318-2324,1989.Marsh, B.R., Sanderson, D.R., and Martini, N. Bronchoscopy. In Atlas ofEarly Lung Cancer National Cancer Institute, National Institute ofHealth and Human Services, Igaku-Shoin Ltd., Tokyo, pp 77-80, 1983.138Matthews, M.J., and Linnoila, R.I. Pathology of Lung Cancer - An Update.In Lung Cancer Comprehensive Treatise, Bitran, J.D., Golomb, H.M.,Little, A.G., and Weichselbaum, R.R. (eds), Grune and Stratton, Inc.,Orlando, Florida, pp 5-34, 1988.Melamed, M.R., Flehinger, B.J., Zanian, M.B., Heelan, R.T., Perchik, W.B.,and Martini N. Screening for Early Lung Cancer. Results of theMemorial Sloan-Kettering Study in New York. Chest 86: 44-63, 1984.Miller, W.E., Muhm, J.R., Stitik, F.P., and Woolner, L.B. Radiology andPathologic Correlations. In Atlas of Early Lung Cancer National CancerInstitute, National Institutes of Health, U.S. Department of Health andHuman Services, Igaku-Shoin, Tokyo, pp 215-269, 1983.Minna, J.D., Higgins, G.A., and Glatstein, E.J. Cancer of the Lung. InPrinciples and Pratice of Oncology, DeVita, V.T., Hellman, S.,Rosenberg, S. (eds), J.B. Lippincott, Philadephia, pp 507-599, 1989.Mitchell, J.B., Cook, J.A., and Russo, A. Biological Basis forPhototherapy. In Phototherapy of Cancer, Morstyn, C., and kaye, A.H.(eds), Harwood Academic Publishers GmbH, Switzerland, pp 1-22, 1990.Moan, J., and Sommer, S. Fluorescence and Absorption Properties ofComponents of Hematoporphyrin Derivative. Photochemistry andPhotobiophysics 3: 93-103, 1981.Moan, J., and Sommer, S. Oxygen Dependence of the Photosensitizing Effectof Hematoporphyrin Derivative in NHIK 3025 Cells. Cancer Research 45:1608-1610, 1985.Montán, S., Svanberg, K, and Svanberg, S. Multicolor Imaging and ConstrastEnhancement in Cancer-Tumour Localization Using Laser-InducedFluorescence in Hematoporphyrin-Derivative-Bearing Tissue. OpticsLetters 10(2): 56-58, 1985.Montán, S., and Strömbland, L-G. Spectral Characterization of Brain TumorsUtilizing Laser-Induced Fluorescence. Lasers in the Life Sciences1(4): 275-285, 1987.Montag, A.C., Bartels, P.H., Dytch, H.E., Lerma-Puertas, E., Michelassi,F., and Bibbo. Karyometric Features in Nuclei Near ColonicAdenocarcinoma. Analytical and Quantitative Cytology and Histology 11(4): 159-167, 1991.Moore, G.E. Fluorescein as an Agent in the Differentiation of Normal andMalignant Tissues. Science 106: 130-131, 1947.Muishine, J.L., Tockman, M.S., Smart, C.R. Considerations in theDevelopment of Lung Cancer Screening Tools. Journal of the NationalCancer Institute 81(12): 900-906, 1989.139Muishine, J.L., Treston, A.M., Scott, F.M., Avis, I., Boland, C., Pjelps,R., Kasprzyk, Y., and Cuttita, F. Lung Cancer: Rational Strategiesfor Early Detection and Intervention. Oncology 5(5): 25-33, 1991.Nieburgs, H.E. Recent Progress in the Interpretation of MalignancyAssociated Changes. Acta Cytologica 12(6): 445-453, 1968.Neuman, R., and Logan, M. The Determination of Collagen and Elastin inTissues. Journal of Biological Chemistry 186: 549-556, 1950.Oho, K., and Amemiya, R. Practical Fiberoptic Bronchoscopy, Igaku-Shoin,Tokyo, Second Edition, 1984.Palcic, B., Lam, S., Hung, J., MacAulay, C. Detection and Localization ofEarly Lung Cancer By Imaging Techniques. Chest 99: 742-743, 1991.Palcic, B. , and Lam, S. Lung Imaging Fluorescence Endoscope: A Device fordetection of Occult Lung Cancer. Medical Design and Material: June,1991.Parkin, D.M., Laara, E., and Muir, C.S. Estimates of the WorldwideFrequency of Sixteen Major Cancers. International Journal of Cancer41: 184-197, 1988.Parrish, J.A. New concepts in Therapeutic Photomedicine, Photochemistry,Optical Targeting and Therapeutic Window. Journal Invest. Dermatology77: 45-50, 1981.Parrish, J.A., and Wilson, B.C. Current and Future Trends in LaserMedicine. Photochemistry and Photobiolgy 53(6): 731-738, 1991.Parsa, P., Jacques, S.L., and Nishioka, N.S. Optical Properties of RatLiver between 350 and 2200 run. Applied Optics 28 (12): 2325-2330,1989.Pesce, A.J., Rosoén, C-G., and Pasby, T.L. Fluorescence Spectroscopy AnIntroduction for Biology and Medicine, Marcel Dekker, Inc., New York,1971.Philips, D., Hewer, A., Martin, C., Garner, R., and King, M. Correlationof DNA Adducts Levels in Human Lung Cigarette Smoking. Nature 336:790-792, 1988.Pienta, K.J., Partin, A.W., and Coffey, D.S. Cancer as a Disease of DNAOrganization and Dynamic Cell Structure. Cancer Research 49: 2525-2532, 1989.Policad, A. Etudes sur les aspects offerts par des tumeur experimentalesexaminee a la lumiere de woods. Compt Rend Soc Biol 91: 1423-1424,1924.140Pollack, M.A., Taylor, A., Taylor, J., and Williams, R.J. B Vitamins inCancerous Tissues I: Riboflavin. Cancer Research: 2(11): 739-743,1942.Potter, W.R., and Mang, T.S. Photofrin II Levels By in vivo FluorescencePhotometry. In Porphyrin Localization and Treatment of Tumours,Doiron, D.R., and Gomer, C.J. (eds), Alan R. Liss, New York, pp 177-186, 1984.Potter, W.R. PDT