UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluation of digital and geometric unsharpnesses in dental radiographs Radan, Elham 1999

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

Item Metadata

Download

Media
831-ubc_1999-0287.pdf [ 22.55MB ]
Metadata
JSON: 831-1.0089020.json
JSON-LD: 831-1.0089020-ld.json
RDF/XML (Pretty): 831-1.0089020-rdf.xml
RDF/JSON: 831-1.0089020-rdf.json
Turtle: 831-1.0089020-turtle.txt
N-Triples: 831-1.0089020-rdf-ntriples.txt
Original Record: 831-1.0089020-source.json
Full Text
831-1.0089020-fulltext.txt
Citation
831-1.0089020.ris

Full Text

Evaluation of Digital and Geometric Unsharpnesses in Dental Radiographs i y Elham Radan D.M.D., The University of Shahid Beheshti, Iran, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS OF THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Oral Biological and Medical Sciences We accept this thesis as conforming tO/the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1999 ©Elham Radan, 1999 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of O&HS ~ P>^tr'^ The University of British Columbia Vancouver, Canada Date 30 ^ / /f?? DE-6 (2/88) Abstract Recent advances in digital electronic devices have led to developments in digital radiography. Increased interest in imaging systems other than conventional fi lm-based radiography has been stimulated by anticipated reduction in patient radiation exposure wi th increased convenience and image quality. Wi th the new imaging systems developing, it is necessary to evaluate the image quality wi th comparison to conventional film-based radiography. The aim of this study was to develop a definitive evaluation based on an endodontic file model for comparison of the effects of geometric and digital unsharpnesses in dental radiography. The endodontic file has a well-defined outline and small size, which makes it ideal to test resolution. Files of sizes 6, 8, 10 and 15 were placed in the root canal of an extracted lower incisor at 3 different positions wi th 2mm increments. Radiographs were taken at magnification factors of 1.0,1.1,1.2 and 1.3. Each radiograph was digitized at resolutions of 75, 150, 300 and 600 dots per inch (dpi). A digital visual analog scale was used for the assessment of image quality. Ten observers viewed the 192 digital images and identified the positions of the end of the files and the end of the root. Measured distances between these points were compared wi th a gold standard (actual file positions). Analysis of variance and paired T-tests were used to demonstrate significant differences. N o significant differences were found between the resolutions of 300 and 600 dpi , nor between the magnifications of 1.0 and 1.1. Significant differences were found between the magnification of 1.3 and the others, and the resolution of 75 dpi and the others, for al l the files. The effects of geometric unsharpness wi th in the range studied for all the file sizes were not significant except for the magnification of 1.3, which is of little clinical relevance. Based on this study, wi th endodontic files in an extracted tooth, there is little need for resolutions higher than 300 dpi in clinical dental radiography and geometric unsharpness has no influence on diagnostic information. This may apply to many other aspects of diagnostic imaging. i v Table of Contents Abstract i i Table of Contents i v List of Tables v i i List of Figures v i i i Acknowledgements x i I N T R O D U C T I O N 1 Chapter One Advances in Imaging 4 1.1 - Conventional Radiography 5 1.2 - Digital Radiography 8 Chapter Two Review of Literature 11 Chapter Three Image Unsharpness 24 3.1 - Sharpness and Resolution 25 3.2 - Causes of Unsharpness 3.2.1 - Geometric Unsharpness and its Relevance to Clinical Diagnosis 27 3.2.1.1 - Focal Spot 29 3.2.1.2 - Magnification and Geometric Unsharpness 34 3.2.2 - F i lm Unsharpness 37 3.2.3 - Parallax Unsharpness 38 3.2.4 - Movement Unsharpness 38 V 3.2.5 - Digital Unsharpness 39 3.3 - A n Experimental Design to Determine the Position of the Focal Spot 40 Chapter Four Endodontic File Study Model 48 4.1 - Pilot Study 49 4.2 - Definitive Study 51 4.2.1 - Goals 52 4.2.2 - Materials and Methods 53 4.2.2.1 - Test Objects 53 4.2.2.2 - Radiography 57 4.2.2.3 - Digit izing 61 4.2.2.4 - Observation and Interpretation 64 Chapter Five Analysis and Results 70 5.1 - Data Analysis 71 5.2 - Results 72 Chapter Six Discussion and Conclusions 97 6.1 - Discussion and Conclusions 98 6.2 - Different Aspects of Measurements 104 6.2.1 - Resolution 104 6.2.2 - Precision 104 6.2.3 - Accuracy 104 6.2.4 - Validi ty 105 r 6.2.5 - Reliability 106 Bibliography 110 Appendix Pilot Study Paper 125 List of Tables 1. Speed groups of intra oral films. 6 2. History of Kodak intraoral dental film. 7 3. Gradients of the central row images of the pewter plate. 43 4. Sizes of endodontic K-type files. 51 5. Digital file sizes for a 256 grey shade image, one inch square. 52 6-a. Analyses of Variance: Difference in distance. 83 6- b. Bonferroni p-values: Difference in distance. 84 7- a. Analyses of Variance: Observers' assessment of image quality. 85 7-b. Bonferroni p-values: Observers' assessment of image quality. 86 8. Bonferroni p-values for difference in distance from the total study. 89 9. A N O V A for the two observer types. 95 10. The relationship between pixel size and resolution. 98 List of Figures 1. Radiograph of a line pair plate. 26 2. Diagram to show the effects of source to object and object. 28 to f i lm distance on penumbra (Geometric Unsharpness). 3. Diagram to show the effect of focal spot size on penumbra. 30 4. Diagram to show the effect of source to object distance 31 on penumbra. 5. A dental x-ray tube. 32 6. Anode and cathode of an x-ray tube. 33 7. Target face showing focal spot, viewed perpendicular to the face. 34 8. Target face showing focal spot, viewed perpendicular to the long 34 axis of the x-ray tube. 9. Image of the focal spot of the x-ray unit used in this study. 35 10. Diagram of the relationship of source, pinholes, and fi lm 36 used to produce the focal spot image. 11. Diagram to show the effect of source to object and source to fi lm 37 distance on magnification. 12. Pewter plate wi th the grid of 300 Jim pinholes. 41 13. Test column used for standardizing the relationships of the 42 x-ray source, object and film. 14. X-ray unit secured in a reproducible position at the top of the 44 test column. 15. Diagram to show the relationship between x-ray source, 45 pewter plate, and film, on the test column. 16. Scanned image of pewter plate showing focal spot images. 46 Profile plot of one row of focal spot images. 17. Pinhole distances plotted against number of pinholes from 47 i x the central row of the focal spot images. 18. V iew of the root attached to the jig showing the sectioned surface. 53 19. Root attached to a part of the jig wi th self autopolymerizing 54 acrylic resin. 20. Photomicrograph of apex of root wi th protruding endodontic 55 file. Superimposed image of stage micrometer (1mm). 21. Endodontic file wi th 2 m m stops. 56 22. Jig used to maintain the endodontic file position in the root. 56 23. Position of jig on the test column. 58 24. Processed unexposed film and radiograph of penetrameter. 60 25. A n example of 3 positions of file size 15, wi th resolution of 62 600 dpi and magnification of 1. 26. Outline of template superimposed on scanned image of root 63 to enable reproducibel positioning of the images. 27. A n example of the slides wi th root containing endodontic files 65 and visual analog scale. 28. A screen photograph of the marked images on one of the slides. 66 29. Screen photograph of observers' marks at the end of the root 68 and the end of the file and relationship of the coordinates of the two points. 30. Measured distance of 10 m m on the scanned image of a ruler. 69 31. Density of the radiographs plotted against processing order. 73 32. Mean difference in distance for each digital resolution (pooled 75 magnification). 33. Observers' assessments for image quality for each digital 76 resolution (pooled magnification). 34. Mean difference in distance for each magnification (pooled 77 digital resolution). 35. Observers' assessments of image quality for each magnification 78 (pooled digital resolution). 36. Mean difference in distance for each digital resolution and 79 each magnification, files 6 and 8. 37. Mean difference in distance for each digital resolution and 80 each magnification, files 10 and 15. 38. Observers' assessments of image quality for each digital 81 resolution and each magnification, files 6 and 8. 39. Observers' assessments of image quality for each digital 82 resolution and each magnification, files 10 and 15. 40. File 15, removing dpi 75 and magnification of 1.3 from the graph. 87 41. Studentized residuals plotted against estimates for the four 90 file sizes. 42. Difference in distance for first and second readings for the 5 of 91 the repeat images. 43. Observers' assessment of image quality for first and second 92 readings for 5 of the repeat images. 44. Objective and subjective evaluations of the best images (mag 1, 93 dig 600) for 3 different positions. 45. Objective evaluation of the positions using silver standard. 94 46. Four digital images of group # 18 (10 lp /mm)of the line pair 99 plate shown in Figure 1, at the four resolutions used in this study. 47. Scatter plot relating the differences in distance and the observers' 107 subjective assessments for the 1st and 2nd readings. Acknowledgements I am most grateful to my supervisor Dr Colin Price without whose help and support, completion of this thesis would not be possible. Endless hours of discussion gave me such good overview to understand the problem especially through his meticulous care for details. His efforts in helping me to learn a structured way of approaching scientific study, were invaluable. Working in radiology department for over 2 years, I was fortunate to be trained by great radiology experts, Dr Price and Dr McDonnell, who introduced me to the logic behind radiographic interpretations. I would also like to thank my committee members, Dr Alan Hannam, Dr Jeff Coil, Dr Doug Waterfield and Dr Donal McDonnell for their contribution and helpful comments to bring this thesis to fruition. I wish to take this opportunity to thank Dr Don Brunette for his useful suggestions and helping me in terms of critical thinking through his valuable book. The floral radiographs illustrating the chapter 3 and 6 title pages and the digital illustration of chapter 1 are reproduced by the kind permission of Dr A. G. Richards and Dr C. Price respectively. Mr Bruce McCaughey has cheerfully produced the photographs and I much appreciate his help and skill. I would also like to show my appreciation to Dr Lewei Zhang for her assistance to make the photomicrograph and Dr Bob Pretty who kindly helped me with the sectioning device. Finally I would like to express my gratitude to my parents who have always encouraged me in following my dreams and last but not least my husband Mehdi whose unconditional love and support have always given me peace of mind and I couldn't ask for a better soulmate... INTRODUCTION In clinical radiography, diagnostic accuracy is influenced by image quality and observer skills. Image quality is dependent on three factors: contrast, resolution and noise (Okano et al, 1983) "Contrast" is the difference in density between various parts of the image. A n image that shows very light areas and very dark areas has high contrast, also referred to as a short grey scale of contrast, because there are few shades of grey between the black and white images. A radiographic image that is composed of light grey and dark grey zones has low contrast, also referred to as a long grey scale of contrast (Goaz and White, 1994). Contrast depends on differences in thickness, material and imaging method. "Resolution" is the ability of the imaging system to form a distinct image of two or more objects located close together. "Image sharpness" determines how wel l the min imum details of an object are reproduced on the radiograph. It is in fact a measure of how wel l a boundary between two areas of different contrasts is detectable. Although sharpness and resolution are two distinct features, they are to some degree interdependent, being influenced by the same geometric variables (Goaz and White, 1994). "Noise" is anything that is not part of the real image. It occurs in different stages and detracts from image quality. Faster systems, which require less radiation exposure, have more noise because there are fewer x-ray quanta to form the image. Noise can originate in emulsions and processing in conventional radiography or in digital radiography wi th reduced resolutions. In radiographic images, "unsharpness" or blur of the image can be caused by five potential sources: focal spot (the finite size of the x-ray source), source to object and object to receptor distances, receptor, digitization and motion. Unsharpness that is caused by focal spot size, or the geometry of the distances between the source to the object and the object to the receptor, is called "Geometric Unsharpness" (Fitzgerald, 1947). In film-based radiography, receptor unsharpness is caused by the finite size of the emulsion grains, the thickness of the emulsion, and, in the case of double emulsion films, parallax between emulsions. In screen/film systems, the intensifying screens and transfer of light from them add a much greater measure of unsharpness. In digital radiography, unsharpness is due to the limitation of resolution of digital receptors, in direct systems, or limits of the size of the file created during scanning conventional radiographs, in indirect systems. These issues are explained in more detail in chapter 3. Movement unsharpness is the blur that occurs if either the patient or the x-ray source moves during the exposure. In the field of oral radiology, there has been little attention to focal spot characteristics and geometric unsharpness. A study performed by Platin et al, in 1996, shows the effect of focal spot size in diagnostic accuracy, using a dental caries . model (Platin et al, 1996). Al though the interpretation of dental caries is a common task in clinical dentistry, because of the ill-defined edge of the lesions, it depends more on broad area contrast than on resolution. In other words, in dental caries there is no definite edge between the lesion and the sound surface, which makes it a poor model for studying the sharpness and resolution of the image. Platin et al. showed that there was no relevance of the effect of focal spot size and diagnostic accuracy. They concluded there was no need for the annual measurement of the x-ray unit focal spot size which was recommended by the American Academy of Oral and Maxillofacial Radiology as part of the dental radiographic quality control program. In digital images, the unsharpness is caused by limits of resolution in both direct and indirect digital systems. Digital information is any information that is represented in discrete units, whereas the analog information is represented in continuous fashion. Digital images take a large amount of computer memory which causes slow transmission of data, especially through modems. This limits the practicality of high-resolution, large-file-size, digital images. Geometric and digital unsharpnesses are interrelated in many aspects. The factors that cause loss of image sharpness, also cause loss of resolution (Goaz and White, 1994). Recent advances in technology have created innovative imaging modalities which have necessitated the investigation of the combined effects of these two factors. This study was undertaken to explore the importance of geometric unsharpness and the need for higher resolutions in clinical dental radiography. We decided to use an endodontic file model, since it should be a good model for studying resolution due to the defined edge and small diameter of the files. Chapter 1 Advances in Imaging 1.1 - Conventional Radiography X rays were discovered by Wilhelm Conrad Rontgen, a German physicist, on November 8, 1895. There were several coincidences that led to this discovery. Rontgen noticed fluorescence of a barium platinocyanide screen that was located some distance from a discharge tube which he had covered wi th black paper. Because electrons could not escape the glass envelope of the tube to produce fluorescence and because the cardboard permitted no light to escape from the tube, he was quick to realize that some unknown type of ray was produced when the tube was energized. That is why he called them x rays (Curry III et al., 1990). Although his interests were mainly in the realm of physics, the capabilities of the new rays to penetrate soft tissue and form an image of underlying bones d id not escape him. He wrote the first article (Rontgen, 1896) two weeks after his discovery. The article is concise and contains an amazing collection of evidence regarding the properties of x rays. One of the illustrations in his paper was a radiograph of his wife's hand. He was awarded the first Nobel Prize in 1901. Dental radiographs were produced by others in Germany, America and England within the first few weeks of Rontgen's discovery (Curry III et al, 1990). When a beam of x-ray photons travels through an object, its intensity is reduced (attenuated) by absorption and scattering of the photons out of the primary beam. The pattern of the photons that exit the object carries information that can be recorded on an image receptor. Fluorescent screens were the first image receptors at the time of x ray discovery. Rontgen demonstrated that the resultant images could be also captured on photographic emulsions. The emulsions of the day were on glass plates. Two weeks after the announcement of Rontgen's discovery, the first dental radiograph was made requiring an exposure of 25 minutes. The plates were cut down by a dentist, wrapped in black paper and rubber dam material. The glass plates were extremely fragile and were uncomfortable for patients (Goaz and White, 1982). They were soon replaced by films. The early cellulose nitrate films were extremely flammable and dangerous to store. Cellulose tri-acetate films used later were not as flammable, although they would burn if held in a flame (Meredith and Massey, 1968). In 1913, the first commercial hand wrapped x-ray film developed for dentistry was introduced by Eastman Kodak Company. This film had a single emulsion (Goaz and White, 1982). In 1919 the first modern dental x-ray film packets became commercially available with an emulsion, called "Regular" (Kodak). This was very slow by current standards. Since that time, different film sizes and packaging have greatly enhanced the operation of dental radiology. In 1925 "Original Radia-tized" (Kodak) became available that was twice as fast as the earlier ones. Kodak "Original Ultra-speed" was available by 1941 followed by "Improved Radia-tized" (Kodak) in 1955. "Improved Ultra-speed" Kodak, arrived later in 1955 that was 6 times faster than improved radia-tized. In early 1981, a still faster film, "Ektaspeed" by Kodak became available (Richards and Colquitt, 1981). Kodak Ektaspeed Plus was introduced in 1994 with the same speed as Ektaspeed, but improved contrast (Price, 1995). The speed groups of intraoral dental radiographic films are classified in Table 1 (ADA, 1970). Speed Group Speed Range (In Reciprocal Rontgens) A 1.5-3.0 B 3.0 - 6.0 C 6.0 -12.0 D 12.0 - 24.0 E 24.0 - 48.0 F 48.0 - 96.0 Table 1. Speed groups for intraoral dental films. As Table 1 shows, each group of films is twice as fast as the previous one. Kodak Improved Ultra-speed, 1955, is categorized as D speed and Kodak Ektaspeed, 1981, is in the group of E speed films, that is twice as fast as D speed films. If we give an arbitrary factor of 1 for the amount of x rays required to make an exposure of Kodak's original dental film in 1919 (Regular Kodak), the exposure factors for other intraoral Kodak films are as the following in Table 2 (Richards and Colquitt, 1981). F i l m Exposure Factor Year Regular 1.0 1919 Radia-tized 0.5 1920s Original Ultra-speed 0.25 1941 Improved Radia-tized 0.25 1955 Improved Ultra-speed(D) 0.04 1955 Ektaspeed(E) 0.02 1981 Ektaspeed Plus ( E + ) 0.02 1994 Table 2. History of Kodak intraoral dental film. The high speed films in use today have considerably reduced the amount of patient and the operator radiation exposure. In the 1970's, a