Dosimetry and Response. SPIE 1065,1990Prahl, S.A., Keijzer, M., Jacques, S.L., and Welch, A.J. A Monte CarloModel of Light Propagation in Tissue. In Dosimetry of Laser Radiationin Medicine and Biology,SPIE Optical Engineering Press, Washington, VolIS-5, pp 102-111, 1989.Profio, A.E. Radiation Shielding and Dosimetry, Wiley, New York, 167-218,1979.Profio, A.E., Doiron, D.R., and King, E.G. Laser Fluorescence Bronchoscopefor Localization of Occult Lung Tuniours. Medical Physics 6(6): 523-525, 1979.Profio, A.E., Doiron, D.R., and Sarnaik, J. Fluorometer for EndoscopicDiagnosis of Tuniours. Medical Physics 11(4): 516-520, 1984.Profio, A.E. Laser Excited Fluorescence of Hematoporphyrin Derivative forDiagnosis of Cancer. IEEE Journal of Quantum Electronics 20(12): 1502-1507, 1984.Profio, A.E., Balchum, O.J., and Carstens, F. Digital BackgroundSubstraction for Fluorescence Imaging. Medical Physics 13(5): 717-721,1986.Prorok, P.C., Chamberlain, J., Day, N.E., Hakama, M., and Miller, A.B.UICC Workshop on the Evaluation of Screening Programmes for Cancer.International Journal Cancer 34: 1-4, 1984.Rall, D.P., Loo, T.L., Lane, M., and Kelly, M.G. Appearance andpersistence of Fluorescent Material in Tumor Tissue After TetracyclineAdministration. Journal of the National Cancer Institute 19(1): 79-85,1957.Rassmussen-Taxdal, D.S., Ward, G.E., and Figge, F.H. Fluorescence of HumanLymphatic and Cancer Tissues Following High Doses of IntravenousHematoporyphrin. Cancer 8(1): 78-81, 1955.Rava, R.P., Richard-Kortuni, R., Fitzmaurice, Cothren, R., Petras, R.,Sivak, M., Levin, H., and Feld, M. Early Detection of Dysplasia inColon and Bladder Tissue Using Laser Induced Fluorescence. SPIE 1426:68-78, 1991.141Razum, N., Baichum, O.J., Profio, A.E., and Carstens, F. SkinPhotosensitivity : Duration and Intensity Following IntravenousHematoporphyrin Derivatives, HpD and DHE. Photochemistry andPhotobiology 46: 925-928, 1987.Richards-Kortum, R., Rava, R.P., Petras, R.E., Fitzmaurice, M., Sivak, M.,and Feld, M.S. Spectroscopic Diagnosis of Colonic Dysplasia.Photochemistry and Photobiology 53(6): 777-786, 1991.Richards-Kortum, R., Metha, A., Hayes, C., Cothren, R., Kolubayev, T.,Kittrell, C., Ratliff, N.B., Kramer, J.R., and Feld, M.S. SpectraDiagnosis of Atherosclerosis Using an Optical Fiber Laser Catheter.American Heart Journal 118(2): 381-391, 1989a.Richards-Kortum, R., Rava, R.P., Cothren, R., Metha, A., Fitzmaurice, M.,Ratliff, N.B., Kramer, J.R., Kittrell, C., and Feld, M.S. A Model forExtraction of Diagnostics Information from Laser Induced FluorescenceSpectra of Human Artery Wall. Spectrochim Acta 45A(l): 87-91, 1989b.Rivlin, R.S. Riboflavin and Cancer: A Review. Cancer Research 33: 1977-1986, 1973.Rosenthall, D.L. Cytopathology of Pulmonary Disease. In Monographs inClinical Cytology, Volume 11, Wied, G.L. (ed), Karger, Basel, pp 24-41,86-92, 1988.Rotomskis, R., van de Meent, E.J., Aartsma, T.J., and Hoff A.J.Fluorescence Spectra of Hematoporphyrin and Hematoporphyrin-DiacetateAggregates in Buffer Solution. Journal of Photochemistry andPhotobiology B, Biology 3: 369-377, 1989.Sanderson, D.R., Fontana, R.S., Lipson, R.L., and Blades, E.J.Hematoporhyrin as a Diagnostic Tool A Preliminary Report of NewTechnique. Cancer 30(5): 1368-1372, 1972.Saccomanno, C. Carcinoma in situ of the Lung: Its Development, Detectionand Treatment. Seminars in Respiratory Medicine 4(2): 156-160, 1982.Saccomanno, C., Moran, P.G., Schmidt, R.D., Hartshorn, D.F., Brian, D.A.,Dreher, W.H., and Sowada, B.J. Effects of 13-Cis Retinoids onpremalignant and malignant Cells of Lung origin. Acta Cytol. 26: 78-85, 1982.Saccomanno, G., Archer, V.E., Auerbach, 0., Saunders, R.P., and Brennan,L.M. Development of Carcinoma of the Lung as Reflected in ExfoliatedCells. Cancer 33: 256-270, 1974.Salley, J.J. Experimental Carcinogenesis in the Cheek Pouch of the SyrianHamster. Journal of Dental Research 33: 3533-362, 1954.Sartori, M., Sauerbrey, R., Kubodera, S., Tittel, F.K., Roberts, R., andHenry, P.D. Autofluorescence Maps of Atherosclerotic Human Arteries -142A New technique in Medical Imaging. IEEE Journal of QuantumElectronics 23(10: 1794-1797, 1987.Schwartz, S., Absolon, K., and Vermund, H. Some Relationships ofPorphyrins, X-rays, and Tumors. Minnesota University Medical Bulletin27: 7, 1955.Schreiber, H., Saccomanno, C., Martin, D.H., and Brennan, L. SequentialCytological Changes During development of Respiratory Tract TumorsInduced in Hamster by Benzo(a)pyrene-ferric oxide. Cancer Research,1974.Schomacker, K.T., Frisoli, J.K., Compton, C.C., Flotte, T.J., Richter,J.M., Nishioka, N.S., and Deutsh, T.F. Ultraviolet Laser-InducedFluorescence of Colonic Tissue: Basic Biology an dDiagnosticPotential. Lasers in Surgery and Medicine 12: 63-78, 1992.Seifter, E.J. and Ihde, D.C. Small Cell Cancer: A DistinctClinicopathologic Entity. In Lung Cancer Comprehensive Treatise,Bitran, J.D., Golomb, 11.11., Little, A.G., and Weichselbaum, R.R. (eds),Crune and Stratton, Inc., Orlando, Florida, pp 257-279, 1988.Shopland, D.R. Changes in Tobacco Consumption and Lung Cancer Risk:Evidence From Studies of Individuals, In Evaluating Effectivess ofPrimary Prevention of Cancer, Hakama, M., Beral, V., Cullen, J.W., andParkin, D.M. (eds), IARC Scientific Publications N°. 103, Lyon, pp 77-91, 1990.Shopland, D.R., Eyre, H.J., and Pechacek, T.F. Smoking-Attributable CancerMortality in 1991: Is Lung Cancer Now the Leading Cause of Death AmongSmokers in the United States? Journal of the National CancerInstititue 83(16): 1142-1148, 1991.Smith, G.J. The Effects of Aggregation on the Fluorescence and the TripletState Yeild of Hematoporphyrin Derivative. Photochemistry Photobiology41(2): 123-126, 1985.Spiro S.G., Edwards, D., Harper, P.G., Geddes, D.M., and Souhami, R.L.Computed Tomography of the Thorax in the Diagnosis and Management ofMalignant Disease. British Journal of Diseases of the Chest 76: 209-222, 1982.Sporn, M.B. Carcinogenesis and Cancer: Different Perspectives on the SameDisease. Cancer Research 51: 6215-6218, 1991.Stanley, K., and Stjernswàrd, J. Lung Cancer In Developed and DevelopingCountries. Cancer Treatment and Research 45: 1-14, 1989.Stolley, P.D. Editorial Retrospective: Lung Cancer in Women - Five YearsLater, Situation Worse. New England Journal of Medicine 309: 428-429,1983.143Stjernswärd, J. Battle Against Tobacco. Journal of National CancerInstitute: 18(20): 1524-1525 (Editorial), 1989a.Stjernsward, J. Lung Cancer -- A Worldwide Health Problem. Chest 198996(1) Supplement: ls-5s, 1989bStrauss, B.S. The Origin of Point Mutations in Human Tumor Cells. CancerResearch 52: 249-253, 1992.Sutro, C.J., and Burman, M.S. Examination of Pathologic Tissues byFiltered Ultraviolet Radiation. Arch. Path. 16: 346-349, 1933.Svaand, L.O. Light Delivery and Dosimetry. In Phototherapy of Cancer,Morstyn, C. and Kaye, A.H. (eds), Harwood Academic Publishers GmbH,Switzerland, 73-99, 1990.Tang, G.C., Pradhan, A., and Alfano, R.R. Spectroscopic DifferencesBetween Human Cancer and Normal Lung and Breast Tissues. Lasers inSurgery and Medicine 9: 290-295, 1989a.Tang, G.C., Pradhan, A., Sha, W, Chen, J., Liu, C.H., Wahi, S.J., andAlfano, R.R. Pulsed and CW Laser Fluorescence Spectra from cancerous,normal, and Chemically Treated Normal Human Breast and Lung Tissues.Applied Optics 28(12): 2337-2342, 1989b.Thomas, J.P., Hall, R.D., and Girotti, A.W. Singlet Oxygen Intermediacy inThe Photodynamic Action of Membrane-Bound Hematoporphyrin Derivative.Cancer Letters 35: 295-302, 1987.Thomas, P., Rubinstein, L., and the Lung Cancer Study Group. CancerRecurrence After Resection T1NO Non-Small Cell Lung Cancer. AmericanThoracic Surgery 48: 242-247, 1990.Thomas, P., Rubinstein, L., and the Lung Cancer Study Group. LateApperance of malignancies After Surgery For Tl NO Non-Small Cell LungCancer. Lung Cancer 7(supplement): 82, Abstract 299, 1991.Thomson, S. Pathologic Analysis of Photothermal and PhotomechanicalEffects of Laser-Induced Interactions. Photochemistry and Photobiology53(6): 825-835, 1991.Tockman. M.S., Gupta, P.K., Myers, J.D., Frost, J.K., Baylin, S.B., Gold,E.B., Chase, A.M., Wilkinson, P.H., and Muishine, J.L. Sensitive andSpecific Monoclonal Antibody Recognition of Human Lung Cancer Antigenon Preserved Sputum Cells: A New Approach to Early Lung CancerDetection. Journal of Clinical Oncology 6(11): 1685-1693, 1988.Tomson, S.H., Emmett, E.A., and Fox, S.H. Photodestruction of MouseEpithelial Tumors After Oral Acridine Orange and Argon Laser. CancerResearch 34: 3124-3129, 1974.144Topal, M.D. DNA Repair, Oncogenes, and Carcinogenesis. Carcinogenesis 9:691-696, 1988.Twersky, V. Multiple Scattering of Waves and Optical Phenomena. J. Opt.Soc. Am. 52: 145-171, 1962.Twersky, V. Absorption and Multiple Scattering by Biological Suspension.J. Opt. Soc. Am. 60: 1084-1093, 1970.Uei, Y., and Noguchi, Y. Comparison of Cytologic Markers for AutomatedLung Cancer screening. Analytical and Quantitative Cytology andHistology 12(5): 327-332, 1990.Udenfriend, S. Fluorescence Assay in Biology and Medicine Volume I. InMolecular Biology, An International Series of Monographs and Iextbooks,Volume 3, Kaplan, N.O., and Scheraga, H.A. (eds), Academic Press, NewYork, 1962.Udenfriend, S. Fluorescence Assay in Biology and Medicine Volume II. InMolecular Biology, An International Series of Monographs and Iextbooks,Kaplan, N.O., and Scheraga, H.A. (eds), Academic Press, New York,1969.van Gemert M.J.C., and Henning, J.P.H. Model Approach to Laser CoagulationDermal Vascular Lesions. Arch. Dermatol. Res. 270: 429-439, 1981.Vogel, A., Dlugos, C., Nuffer, R., and Bringruber, R. Optical Propertiesof Human Sclera, and Their Consequences for Transscleral Laserapplications. Lasers in Surgery and Medicine 11: 331-340, 1991.Vogelstein, B., Fearon, E.R., Hamilton, S.R., Kern, S.E., Preiisinger,A.C., Leppert, M., Nakamura, Y., White, R., Smits, A.M., and