relatively new method for recording images without fi lm was developed that was called "Xeroradiography" (Goaz and White, 1982). This method was based on the same principle as a photocopier. A plate, which contained a photoconductor, charged wi th static electricity, was exposed to x rays. X rays caused the photoconductor layer to lose its charge in an amount corresponding to the intensity of the x-ray beam. The plate was processed in a device in which charged dry toner particles were attracted to the plate and transferred to a sheet of paper (Curry III et al., 1990). Xeroradiography had attractive features such as pronounced edge enhancement that showed greater details of soft tissue structures, a choice of positive and negative displays and wide range of exposures (greater latitude) (Goaz and White, 1982). It required less radiation exposure for intraoral radiography compared wi th D-speed film (Graft et al., 1979). In addition, it d id not require silver halide containing emulsions and the photoconductor plates were reusable. Processing of the image was also faster, more convenient, and the images could be viewed by reflected light. The disadvantage was, because of the edge enhancement, many false diagnoses of caries under restorations were being made. This stimulated Xerox Company to produce a dental xeroradiography system that used a l iquid toner, instead of the dry toner in late 1970's, which reduced the edge enhancement. When Kodak introduced its E-speed fi lm in 1981, there was little advantage in using xeroradiography especially wi th no edge enhancement. The large radiation dose required for extraoral radiography (45 times more than conventional screen/film combinations) was another major drawback in this system. Lack of recent dental literature referring to this method is evidence of failure to hold a place in the market. 1.2 - Digital Radiography In the field of oral radiology, there has always been increased interest in alternative imaging systems other than conventional films in order to reduce patients' radiation exposure and increase convenience and image quality. Computer technology, as we know it today, is the most significant advance in diagnostic imaging. Digital radiography has developed to its present state by using digital electronic devices. It can be either direct (direct sensor systems) or indirect (scanning conventional films). Conventional fi lm produces an analog image that contains continuous data in a steadily graded range of different grey shades. Digital images, on the other hand, contain discrete information that is based on the binary number system, in which two digits (0 and 1) are used to represent information. These two characters are called "bits", for Binary digiT (Goaz and White, 1994). Computer language is in specific units called "words". A common unit of information is 8 bits in length, called "byte". Every bit in an 8-bit word is either 0 or 1, which can give as o much as 2 (256) possible words or bytes in this language. Therefore an 8-bit system can represent 256 shades of grey ranging from 0 (white) to 255 (black) in value (Goaz and White, 1994). The pictures are made up of small areas called "pixels". Each pixel has potentially 256 grey shades. The resolution of a digital image depends on the number of pixels in a given area (pixels per inch, cm, mm, etc.). Common usage in computer language has tended to replace such units wi th "dots per inch" (dpi). In dental digital radiographic systems, an intraoral sensor is used in place of film and the image is displayed on a computer screen. The sensors that are being used at the moment are either charge-coupled devices (CCD) or photostimulable phosphor plates (PSP). In the systems using C C D , digital data are processed electronically and displayed on a computer screen almost instantaneously. The charge-coupled device is a semiconductor that can store charge on a signal from the outside and transfer it to a readout point (Curry III et al, 1990). It consists of a chip wi th an active area that is divided into a two-dimensional array of pixels. When either direct x-rays or light from an intensifying screen interacts wi th the pixels of a C C D , an electric charge is created that is stored by the pixels. Charges are removed electronically from each individual pixel, creating an analog output signal. This analog signal is then converted into discrete units by an analog digital converter ( A D C or digi t izer) , since computers function wi th digital information (Goaz and White, 1994). PSPs are rigid or flexible plates coated wi th a barium halide compound which absorbs radiation. When these plates are exposed to x-rays by a conventional x-ray equipment, they absorb some of the energy in the beam and store a portion of this energy in high energy traps to create a latent image. Dur ing processing, this latent image is scanned by a laser beam and the image can be viewed, typically from one half to several minutes. Digital systems have the advantages of: • Reduced radiation exposure. • N o processing required. • Computer stored Images. • Transmission of the images via electronic mail . • Image manipulation. The disadvantages are: • Small receptor size which may require multiple images to complete a survey. • In the case of C C D s , physical connection to a cable, and a rather bulky sensor, which is not easy to position in the mou th . • Image Quality is inferior compared to the film. • Initial cost is high. Chapter 2 Review of Literature 1 2 Studies that have used endodontic file models in radiography go back to 1969 when Vande Voorde et al. estimated endodontic working length to assess the right-angle paralleling diagnostic radiographs (Vande Voorde and Bjorndahl, 1969). This paper is valuable in the sense of historical approach to the method of measurement using endodontic files. But it does not show the application of it towards image quality. Okano et al. in 1983 used an endodontic file model to find the effective exposure level and diagnostic performance in endodontic radiography (Okano et al., 1983). They concluded that a significant reduction in exposure wou ld have a relatively small effect on the precision of endodontic distance measurements. Shearer et ah, in 1990, used a root canal model (without a file) to perform an in vitro comparison of RadioVisioGraphy (RVG, a CCD-based digital intraoral radiographic system) and conventional fi lm radiography (Shearer et al., 1990) and concluded that R V G may be equal to conventional fi lm radiography for the imaging of root canals in vitro. A similar study was performed in 1991, using endodontic files in the root canals, by the same authors and the results showed no significant difference between conventional f i lm and enhanced R V G images (Shearer et al., 1991). The enhancement facility for estimation of root canal treatment, when ut i l iz ing R V G , was suggested. Another study in 1991 by Cox et al. was performed to evaluate examiner accuracy by eyesight alone in making proper file length adjustments from radiographs (Cox et ah, 1991). They concluded there is more than 80% clinical accuracy when file adjustments are wi th in + 2 mm. Sanderink et al. , in 1994, used an endodontic file model to compare digital systems available at the time and E-speed dental x-ray film, and concluded that R V G in normal mode and Sens-A-Ray systems were comparable wi th conventional radiography in determining root canal length wi th size 15 files (Sanderink et ah, 1994). However, all systems were inferior to fi lm when size 10 files were used. Leddy et al. , in 1994, performed an in vitro investigation to determine the accuracy of endodontic file length adjustments using R V G images approximately two times the size of the tooth, compared wi th conventional radiographs (Leddy et ah, 1994). Results showed no significant difference in making accurate file length adjustments using conventional radiography versus R V G . In 1995, Ellingsen et al. performed an in vitro study to determine whether the R V G imaging system offered any advantages over conventional radiographs, viewed wi th magnification, for the identification of endodontic file tips in relation to the radiographic apex (Ellingsen et al., 1995). Results showed that R V G wi th zoom in the negative to positive mode was statistically equivalent to D-speed radiographs and superior to E-speed radiographs. The standard zoom was also superior to E-speed radiographs. D-speed radiographs were statistically superior to E-speed radiographs, being judged better than E-speed 90% of the time. They also performed an in vivo evaluation in which the results showed the D-speed radiographs were superior to all R V G mode images used in the study (Ellingsen et ah, 1995). D-speeds also provided superior recognition of small file tips when compared to E-speed radiographs 100% of the time. Four of the R V G images were equivalent to E-speed radiographs. Scarfe et ah, 1995, studied the detection of accessory lateral canals using R V G digital system and E-speed film (Scarfe et ah, 1995). They concluded that R V G images were slightly more sensitive than conventional fi lm images but the diagnostic accuracy of both systems in detection of lateral accessory canals was low. In 1996, Borg et ah, used an endodontic file model to assess the image quality of PSP systems and concluded that Digora intraoral image plate system provided reliable endodontic measurements even at very low exposures (Borg and Grondahl, 1996). Velders et ah, 1996, used an endodontic file model to compare the image quality of two digital sensor systems to determine the effect of dose reduction (Velders et ah, 1996). They concluded that wi th two digital sensor systems tested (Sidexis and 1 4 Digora), a dose reduction of approximately 95% compared wi th Ektaspeed films is possible to determine the length of a premolar root and size 20 and 25 files. Another study in 1996, by Garlock et al, used an endodontic file model to compare R V G - P C i and E-speed fi lm in file length determination wi th known variation in k V p and instrument orientation (Garlock et al, 1996). They concluded that R V G - P C i is particularly sensitive to overexposure, resulting in apparent shortening of the instrument