Bos, J.L.Genetic Alterations During Colorectal-Tumour Development. New EnglandJournal of Medicine 319: 525-532, 1988.Wagnières, G., Depeursinge, Ch, Monnier, Ph., Savary, M., Cornaz, P.,Châtelain, A., van den Bergh, H. Photodetection of Early Cancer byLaser Induced Fluorescence of a Tumour-selective Dye: Apparatus Designand Realization. SPIE 1203: 43-52, 1990.Wan, S., Anderson, R.R., Parrish, J.A. Analytical Modeling for The OpticalProperties of the Skin With In Vivo and In Vitro Applications.Photochemistry and Photobiology 34: 493-499, 1981.Weber, J.E. Applications of Multivariate Analysis in Diagnostic Cytology.Analytical and Quantitative Cytology and Histology 10(1): 54-72, 1988.Weil, G.J., Chused, T.M. Eosinophils Autofluorescence and its Use inIsolation and Analysis of Human Eosinophils Using FlowMicrofluorometry. Blood 57 (6): 1099-1104, 1981.145Weinstein, I.B. Cancer Prevention: Recent Progress and FutureOpportunities. Cancer Research (Supplement) 51: 5080s-5085s, 1991.Weishaupt, K.R., Gomer, C.J., and Dougherty, T.J. Identification ofSinglet Oxygen as The Cytotoxic Agent in Photoactivation of a MurineTumor. Cancer Research 36: 2326-2329, 1976.Welch, A.J., van Gemert, M.J.C., and Bradley, A.B. Optical and ThermalEvents in Laser Angioplasty. In Primer and Laser Angioplasty,Ginsburg, R., and White, J.C. (eds), Futura Publishing, Inc., New York,pp 89-119, 1989.Welch, A.J., Yoon, G., van Gemert, M.J.C. Practical Models for LightDistribution in Laser-Irradiated Tissue. Laser in Surgery and Medicine6: 488-493, 1987.Welsh, F., O’Connor, M.J., and Langfitt, T.W. Regions of Cerebral IschemiaLocated by Pryidine Nucleotide Fluorescence. Science 198: 951-953,1977.Wied, G.L., Bartels, P.H., Bibbo, M., and Sychra, J.J. CytomorphometricMarkers for uterine cancer in Intermediate Cells. Analytical andQuantitative Cytology 2(4): 257-363, 1980.Wilbanks, G.D., and Richart, R.M. Fluorescence of Cervical IntraepithelialNeoplasia Induced by Tetracycline and Acridine Orange. Amer. J.Obstet. Gynec. 106(5): 726-730, 1970.Wilson, B.C., and Adam, C. A Monte Carlo Method for the Absorption andFlux Distributions of Light in Tissue. Medical Physics 14: 835-841,1987.Wilson, B.C., Patterson, M.S.,, Flock, S.T., and Wyman, D.R. Tissue OpticalProperties in Relation to Light Propagation Models and In VivoDosimetry. In Photon Migration in Tissues, Chance, B. (ed), PlenumPress, New York, pp 25-42, 1990.Wilson, B.C., and Jacques, S.L. Optical Reflectance and Transmittance ofTissues: Principles and Applications. IEEE Journal of QuantumElectronics 26 (12): 2186-2199, 1990.World Health Organization Histological Typing of Lung Tumours. SecondEdition. World Health Organization, Geneva, Switzerland. AmericanJournal of Clinical Pathology 77: 123-136, 1982.Woolner, L.B. Pathology of Cancers Detected Cytologically. In Atlas ofEarly Lung Cancer: National Cancer Institute, National Institutes ofHealth, U.S. Department of Health and Human Services, Igaku-Shoin Ltd.,Tokyo, pp 108, 1983.Woolner, L.B., Fontana, R.S., Cortese, D.A., Sanderson, D.R., Bernatz,P.E., Payne W.S., Palvolero, P.C., Piehler, J.M., and Taylor, W.F.146Roentgenographically Occult Lung Cancer: Pathologic Findings During a10 Year Period. Mayo Clinic Proceedings 59: 453-466, 1984.Wooten, R.S., Smith, K.C., Ahlquist, D.A., Muller, S.A., and Balm, R.K.Prospective Study of Cutaneous Phototoxicity After SystemicHematoporphyrin Derivative. Lasers in Surgery and Medicine 8: 294-300,1988.Yang, Y., Ye, Y., Li, F., Li, Y., and Ma, P. CharacteristicsAutofluorescence For Cancer Diagnosis and Its Origin. Laser in Surgeryand Medicine 7: 528-532, 1987.147Chapter 7: APPENDIX7.1 Discriminant Function Analysis*Discriminant analysis is concerned with the problem of classifyingobjects into groups or populations on the basis of a set of feature values.In discriminant function analysis, a function of the observed variables isestimated such that the associated probability or cost of misclassificationis minimized for a set of cases whose classification is known.Classification algorithms cannot be used to determine if natural groupingsexist within an undifferentiated population. In addition, classificationalgorithms should not be used as a verification of a suspected groupingwithin a population.Sample based classification rules for discriminant analysis areusually constructed from random sets called “learning” or “training”samples. The aim is to identify characteristics features of the samplepopulation, and use these to develop a classification rule that partitionsthe set of all possible sample observations into subsets which correspondto selected groups or classes. In practice, no classification rule canprovide an error-free assignment. This also holds for the training samplesince the classes or groups overlaps. In general, the criterion forclassification is to minimize the total error of classification (TEC). Inthe two group case, the total error of classification is as follows:*References used in the preparation of Appendix 7.1 were Weber, 1988 andLachenbruch, 1975.148“True” Classification Classification AlgorithmGroup A Group BGroupA a bGroupB c dTEC = b+c/(a+b+c+d)In addition to the TEC, prior probabilities of occurrence and costs ofmisciassifications are sometimes taken into account when defining aclassification rule. Discriminant functions used for assigningobservations to defined classes or groups on the basis of a set of featurevalues are generally based on Bayes’ Classification Rule. Bayes’ Rulestates:for N groups Gk (l k N), assign objects to group G1 IfP(G.I) > P(GI) for all j not equal to iwhere X is the vector of measured feature values. Conditionalprobabilities of the form P(GkI) are difficult to determine, but P(lGk)can be estimated from sample of objects from group K. The application ofBayes’ rule can be simplified using Bayes’ Theorem:P(GkI) = P(XIGk)P(Gk)/P(XIGk)P(Gk)149P(Gk) is the “a priori” probability (i.e the probability that an objectcomes from group k). In practice, therefore, the Bayes’ Classificationrule is used in the form:assign the object to group C1 ifP(IG.)P(G.)/>P(XIGk)P(Gk) > P(XIG.)P(C.)/P(IGk)P(Gk)or P(IG.)P(G.) > P(G)P(G.) for all j not equal to iEstimation of P(IGk) for a sample with features unknowndistributions requires the collection and analysis of large volumes ofdata. The amounts of data can be reduced if certain assumptions are madeabout the feature distributions for the population of interest. Inpractice, the most commonly used assumption in discriminat analysis is thatX is a p-dimensional vector of observations belonging to multivariatenormal distributions. The conditional probabilities for each group can becalculated relatively simply according to the multivariate normaldistribution. The general equation of the multivariate normal distributionis:PQ) = exp[-((-)’ II1(&)/21Applying Bayes’ Rule:P(XIGk) = [2 /21;Ih/2]_1 exp[-((X-k)’IZI(X k))/2]where n is the number of features, k is the vector of feature means forgroup k, and is the covariance matrix for group k. and must beestimated for sample populations known to belong to group k. After150substituting the above equation into Bayes’ Rule and simplifying by takingthe logarithm to eliminate exponential terms, Bayes’ Rule for theniultivariate normal case is:assign the object with feature values to Gk ifd.(X)-ln(P(G.)) <d3(X)-ln(P(G)) for all j not equal to iwhere dk(X) = l/2lnIj+l/2(X-k)’II1(X-jk) is the “discriminant score”.This version of Bayes’ Rule is known as the Quadratic DiscriminantFunction. Within the n-dimensional space defined by the n features theboundaries between the regions to which a given object could be classifiedare quadratic surfaces. A quadratic Bayesian decision boundary for a two-dimensional, two-group, two-feature case is illustrated in figure 32a.For the special case when the covariance matrices are the same forall the groups, i.e. E. for all k, the decision boundaries reduce tolinear equations. In this case, Bayes’ rule for the multivariate normalcan be expressed as:assign the object to group G1 iff1()+ln(P(G)) > f()ln(P(G)) for all i not equal jwhere f(X) = .t21Xh/j&.1, > is the common covariance matrix for allthe groups and f() is the linear discriminant function.151The groups are separated by linear hyperplanes in the feature space.For the two-feature case, the Bayesian decision boundary is a straight lineas illustrated in figure 32b.Estimation of Error RatesError rates or misclassification probabilities for the discriminantfunction may be estimated by one of the following methods:i) Test the classification rule on an independent sample of size M. Formisclassified, e objects the estimated “true” error is E = e/M.ii) The misclassification rate for the training sample that was used togenerate the discriminant function can be calculated. This is knownas the apparent error rate. Usually this underestimates the “true”error rate since the data used to develop the discriminant functionare also used to test its performance.iii) A classifier can be designed with one case removed from the learningset. The discriminant function is then tested by classifying thecase that was left out. This process is repeated for all cases inthe sample. This technique is known as the “leaving-one-out” method.It has the disadvantage of requiring that M classifiers be designed.It is therefore generally only useful for small sample.iv) A sample reuse (jackknife) procedure (Lachebruch, 1975) whichinvolves classifying each observation using the classification ruleobtained by omitting that observation from the training sample.Briefly, the procedure starts with a group of N1 observations. Anobservation is then left out from this group and a classificationfunction based on the remaining N1-l and N2 