and the proprietary software supplied wi th the R V G - P C i was not sufficiently accurate for endodontic assessment. Conover et al, 1996, compared the linear measurements made from a PSP system and conventional dental radiographs and found no significant difference (Conover et al, 1996). A study in 1997, by Versteeg K H et al, was performed to compare estimations of distances from the tip of an endodontic file to the radiographic apex of teeth on digital images (that are much larger than conventional radiographs) and conventional radiographs in order to determine if observers get adjusted to larger images than they are accustomed to (Versteeg et al, 1997). The conclusion was that the estimates of distances on digital images were comparable wi th or even better than those of conventional radiographs. Versteeg C H et al, 1997, studied the effect of altering image size in digital image systems on diagnostic quality, using an endodontic file model, and concluded that relevant diagnostic information may be lost when images are reduced in size (Versteeg et al, 1997). Another study by Versteeg et al , in 1998, using the endodontic file model, was performed to analyze the effect of reduction in size of diagnostic images and concluded that it may cause less detectability as wel l as loss of diagnostic information (Versteeg et al, 1998). Cederberg et al, 1998, used an endodontic file model in vivo to compare the difference in interpretation of the position of endodontic file tips between storage phosphor digital system (PSP) and radiographic film (Cederberg et al, 1998). The results showed smaller difference between file tips and root apex in digital images 1 5 than in fi lm, which suggested the former technique was more accurate to assess trial file length. Exposure levels in digital systems were explored in a number of studies. Goshima et al, 1996, evaluated image density (pixel values) and image contrast, due to variation in beam energy (kVp), for the Sens-A-Ray intraoral radiography sensor (Goshima et al, 1996). They concluded that wi th the C C D receptor, being more sensitive to x-ray photons of relatively low k V p , it was possible to use low k V p techniques in Sens-A-Ray without increasing the entrance dosage. A study by Farman TT et al, 1996, compared charge-coupled and storage phosphor sensors wi th Kodak Ektaspeed Plus films for determining density versus exposure and found that the sensor types are unlikely to affect differentiation between resin composites and dental enamel or recurrent caries (Farman et a/., 1996). Hayakawa et al, 1996, studied the optimum exposure at various tube voltage settings for computed digital radiography (CCD-based system) and concluded that was a fast system capable of operating at a wide range of k V p settings (Hayakawa et al, 1996). In 1997, Hayakawa et al. concluded that RVG-S provided greater dose savings than Sens-A-Ray (Hayakawa et al, 1997). Hildebolt et al, 1997, studied the response of PSP and digitized Kodak Ektaspeed Plus film to small variation in x-ray exposure and found that there was a significant direct linear relationship between exposure and digital grey-scale values for PSP images but not for digital images of Ektaspeed Plus films (Hildebolt et al, 1997). Scarfe et al, 1997, studied the tissue radiation dosage using R V G - S wi th and without the niobium filtration and found 31 - 39% reduction in average skin entrance wi th standard aluminum filter and 51 - 61% with the addition of niobium to attenuate the beam (Scarfe et al, 1997). However, adding niobium filtration resulted in increased dosage to the deeper soft tissues such as the thyroid gland. A study by Huda et al, 1997, compared the imaging performance of a PSP system and E-speed film and concluded, at the same radiation exposure, low contrast detectability of the PSP was superior to that of fi lm (Huda et al, 1997). They also mentioned that the l imit ing spatial resolution of the PSP was approximately 6.5 line pairs per millimeter ( lp/mm) whereas for fi lm it was in the range of 11 to 20 l p / m m . In 1998, Farman TT et al, compared C C D panoramic radiography and conventional film/screen for the effect of beam energy and radiation exposure and concluded that the digital panoramic receptor produces satisfactory images wi th saving of approximately 70% entry exposure compared wi th the conventional film/screen combination (Farman et al, 1998). Jones et al, 1998, estimated skin exposure as an indicator for comparing R V G versus Kodak Ektaspeed Plus dental radiography (Jones et al, 1998). They found that R V G systems, when combined wi th a standard conventional timer, actually generated greater radiation exposure than standard routine radiography. A study by Farman A G et al, 1998, compared the spatial resolution and dosage considerations for panoramic systems wi th standard film/screen and charge-coupled device and found that maximum spatial resolution of f i lm approached 5 l p / m m whereas wi th the C C D it was just above 4 l p / m m (Farman and Farman, 1998). The image layer was reduced slightly in wid th when using the C C D receptor, and the C C D resulted in skin exposure reduction exceeding 70%. Yoshiura et al, 1998, compared the effect of an additional scintillator layer on the psychophysical properties of a C C D detector for digital dental radiography (Yoshiura et al, 1998). It was concluded since the detector covered by a scintillator layer was more sensitive, it should be preferred for clinical practice since the dose to the patient is reduced. A number of studies have used periapical and periodontal lesion models to compare digital and conventional radiography. Kullendorff et al, in 1996, studied the diagnostic accuracy of direct digital radiography (DDR) versus E-speed film for detection of simulated periapical bone lesions and found no significant difference (Kullendorff et al, 1996). Kullendorff et al, 1997, performed a clinical study to compare D D R , wi th and without image processing, wi th that of conventional E-speed fi lm for the detection of bone lesions (Kullendorff et al, 1997). They concluded that film radiography was slightly better than D D R and the effect of image processing was not significant. Adosh et ah, 1997, compared the periodontal bone loss surgically and by R V G digital system and found that there was a significant difference and R V G measurements were marginally higher than surgical values (Adosh et al, 1997). A study by Borg et al., 1997, compared bone loss measurements on digital radiographs from a C C D (Sens-A-Ray) and a storage phosphor (Digora) system (Borg et al., 1997). It was concluded that digital radiographs are comparable to film-based radiographs for measurement of buccal bone loss, but wi th lower exposure especially wi th storage phosphor system. Farman TT et al., 1998, performed a study to evaluate the efficacy of a C C D detector for panoramic radiography compared to conventional film/screen radiographs using the same machine and patient population (Farman and Farman, 1998). F i l m radiographs marginally outperformed digital images for three criteria. The two modalities showed no difference for one criterion (periodontal bone status). It was concluded that digital images are clinically equivalent to conventional film/screen images for panoramic dental radiography. Another study by Farman A G et al, 1998, compared the efficacy of digital versus analog radiography, in vivo, for measuring periapical lesion dimensions and concluded that Visualix-2 images were preferable to film-based radiographs (Farman et al., 1998). Borg et ah, 1998, compared fi lm and digital radiography for detection of root resorption cavities and concluded all the systems were clinically acceptable but PSP system required considerably lower exposure than for both film and C C D systems (Borg et al, 1998). 1 8 Mistak et al, 1998, compared D D R versus conventional radiography for interpretation of artificial bone lesions and concluded there was no significant difference between D D R stored images, D D R transmitted images and conventional film (Mistak et al, 1998). Versteeg et al.,, 1998, compared the image quality of periapical radiography obtained from C C D image receptors wi th film and concluded periapical radiography wi th C C D sensor leads to more errors resulting in more retakes than conventional film (Versteeg et al., 1998). Reukers et al., 1998, studied in vivo and in vitro accuracy of digital reconstruction for assessment of apical root resorption and found it a reliable method (Reukers et al., 1998). Many studies have looked at image quality. In 1989, Kircos et al. studied the effect of developer temperature changes on image quality and concluded that temperature changes during developing have small effect on D and E-speed intraoral films but cause increase in base and fog for screen film systems that might compromise their quality (Kircos et al, 1989). In another study, they looked at image quality for intraoral films using rare earth filters and concluded that these improve image quality and reduce x-ray exposure (Kircos et al, 1989). Langen et al, in 1993, used fractured skulls to compare digital and conventional film/screen radiography and concluded both techniques were equally good (Langen et al, 1993). Kassebaum et al, in 1989, studied the spatial resolution for digitizing dental radiographs and concluded that digitization parameters of Kodak Ektaspan provided images wi th adequate diagnostic accuracy (Kassebaum et al, 1989). O h k i et al, 1994, studied the image quality of digitized conventional intraoral radiographs using a caries model and found that the images digitized by a drum-scanner had the best diagnostic accuracy, and sufficient diagnostic accuracy could be attained on low cost video monitors of personal computers (Ohki et al, 1994). A study by Scarfe et al, 1995, compared the measurement accuracy of two C C D based intraoral digital systems wi th analog fi lm and concluded that the mouse-driven computerized measurement 1 9 cursor and intraoral f i lm assessment using a ruler provided the most accurate and consistent measurement (Scarfe et al, 1995). Borg and Grondahl, 1996, performed a subjective evaluation for comparison of image quality