observations (for two152group case) is calculated. The observation left out is thenclassified. These steps are repeated until all the N1 observationshave been classified. These steps are then repeat through for the N2group. The jackknife misclassification rate is then nlM/nl for thegroup N1 and for group the N2, where and 2M are the numberof left out observation misclassified in group N1 and N2respectively. n1 and n1 are the number of observations in therespective groups N1 and N2.Feature SelectionTo reduce the number of features that is required for objectclassification, it may be beneficial to combine a feature selectionprocedure with the discriminant function analysis. The aim is to reducethe number, n, of original features variables to the m variables that givethe best error rate. The criteria for feature selection are typicallybased on analysis-of-variance statistics. The most frequently used methodfor selecting features to include in the discriminant analysis is thestepwise procedure. This can be performed as a stepwise forward method, astepwise backward method, or a full stepwise method. In stepwise forwardanalysis, the variable with the maximum measure of goodness (i.e, bestdiscriminates between (or among) the group) is selected first. Thevariable next is the one which, when combined with the first variablemaximises the measure of goodness. This process continues until inclusionof another variable does not significantly improve discrimination. In thestepwise backward procedure, all the variables are initially included andat each stage, the variable that results in the smallest decrease in153measure of goodness is discarded. For the full stepwise procedure,analysis begins with an initial subset of the variables in the discrintinantanalysis. Possibilities include the empty set and the full set of all thevariables. Each variable is tested separately for possible exclusion andthe variable with the smallest non-significant reduction in the criterionis dropped (i.e. the decrease in the measure of goodness produced byremoving a variable is examined and if the decrease is below a specifiedthreshold, the variable is removed). This is repeated until no morevariables can be removed without causing a significant reduction in thecriterion. Each variable not included in the discriminant function is thentested for possible inclusion, and variable which provides the greatestimprovement in the criterion are added back to the set. In practice, theentering and removal thresholds should be different to avoid cycling. Thefeatures are entered into the analysis corresponding to which has thehighest conditional F ratio (this is the measure of goodness used instepwise methods). The conditional F ratio is calculated for eachunentered feature through a one-way analysis of covariance, where thecovariates are the features already entered (for the first step, where nofeatures have yet been entered, the feature with the highest value for theF statistic is entered). The features are entered into the discriminantfunction one at a time, and are allowed to remain in the discriminantfunction if they satisfy a user-specified minimum F-value. The minimum Fvalue for feature inclusion in the discriminant function was set to be0.005. Features that are highly correlated with those already entered intothe analysis will not, however, be entered. Once the final set of featureshas been selected the discriminant function is then calculated accordingly.154Figure 32C%JL4.4Examples of (a) Linear and (b) quadratic decision boundary fora two dimensional, two group and two feature case by adiscriminant function. Ellipses represent equal probabilitycontours of normally distributed, bivariate feature data.abquadratic decision boundaryfeature 1group Agroup Blinear decision boundaryfeature 1155For this thesis, a commercially available program, BMDP (BiomedicalData Processing, BMDP Statistical Software, Los Angeles, California), wasused to generate and test discriminant functions. A stepwise analysis ofthe data was performed using the 7M discriminant analysis routine of thissoftware package.1567.2 In vitro SpectroscopyTissue SpecimensThoracotomy specimens of normal and malignant tissues from patientsundergoing surgery for lung cancer were obtained at the time of lungresection. These fragments were rinsed in saline or phosphate bufferedsaline (PBS) to remove any excess blood. Measurements of fluorescencespectra were made in freshly obtained specimens, in specimens snap frozenin liquid nitrogen at -70°C, thawed and kept moist at 4°C in PBS for up to3 hours. In addition to the fluorescence spectra of the composite normalbronchial tissues, autofluorescence spectra were also measured before andafter removal of the surface epithelium and then the subutucosa tissue tocharacterize the autofluorescence of the different layers of the bronchialwall. Following the fluorescence spectroscopy studies, these specimenswere fixed in PBS with 10% formalin for histological confirmation. Thetissues were processed routinely and stained with Hematoxylin and Eosin(H&E). Laser induced fluorescence spectra were also obtained from intactspecimens and after homogenization of the specimens using a homogenizer.Extracellular matrix constituents and other chromophoresBasement Membrane MatrigelTM were obtained from CollaborativeResearch Incorporation. Matrigel is a solubilized basement membranepreparation from Engeibreth-Hoim-Swarm (EHS) mouse sarcoma. Its majorcomponent is laminin, collagen IV, heparan sulfate proteoglycans, entactinand nidogen. It also contains tumour growth factor-beta (TGF-beta),fibroblast growth factor, and tissue plasminogen activator. Matrigel is157formulated in Dulbecco’s Modified Eagle’s Medium with 10 ug/mi gentanlycin.It is kept at -20°C and thawed at 4°C overnight before being used. It ismaintained in a gelled consistency at 22°C to 37°C, and fluorescencespectra and imaging measurements were made at room temperature with gelthickness of 0.5 to 10 nun in either petri dishes or optical glass cuvettes,Cell CultureThe cells used in all of the studies in this part of the project wereHTB 58 SK-Mes-1 human lung carcinoma, A549 human lung adenocarcinoma, andCCL-210 CCD-19L4 human lung diploid (normal), obtained from ATCC. Inaddition, short-term cultures of human skin fibroblast were obtained from ahealthy donor.The cells were routinely grown as monolayer culture in 75-cm2polystyrene tissue culture flasks (Falcon). Chinese Hamster V79 lungfibrobiast lines and A549 human lung adenocarcinonia were maintained inEagle’s F-is minimum essential medium (MEM) (Gibco) supplemented with 10%fetal bovine serum (Gibco). HTB 58 SK-Mes-l human lung carcinoma lineswere maintained in minimum essential medium (HEM) with non-essential aminoacids and sodium pyruvate (Gibco) supplemented with 10% fetal bovine serum.CCL-210 human lung diploid lines were maintained in minimum essentialmedium (MEM) supplemented with 10% fetal bovine serum. The short-termculture of human skin fibroblasts were maintained in Eagle’s minimumessential medium supplemented with 20% fetal bovine serum.The cell cultures were incubated at 37°C in a 5% CO2 humidifiedatmosphere environment. Cells were maintained in exponential growth andwere routinely subcultured when grown to confluence. Subculturing was158usually achieved by try-psinization. The cells were first washed with about5 ml of 0.1% trypsin solution (Gibco) and then exposed to about 3 ml of thetrypsin for about 6 minutes at 37°C. Trypsin action was neutralized by theaddition of fresh medium, which was then pipetted vigorously to obtainsingle cells. Cells were then plated in new tissue culture flask.Samples of cells for each experiment were removed from the growthmedium by centrifugation at 600 rpm, at 4°C (Model RC-3, Sorvall) for about6 minutes. For the experiment using the fluorescence activated cellsorter, the cells were re-suspended in phosphate buffer saline (PBS) andvortexed for 5 seconds to disperse the cells uniformly. For measurementsof the fluorescence with cells layer on basement membrane, the cell sampleswere re-suspended in fresh medium and allowed to attach to the basementmembrane in petri dishes (Falcon) for about 2 hours.Animal ModelSquamous cell carcinoma was induced in the buccal cheek pouch ofSyrian Golden hamsters using a modification of the method reported bySalley (Salley, 1954). Cotton sutures impregnated with 9,10-dimethyl 1,2-benzanthracene (DMBA) and covered with a silicone sheath were sewn into thehamster cheek pouch. The development of dysplasia at 6-12 weeks, carcinomain situ in 12-14 weeks and large invasive cancers at 12-14 weeks wereobserved. Fluorescence spectroscopy using the optical multi-channelanalyzer and fluorescence imaging were performed on these hamsters.159Optical Properties MeasurementsA standard integrating sphere (Oriel, Strattford, CT, Model 70491)was used for the optical transmittance and reflectance measurements. Theset-up for these experiments is shown in figure 33. The integrating spherewas used to measure the total transmittance, T and total reflectance, Rfrom the freshly obtained bronchial fragments and fresh samples from thehamster cheek pouch. All measurements were performed in vitro with thespecimens sandwiched between two glass slides and kept moist inphysiological saline or PBS (phosphate buffer saline). In all cases, thespecimens were illuminated with a 442 mu collimated laser beam and theintegrated reflectance and transmittance were measured using the lineardetector of the optical multi-channel analyzer.For the reflectance measurements the sample was placed at thereflectance port of the integrating sphere as shown in figure 33. Thespecular component of the reflectance was captured by a light trap withinthe integrating sphere. The inner surface of