in fi lm, C C D and PSP systems and concluded higher image quality was achieved over a much wider exposure range wi th PSP system than wi th either film or the C C D systems (Borg and Grondahl, 1996). L i m et al, 1996, compared digital intraoral radiographic images (Digora) wi th conventional E-speed film in terms of image quality, radiation dosage and diagnostic value (Lim et al, 1996). It was concluded that new imaging plates had slightly lower resolution power compared to E-speed film but gave favourable results in terms of contrast differentiation, and reduction of x-ray exposure was 53% compared to E-speed films. Wenzel et al, 1996, concluded that for caries diagnosis, a compression ratio of 1:12 can be justified without affecting accuracy and image quality (Wenzel et al, 1996). Studies regarding image processing include Wenzel and Hintze, 1993, who studied the effect of dentists' perceptions of the quality of digitally captured radiographs after the application of various image treatment filters, and found that the majority of dentists preferred the treated images over the original one (Wenzel and Hintze, 1993). Kullendorff and Nilsson, 1996, studied the effect of diagnostic accuracy of direct digital radiography in the detection of artificially made periapical lesions, after applying image procession, and found that the image processing was most effective when adjustment of contrast and brightness was used (Kullendorff and Nilsson, 1996). There was no significant difference when more complicated processing procedures were used. Moystad et al, 1996, compared the accuracy of approximal caries detection using enhanced and unenhanced PSP images and dental film in vitro (Moystad et al, 1996). It was concluded that enhancement improved detection of approximal caries compared wi th unenhanced images and Ektaspeed film. In 1997, Visser et al, studied the effect of image manipulation and found that the altered diagnostic content of the manipulated images was not detected by dentists, and suggested additional measures for data protection of digital radiographs (Visser and Kruger, 1997). Versteeg et al, 1998, studied the effect of calibration and automatic grey scale adjustment to detect simulated bone lesions using a PSP system and concluded that digital images should be calibrated for the highest exposure to optimize the wide latitude, and auto grey scale was not useful (Versteeg et al, 1998). Tyndal l et al, 1998, compared Kodak Ektaspeed Plus fi lm wi th Siemens Sidexis digital imaging system for caries detection in vitro (Tyndall et al, 1998). They found unenhanced digital Sidexis equivalent to fi lm but enhanced images were significantly inferior than unenhanced digital and fi lm images. A few studies have explored a new method of creating three-dimensional (3-D) radiography displays based on optical aperture theory know as tuned aperture computed tomography (TACT). Webber et al, in 1996, studied the comparison of film and C C D (2-D) and T A C T (3-D) for identifying the location of crestal defects around endosseous implants and found that T A C T performed significantly better than the other two methods (Webber et al, 1996). In 1997, Webber et al. introduced T A C T as a new 3-D method wi th a number of advantages over conventional plain film and tomographic imaging (Webber et al, 1997). It was concluded that, depending on the diagnostic task, T A C T performance is superior or comparable to conventional modalities. Tyndal l et al, 1997, compared T A C T wi th D D R and film for primary caries detection in vitro (Tyndall et al, 1997). They found that T A C T and film were not significantly different but significantly superior to D D R . In 1998, Nai r et al, studied the efficacy of T A C T images for the detection of recurrent caries in vitro and concluded that T A C T was significantly more effective than film or D D R (Nair et al, 1998). M o n d o u et al, in 1996, proposed a method for objective assessment of digital radiography intraoral systems wi th each other and wi th conventional radiographs (Mondou et al, 1996). Evaluation of digital subtraction radiography (DSR) has been investigated in some studies. E l lwood et al, in 1997, evaluated DSR using Digora radiographic imaging system and a semi-automated image processing software (Ellwood et al, 1997). It was concluded that this system provided adequate precision evaluation. A study by Yoshioka et al., 1997, explored the accuracy of in vitro quantitative digital subtraction for direct digital dental radiography using a commercially available image analysis system (Yoshioka et al, 1997). They concluded the quantification of digital subtraction using their method was sufficiently reliable for its application to clinical practice to be evaluated. Delano et al, 1998, studied DSR to establish an objective method for evaluating the treatment outcome of apical periodontitis and concluded it may be a useful tool in endodontic apical surgery assessment (Delano et al, 1998). Christgau et al, 1998, performed a study to determine the accuracy of DSR to detect small changes in calcium mass in alveolar bone adjacent to roots and concluded high accuracy of DSR (Christgau et al, 1998). In another study, they found very high correlation between the objective, quantitative assessment of subtle changes in alveolar bone by DSR and the true changes in bone thickness (Christgau et al, 1998). Many studies have used caries models to evaluate digital radiography systems. Wenzel et al, 1995, compared the diagnostic accuracy of C C D and PSP digital systems for caries detection in vitro and concluded they performed almost equally wel l and compressed images were as accurate as uncompressed ones (Wenzel et al, 1995). Tirrell et al, 1996, used chemically created lesions to compare digital and conventional radiography and found no significant difference other than, after 12 -24 hours, digital imaging demonstrated lesions significantly earlier than conventional radiography (Tirrell et al, 1996). Kang et al, 1996, compared observer differentiation of mechanically created defects versus normal proximal caries and concluded mechanical defects were 1.40 times more detectable than natural lesions and the lesion depth influenced the probability of correct identification (Kang et al, 1996). They also found that there was less discrimination between natural and artificial lesions, using digital images compared to fi lm. Nielsen et al, 1996, compared the PSP images wi th Kodak Ektaspeed Plus films for caries detection in primary molars in vitro and found no significant difference between the two modalities (Nielsen et al, 1996). Svanaes et al, 1996, used a caries model in vitro to study the effect of an intraoral PSP system compared wi th conventional radiography and found PSP images comparable to E-speed film (Svanaes et al, 1996). They also found that the magnified PSP images showed a significantly higher diagnostic accuracy than non magnified ones. Versteeg et al, 1997, compared PSP images wi th conventional f i lm for detection of approximal caries in vivo and found that caries depth in PSP images was underestimated compared wi th film-based images (Versteeg et al, 1997). Lawrence et al, 1996, performed a clinical study to evaluate the effect of fluoridation on approximal caries progression, using serial digitized bitewing images and conventional film (Lawrence et al, 1997). They concluded that the two radiographic methods were strongly correlated. Huysmans et al, 1997, studied the effect of exposure time on in vitro caries diagnosis using the Digora system (Huysmans et al, 1997). They concluded that the diagnostic performance was not impaired for the exposure times as short as 6% of E-speed fi lm exposure, but wi th reduction to 3% of E-speed film exposure, caries diagnosis may be impaired. Price and E r g u l , 1997, used a caries model in vitro to compare Kodak Ektaspeed Plus film wi th Sens-A-Ray (CCD-based) digital system and concluded that film was statistically superior to Sens-A-Ray but the magnitude was small (Price and Ergul , 1997). They also found that the effect of scattering medium was insignificant. White and Yoon, 1997, also compared the performance of digital and conventional images for detecting proximal surface caries in vitro and concluded the direct digital CCD-based system performed as wel l as E-speed film and the two systems were diagnostically comparable (White and Yoon, 1997). In a review by Versteeg et al, in 1997, it was concluded that digital imaging certainly has great potential especially wi th respect to diagnostic quality and automated image analysis (Versteeg et al, 1997). In 1998, Nai toh et al, used a caries model to evaluate the observer agreement in determining the depth of proximal caries, using CCD-based D D R and conventional film-based radiography (Naitoh et al, 1998). They found similar results wi th both modalities. A review by Whaites and Brown, in 1998, regarding an update on digital imaging, included quality and standards for radiography wi th the recommendations from the latest guidelines (Whaites and Brown, 1998). Another review was performed by Wenzel, 1998, regarding digital radiography and caries diagnosis (Wenzel, 1998). In this review, the unknown part of digital radiography is also pointed out. For instance there is no evidence that the number of retakes has been reduced or how many images are needed wi th the various C C D systems when compared wi th conventional film. It is not known how stable these systems are in daily clinical use. Controll ing the cross-infection is another concern in relation to processing the PSP and providing a barrier for C C D sensors. There is also little evidence to show how enhancement facilities are changing working practices or treatment decisions. 