the integrating sphere had acoating of barium sulphate (BaSO4). The 100% reflectance (or reference),Rr was measured by placing a high reflectivity barium sulphate coated platein the reflectance port of the integrating sphere. The total reflectanceof the sample is then given by R = (Rm/Rr)P; where p is the givenreflectance coefficient at 442 rim of the barium sulphate plate standard.The total transmittance of the sample was measured in the same mannerexcept that the sample was placed in the transmittance port of theintegrating sphere while the reflectance port and light trap were replacedwith masks coated with barium sulphate.160Figure 33 Schematic diagram of the set-up of the transmittance andreflectance measurements using an integrating sphere.FIBER OPTIC TO CMASPECULAR REFLECTIONLIGHT TRAPBEWEEN GLASS SLIDES:NCIDCNT BEAM 342 r. LASERSPECIMEN BETWEEN CLASS SLIDESINTEGRATING SPHERE/ i\I R N SM ITT A N CE REFLECTANCE161The background corrections were performed using two glass slides witha drop of saline or PBS in between. Immediately following measurements,the tissue specimens were place in 10% formalin for routine histologicalpreparation and Hematoxylin and Eosin staining for histologicalexamination. Sections were made on the specimens such that the thicknessof the specimens could be measured using quantitative microscopy. Theabsorption and scattering coefficients were then calculated.1627.3 Optical Multi-Channel AnalyzerDescription of the Spec troanalyzer ComponentsThe basic components of the system consist of a spectrum acquisitionmodule link to a processing system for real-time digital signal processing,which is controlled by a host computer.The emitted autofluorescence was focussed onto the entrance slit ofthe spectrograph (Jarrell-Ash Monospec 27, with a 300g/nim gratings, blazedat 500 rim, and a 0.29 rim/element resolution) with a f/l.5 collimating lensand an f/3.14 focussing lens to match the f/3.8 spectrograph. Thespectrograph was fitted with a 2 mm-width slit. The spectrograph uses acrossed path Czerny-Turner design to minimize re-entry spectra. Lightpasses through the entrance slit reflect from the collimating mirror to agrating. The 300g/mm gratings blazed at 500 run diffracts the light suchthat the angle by which it reflects back to the focusing mirror is afunction of wavelength. The dispersed spectrum is then collected at theexit turning mirror to a detector. The spectrograph was adjusted to coverthe 450 rim to 750 run range. Detailed aspects of the optics of spectroscopyis discussed in Optics of Spectroscopy (Lerner, 1990).The light at the exit port was recorded with an intensified 1024-elements silicon linear diode array detector (EG & G PARC, model 1455R700HQ). The model features a 18 nun diameter intensifier with manual gaincontrol. The intensifier is red enhanced, high quantum efficiency of about15.7% at 550 run and a 1024-elements silicon diode array with 700 “active”elements.163The detector was interfaced to the Optical Multi-Channel Analyzer bya detector interface/controller with a 14-bit analog digital converter.The OMA controller was interfaced to a 80386 IBM compatible computer fordata storage and spectral display and processing. The spectra processingand display of the incoming analog spectrum was performed in real time.Background subtraction was performed on each spectrum during theacquisition mode before transmitting the data to the host computer tocorrect for the dark noise of the detector. The operation and installationof the system is described in detail in the EG & C Princeton AppliedResearch manual.System CalibrationSpectral calibration for the x-axis (wavelength axis) of the OMA wasperformed using the spectrum of a low-power-low-pressure mercury-argon lamp(Oriel, Strattford, CT, Model 6035). The calibration is linear, meaningtwo reference wavelength for two different points of the reference mercurycurve on the screen are selected. The program then generates thewavelengths on the x-axis and stores this information into a parameterfile.Correction for non-linear spectral response of the detector wasperformed using a standard lamp. The correction factors are obtained byobserving the wavelength-dependent output from a calibrated light. Thedetection system was then calibrated by measuring the intensity of thestandard lamp versus the wavelength using the detector system. Thesensitivity (correction factors) of the detection system is then calculatedby dividing the measured intensity of the standard lamp versus the spectral164output data provided with the lamp (known standards). All spectra are thencorrected by dividing or multiplying the correction factors to account forthe non-uniform response of the detector. However, in the wavelengths ofinterest, it was found that the detector response was almost uniform.

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0085057/manifest

Comment

Related Items