2 4 Chapter 3 Image Unsharpness 3.1 - Sharpness and Resolution Sharpness is the ability of the imaging system to define an edge. It measures how wel l the min imum details of an object are reproduced on the radiograph. In other words, it is a measure of how well a border between two areas of contrasting radiodensity is delineated (Goaz and White, 1994). Sharpness and contrast are closely related in the subjective response they produce. If contrast is high, a sharp edge can be easily seen, but if contrast is low, the sharp edge may be poorly visible. Unsharpness, or blur as it is often called, is the inability of the imaging system to record a sharp edge (Curry III et al, 1990). Resolution on the other hand, is the resolving power or the ability to record separate images that are placed very close together. A n imaging system may have the ability to record sharp edges but be less able to show fine details. Another system may resolve fine details but yet yield fuzzy unsharp edges. Sharpness and resolution are different, but they are related in the subjective response they produce (Curry III et al, 1990). Resolution is frequently measured by making a radiograph of a line pair plate (Figure 1). This consists of a thin sheet of lead wi th rectangular spaces arranged to give lines and gaps of decreasing, equal, width. That shown in the Figure has dimensions from 1 line pairs per millimeter ( lp /mm), in group # 1, to 10 l p / m m , in group # 18. In the case of the latter, there are 10 pairs of lines and spaces per millimeter, each line and space being 0.05 mm. Resolution of radiographic fi lm is in the region of 20 line pairs per m m whereas the resolution of a digital system, for example Digora, is reported to be 6.5 line pairs per m m (Huda et al, 1997). I 2 :< .1 5 « 7 II III III III III II III III III I I I I I I I I I I III 17 K i ir> M KI 12 | | |o a Figure 1. A radiograph of a line pair plate (Buckbee-Meers Inc., Minneapolis, M I , U S A ) on D-speed dental x-ray film. Groups represent increasing numbers of line pair per m m , ranging from 1 l p / m m (group 1) to 10 l p / m m (group 18). 3.2 - Causes of Unsharpness 3.2.1 - Geometric Unsharpness and its Relevance to Clinical Diagnosis The principles involved in the formation of radiographic images are similar to those in the production of shadows by light. The clearest shadows are produced wi th a small light source, a large distance between the source and the object and a small distance between the object and the screen. Conversely a larger source, smaller source to object distance and greater object to screen distance results in a more blurred image . The blur in the case of both light and radiation is called "geometric unsharpness". Geometric unsharpness is affected by two factors: the size of the source and the distances between the source and the object and the object and the receptor. The size of the source in the case of x rays is called the "Focal Spot". The image on a radiograph is a two-dimensional representation of a three-dimensional object. To obtain the maximum value from a radiograph, the clinician must mentally reconstruct, as far as possible, an accurate three-dimensional image of the anatomic structures from one or more of these two-dimensional images (radiographs). This task is greatly facilitated when the radiographs are of high quality. The principles of projection geometry describe the effect of focal spot size and its position (relative to the object and the film) on the quality of the resulting image (Goaz and White, 1994). When x rays are emitted from the focal spot on the target of an x-ray tube, they originate from all points over the actual target and radiate in all directions. Some of the rays from each of these point sources w i l l pass tangentially to the edge of the object, forming separate images of the object's edges. Because these rays originate from different points and travel in straight lines, their projections on the fi lm w i l l not be in exactly the same spot. As a result the image of the edge is not sharp and distinct but broad and fuzzy. This zone of unsharpness is the penumbra (Figure 2). Figure 2. Diagram to show the effects of source to object and object to film distance on penumbra (Geometric Unsharpness). 2 9 "Penumbra" (from Lat in ipene, meaning almost, and umbra, meaning shadow), often termed "edge gradient", is defined as the region of partial i l lumination that surrounds the "umbra" or complete shadow (Curry III et al., 1990). The blurring causes a loss in image clarity by reducing sharpness. There are three ways to minimize these effects and improve the quality of the radiographs: 1- Use as small and effective focal spot as practical (Figure 3). 2- Increase the distance between the focal spot and the object (Figure 4). 3- Decrease the distance between the object and the film (Figure 2) (Goaz and White, 1994). There are practical limits to all these issues. The focal spot must be of a finite size in order to dissipate the heat produced from electrons striking the target. Increasing the distance between the source and the object or using long cones in the x-ray units, w i l l l imit the practicality of handling the machines. There is also a limit of how close we can place the film to the teeth in the patient's mouth because of the anatomical limitations. In 1947, Fitzgerald mentioned that wi th the focal spot moved back from the film to a greater distance, the paths of the x rays tend to be nearly parallel and the distortion of the image by magnification is considerably lessened (Fitzgerald, 1947). This idea led to the use of long cones that in theory keeps the geometric unsharpness minimal . The concern is how relevant this is in clinical diagnosis objectively, that is one of the issues that w i l l be discussed here. 3.2.1.1 - Focal Spot The focal spot is the area on the target from which the x-rays originate and it is necessarily of finite size in order to aid the dissipation of heat produced when electrons strike the anode of the tube. A n x-ray machine consists of an evacuated glass envelope which has a cathode at one end and an anode at the other (Figure 5). 3 0 Figure 3. Diagram to show the effect of focal spot size on penumbra (Geometric Unsharpness). Figure 4. Diagram to show the effect of source to object distance on penumbra (Geometric Unsharpness). The cathode contains a filament which is electrically heated to provide a source of electrons. A cup-shaped depression ih the cathode focuses the electrons on to a target on the anode (focal spot) faced wi th tungsten (Figure 6). The tungsten target is embedded in a large block of copper (Figure 7). Copper is a good thermal conductor which dissipates heat from the target, reducing the risk of thermal damage which might lead to melting of the target surface. The purpose of the target is to convert the kinetic energy of the electrons into x-ray photons. Tungsten is used because it has high atomic number, high melting point, and low vapour pressure at the 32 working temperatures of an x-ray tube. Since the production of x rays is a rather inefficient process, w i th more than 99% of the electron kinetic energy converted to heat, there is an obvious requirement of high melting point (Goaz and White, 1994). Radiographic image quality is partly dependent on the size of the target. The sharpness of the radiographic image increases as the size of the focal spot, decreases and vice versa. A s the focal spot decreases in size, the heat generated per unit target area becomes greater. To take advantage of the smaller focal spot, and yet distribute the electrons over a larger surface, the target is placed at an angle (about 70 degrees) wi th respect to the electron beam. The effective focal spot (the projection of the focal spot perpendicular to the electron beam) w i l l be smaller than the actual size of the focal spot (Figure 8). Figure 9 shows the radiographic image of focal spot of the G E 100 x ray unit (the one used in our study), produced wi th a 300 u\m pinhole placed Figure 5. A dental x-ray tube. Figure 6. Anode (left) and cathode of an x-ray tube. The pattern of the electron bombardment on the target face can be seen. This forms the source of x rays (focal spot). The focusing cup around the filament of the cathode focuses the electron beam. 250 m m from the source. The source to receptor distance was 1000 mm. This produces an image of the focal spot size wi th a magnification of x3 (Figure 10). 3.2.1.2 - Magnification and Geometric Unsharpness While geometric unsharpness, as described earlier, is a result of the finite size of focal spot, and the source to fi lm and object to fi lm distances, magnification depends only on difference i n distances. When an object is placed i n the path of x-rays, because of the diverging beam, the shadow projected on the f i lm w i l l show some Figure 7. Target face showing focal spot, viewed perpendicular to the face. Figure 8. Target face showing focal spot, viewed perpendicular to the long axis of the x-ray tube. The effective size of the focal spot is smaller than the actual size. 2 m i n Figure 9. Image of the focal spot of the x ray unit used in this study (300 urn pinhole placed 250 m m from the source and the film 1000 m m from the source - Figure 10). degree of enlargement (Figure 11). The amount of magnification (M) can be defined as: size of the image j , * Q ~ size of the object In the clinical situation, the object within the patient's body might not be accessible for measurement, but since it is usually possible to determine the x-ray source to fi lm distance and the object-film distance, the magnification can be also measured as: source to film distance ^ ~ source to object distance 36 When the object is in contact wi th the film, there w i l l be no enlargement or a magnification of one. Distortion happens when there is unequal magnification of different parts of the same object. It is the result of the object not being directly in the central part of the x-ray beam, and also occures when the film is bent. Wi th a pin point focal spot, there is no geometric unsharpness, but there w i l l be magnification if the object is not in contact with the film. Since the x-ray units have finite focal spot size, the magnification resulting from difference in distance w i l l be accompanied by an increased geometric unsharpness. More geometric unsharpness correlates wi th more magnification of the image. Figure 10. Diagram of the relationship of source, pinholes, and film used to produce the focal spot image (shown in Figure 9). 37 Mm Object Figure 11. Diagram to show the effect of source to object (LI) and source to film distance (L2) on magnification. 3.2.2 - Film Unsharpness X-ray fi lm is photographic fi lm consisting of a radiation-sensitive emulsion that is coated on both sides of a transparent sheet of plastic, called the fi lm base (Curry III et ah, 1990). Dental f i lm emulsions are thicker than photographic emulsions in order to reduce the exposure. This increased thickness is possible because x rays penetrate more than light. Double coat emulsion also eliminates the tendency of the film to curl. Us ing emulsions wi th larger grain size w i l l further reduce the radiation exposure. However, in x-ray fi lm, the size of the silver halide grains limits the sharpness of the image (Goaz and White, 1994). The finer the grain size is, the sharper the image w i l l be. Slow speed films have fine grains and faster films have larger grains. Even in the case of fast, coarse grained emulsions used in dental x-ray films, this is a rarely significant cause of degradation of image quality and is completely negligible (Meredith and Massey, 1968). 3.2.3 - Parallax Unsharpness Parallax unsharpness arises because of double emulsion films. It is caused by the fact that the observer is looking at two patterns rather than a single one. Since the two images are separated from each other by the thickness of the film base, it is not possible to look straight on both images simultaneously. The result is the edges of the shadow do not overlie each other properly and unsharpness results. For the dry film the magnitude of the effect is negligible. In a wet film, however, because of the expansion of the gelatin layer, the separation of the two images is increased and the unsharpness can be noticeable. It is for this reason that the details in a dry radiograph are much more observable than a wet one. With the production of newer, thinner emulsions, which don't expand much during processing, this effect is considerably reduced (Meredith and Massey, 1968). 3.2.4 - Movement Unsharpness Movement of the film, object, or x-ray source during the exposure can also result in image unsharpness. Film or object movement result in multiple or biurred images. Movement of x-ray source, in effect, enlarges the size of the focal spot and decreases image clarity. The use of higher mA and kVp values and correspondingly shorter exposure times can help minimizing this problem (Goaz and White, 1994). 3.2.5 - Digital Unsharpness Digital unsharpness is due to the resolution of the imaging system. There are two types of digital unsharpnesses, direct and indirect. Unsharpness due to direct digital capture is limitation of the resolution of the digital receptors in direct digital systems. Digital unsharpness can also occur in indirect digital images (scanned conventional films) which depends on the size of the file created during scanning. Smaller files that have less pixels per inch or dots per inch (dpi), have less resolution than larger ones wi th more pixels per inch. Therefore, smaller file sizes have more digital unsharpness (e.g. 75, 150 versus 300 and 600 dpi). The higher the number of pixels, the finer the resolution and the closer we approach to the original image. Optimally, a properly displayed digital image w i l l be identical to the original to an observer. 40 3.3 - An Experimental Design to Determine the Position of the Focal Spot Based on a previous study (Price and McDonnell, 1990), we knew the approximate distance from the actual focal spot to the emission point of the x-ray head, in the x-ray unit we were using. This experiment was designed in order to obtain its position more accurately. A pewter plate with a grid of 300 um holes at 5 mm centres was used for this purpose (Figure 12). To standardize the relationships of the x-ray source, object, and film a "test column", consisting of three vertical steel rods and four horizontal aluminum platforms, was used (Figure 13). Two of the platforms were adjustable, and their positions could be established, by a Vernier scale and a vertical steel ruler, with an accuracy of 0.1 mm. A n occlusal size Ektaspeed Plus Kodak film was placed 900 mm from zero on the scale of the test column. A General Electric GE 100 x-ray unit (General Electric Company, Milwaukee, WI, USA) was secured in a reproducible position at the top of the column (Figure 14), The first radiograph was produced with the pewter plate placed directly on the film in order to obtain an image with no magnification. The exposure factors were 90 kVp, 15 mA, 3.5 mm aluminum filtration, and 2 seconds. The second set of radiographs was taken with the pewter plate at a fixed position, on the top platform, at 200 mm from zero on the scale or approximately 300 mm from the source (source to zero on the scale, approximately 100 mm). The films were placed on the second platform at distances of 400, 450, 500, 550 and 600mm from zero on the scale or approximately 600mm ± 50 and 600mm ± 100 from the x-ray source (focal spot). This produced increased magnification of approximately 2 (source to film distance 2 times larger than source to object distance (Figure 15). The exposure factors were 90 kVp, 15 mA, 3.5 mm aluminum filtration and 1 second. The radiographs were processed using an AT 2000 processor (Air Technique Inc., Hicksville, NY, USA) at 28 °C. Figure 12. Pewter plate wi th the grid of 300 urn pinholes. The 6 images were scanned wi th a Hewlett Packard 3c flatbed scanner and a 4c transparency adapter (Hewlett Packard Company, Greeley, C O , U S A ) as 8-bit grey scale images at resolutions of 600 dots per inch (dpi) and imported into N I H Image Program (Wayne Rasband, National Institute of Health, U S A ) . Us ing the facilities of the program, the profile plots of the selection that was made from the horizontal central row of the pinholes for all the images were obtained (Figure 16). Plot profile generates a density profile plot based on the rectangular selection. The wid th of the plot is equal to the wid th of the selection and each point i n the plot represents the average grey value of the pixels in the corresponding column in the selection. The x coordinates of the highest points in plots or the darkest part of the image were derived. Using Cricket Graph Program (Rafferty & Nor l ing , P A , U S A ) a graph consisting of number of pinholes in the x-axis, and the distances (x coordinates) in the y-axis for each image was obtained (Figure 17). A l l the graphs had r > 0.999. The slope or the gradient of the regression lines was the mean distance between the Figure 13. Test column used for standardizing the relationships of the x-ray source, object, and fi lm. pinholes. In the image wi th the magnification of 1 (pewter plate directly on the film and 900 m m distance of film below zero on scale), the gradient or average distance between the pinholes was 4.978 m m for the central horizontal row of the pewter plate. The gradients of the central row of the other images, taken wi th different distances from the source, were also obtained (Table 3). The data were entered into 43 an Excel spreadsheet (Excel, Microsoft Corporation, Redmond, W A , U S A ) . The magnifications were obtained by dividing the gradients (distances between the pinholes in the images) by 4.978 (actual distance between the pinholes in the image wi th the magnification of 1). The focal spot distance was derived from the following relationships: x + y Magnification = X + 2QO _ y - 200M x ~ M - 1 y (mm) *~ Gradient x (mm) 400 >0.999 8.2745 102.017 450 >0.999 ' 9.112 101.040 500 >0.999 9.9189 102.253 550 >0.999 11.587 101.286 Table 3. Gradients of the central row images of the pewter plate. x = Distance from focal spot to zero on the scale in the G E 100 x-ray unit y = Distance from zero on the test column scale to the film M = Magnification The mean distance from the focal spot was derived: x = 101.49 m m Figure 14. X-ray unit secured in a reproducible position at the top of the test c o l u m n . Focal spot X-ray Head Test Column / / / / / / / / / / / / / / / / / / / / / / / / / / / / sssssss Pinhole Object (Pewter Plate) Film / / / / / / / / / /sssss/ss SSSSSSJ/S SSSSSSJSS j II IIP / / / / / / / / / / / • / • ••••I ••••••• • • • • • • • • • / • • • • / / / / / / / ••••••••••••••••••••••••••• ••••••••••••••••••••••••••• ••••••••••••••••••••••••••• ••••••••••••••••••••••••••• ••••••••••••••••••••••••••• • • • • • . • • • • • . / / / / / , • • • • • , • • • • • . / / / / / , • • • • • Ruler •Zero 200 mm 500 mm Figure 15. Diagram to show the relationship between x-ray source, pewter plate, and f i lm, on the test column. Distance of pinhole object (Pewter Plate) below zero on scale: 200 mm. Distance of film below zero on scale: 400, 450, 500, 550, 600 m m . 46 Plot 220.83 191 -+-mm 70.27 Info X:34.37mm <203) V:218.83 Figure 16. Scanned image of pewter plate (Figure 12) showing focal spot images. Profile plot of one row of focal spot images ( N I H Image). 50 0 2 4 6 8 1 0 Pinhole Number Figure 17. Pinhole distances (y-axis) plotted against number of pinholes (x-axis) from the central row of the focal spot images (Figure 16). 4 8 Chapter 4 Endodontic File Study Model 4 9 4.1 - Pilot Study A pilot study was designed to obtain an overview of what would be explored and to plan the definitive study. A paper outlining this study was published, (Appendix-page 125), (Price and Radan, 1997). An Endodontic File Model to Compare Geometric and Digital Unsharpnesses in Dental Radiography a - Goal The goal was to develop an endodontic file model for comparison of the effects of geometric and digital unsharpnesses. b - Study Design Endodontic files, of sizes 8,10, and 15, were placed in the root canals of extracted teeth and radiographed, with a GE 100 x-ray unit, at magnification factors of 1.00, 1.14, and 1.33. Each radiograph was digitized, using a flatbed scanner, at resolutions of 150, 300, and 600 dpi. Twelve observers, in 4 groups of lay, senior dental students, general dentists and specialists, identified the positions of the end of the files and the end of the roots on the digital images. Coordinates were used to obtain differences in distances, with means and standard errors. A silver standard was used to obtain the differences in distances. This was derived from the measurements of the best quality images (600 dpi and magnification of 1). Analyses of variance were used to demonstrate significant differences. 50 c - Results N o significant differences were found between resolutions of 300 and 600 dpi , nor between the geometric unsharpnesses attending magnification factors of 1.00 and 1.14. Significant differences were found between geometric unsharpnesses with magnification factors of 1.33 and the others for all file sizes, and between resolutions of 150 dpi and the others for file sizes 8 and 10. In the case of file size 15, the mean differences in distance were all wi thin 0.2 mm. d - Conclusion The endodontic file model is useful in the comparison of geometric and digital unsharpnesses. Determinations of the position of file size 15 was little influenced by the image qualities used in this study. Little clinically relevant information is gained by scanning radiographs wi th resolutions greater than 300 dpi . 5 1 4.2 - Definitive Study The definitive study was designed based on the experience gained from the pilot study, wi th extended ranges of file sizes, file positions and geometric and digital unsharpnesses. The pilot study showed that the endodontic file model was reliable for studying the resolution because of the fine diameter and wel l defined outline of the images. It was decided to add one more file (file size 6) to the definitive study, since that is the finest endodontic file available. Therefore there were endodontic files of sizes: 6, 8, 10 and 15. The files selected have diameters ranged from 0.06 m m (file 6) to 0.15 m m (file 15), Table 4. File # Diameter at the tip (mm) 6 0.06 8 0.08 10 0.10 15 0.15 Table 4. Sizes of endodontic K-type files. File #15 is the min imum size file for accurate determination of root canal length (Glassman and Serota, 1997), which was the main subject in the study for investigating the effects of digital and geometric unsharpness. For the digital unsharpnesses, it was also decided to extend the range of resolutions to include 75 dpi in addition to 150, 300 and 600 dpi used previously. The 600 dpi resolution takes 4 times as much computer memory compared to 300 dpi . Table 5 shows the digital file sizes for a 256 grey shade image, one inch square. 52 Resolution { d p Matrix size (pixel; s Digital file size (kB) 75 5,625 5.5 150 22,500 22 300 90,000 88 600 360,000 352 Table 5. Digital file sizes for a 256 grey shade image, one inch square. There was no significant difference between the values recorded by the different observer types i n the pilot study. Therefore, it was decided to use only two groups of observers, 5 general dentists and 5 senior dental students. 4.2.1 - Goals The objective was to find the importance of the effects of geometric unsharpness and the need for higher resolutions in clinical dental radiography. The nu l l hypotheses comprised: • Files smaller than size 15 w i l l be increasingly difficult to quantify radiographically. • The presence of geometric unsharpness w i l l detract from image quality and provide less diagnostic information. • Increased digital unsharpness (e.g. 300 dpi compared wi th 600 dpi) w i l l cause poorer image qualities resulting in less radiographic information. 4.2.2 - Materials and Methods 4.2.2.1 - Test Objects A n extracted human lower incisor was used as a test object to enable determination of the position of the ends of endodontic files relative to the apex of the tooth. The decision was made not to use tissue equivalent, in order to reduce the effect of an additional noise. The tooth was sectioned transversely at the cemento-enamel junction (CEJ) using a low speed wafering saw wi th a circular diamond blade (Isomet by Buehler, Lake Bluff, IL, USA) . This provided a perpendicular surface to facilitate reproducible registration of the files in the root canal (Figure 18). The root was attached to a sheet of 1 m m thick acrylic resin, using autopolymerizing acrylic resin (Figure 19). Figure 18. V iew of the root attached to the jig showing the sectioned surface. Figure 19. Root attached to a part of the jig wi th self autopolymerizing acrylic resin. Endodontic K-type files of sizes 6, 8,10 and 15 were inserted inside the root canal, starting wi th the smallest file. The files were placed in 3 positions and radiographs were taken before proceeding to the next larger file. There were 3 positions in 2 mm increments consisted of 1 m m through the apex, 1 m m short of the apex and 3 m m short of the apex. The root canal was instrumented wi th the first file, size 6, until the file protruded beyond the apical foramen of the tooth. A low-power light microscope was used to determine the first position of each file. A n eyepiece graticule was calibrated wi th a stage micrometer (0.01 m m accuracy) and the file was manipulated unti l it extended 1 millimeter beyond the apex (Figure 20). The file was fixed in this position by means of a stop fabricated from a disc of vacuum-formed splint material and autopolymerizing acrylic resin (Figure 21). The disc rested against the sectioned surface of the tooth. A jig was constructed from a sheet of 2 m m thick vacuum-formed splint material, wi th an aperture of 3 cm diameter to reduce the attenuation of the x-ray beam. Provision was made for the acrylic sheet attached to the tooth to be registered on this jig (Figure 22). Two stainless steel hooks were attached to the sheet and a piece of 0.7 m m orthodontic wire served as a spring to maintain sufficient pressure on the file to keep the stop in contact wi th the tooth (Figure 22). Figure 20. Photomicrograph of apex of root wi th protruding endodontic file. Superimposed image of stage micrometer (1mm). Figure 21. Endodontic file with 2 m m stops. Figure 22. Jig used to maintain the endodontic file position in the root. For determining the other two positions, two additional stops, each 2 m m thick, were prepared from vacuum-formed splint material. The thickness of this material was obtained using a fine abrasive sheet, and was established wi th a micrometer. These stops permitted the file to be placed at 1 m m and 3 m m short of the apex of the tooth. Radiographs were produced using the number 6 file before instrumentation wi th the other three sizes in ascending order. The three positions were registered for all files in the manner described above. 4.2.2.2 - Radiography A test column, described previously (chapter 3 page 40) was used to standardize the relationship of the x-ray source, object and the film. Two of the platforms were adjustable, and their positions could be established, by a Vernier scale and a vertical steel ruler, wi th the accuracy of 0.1 mm. A General Electric G E 100 x-ray unit (described in page 40) was secured in a reproducible position at the. top of the column (Figure 14, page 44). The accurate position of the focal spot had been previously determined by triangulation, using images of a pewter plate wi th a grid of 300 um holes at 5 m m centres (chapter 3). a - X-Ray Exposure Ektaspeed Plus films (Kodak Canada Inc., Toronto, Ontario, Canada), size 2, were placed 1000 m m from the focal spot on the aluminum platform of the test column. Provisions were made to register the jig in a reproducible position on the platform of the test column, using the three vertical rods (Figure 23). One series of radiographs was produced wi th the root as close as possible to the films, wi th approximately 5 m m file to fi lm distance, producing a magnification of 1.005 and a negligible geometric unsharpness. The distance between the root and the films was Figure 23. Position of jig on the test column. 59 altered to give increased magnification and greater geometric unsharpness. File to fi lm distances were obtained wi th the known magnification from the following equation: SFD FFD = SFD -Where : F F D = File to film distance SFD = Source to film distance M = Magnification File to film distances of 91, 167, and 231 m m were used to give magnifications of 1.2 and 1.3. In clinical dental radiography, magnification is typically in the region 1.1. Exposure factors were 90 k V p , 15 m A , 3.5 aluminum total filtration and 36 cycles / 60th of a second). •1, of (36 b - Processing The films were processed in random order, by a sequence generated in Microsoft Excel Program. This randomization was performed to avoid any k ind of systematic error caused by changes in the temperature or the freshness of the solution during the processing. In other words, because of the delay between the exposure and processing for each film, if the radiographs were processed by the same sequence of exposure, there wou ld be a smaller delay for the films that were exposed later than the ones exposed earlier. The change of temperature and freshness of the solution would apply to the same degree that was mentioned earlier. Processing was performed two and a half hours after exposure, which is wi th in the standards proposed by International Standards Organization (ISO), to minimize any effects due to latent-image instability. ISO recommends that the films shall be processed not less than 30 minutes nor more than 8 hours after exposure (ISO, 1981). Processing was performed by an A T 2000 processor (Air Technique Inc., Hicksvi l le , N Y , USA) at 26-28 °C for a total processing time of 5 minutes and 20 seconds, using Kodak Readymatic chemistry. Two films, consisting of an unexposed f i lm and one exposed wi th a penetrameter, (Figure 24) were used at the start and end of processing and between each batch of 12 in order to confirm consistency of the processing. The density values of these test objects, including 4 different areas of the unexposed films (base plus fog) along with 6 areas of the films exposed wi th the penetrameter, were measured after processing, using a calibrated Macbeth transmission densitometer, T D 502 (Kollmorgen Corporation, Newburgh, N Y , U S A ) , wi th a 1 m m diameter circular aperture. The means and standard deviations of the density values for all the test films were obtained. Figure 24. Processed unexposed film and radiograph of penetrameter. 4.2.2.3 - Digitizing The 48 radiographs (4 files x 3 positions x 4 magnifications) along wi th the radiographs of the penetrameter were scanned wi th a Hewlett Packard 3c flatbed scanner and a 4c transparency adapter (Hewlett Packard Company, Greeley, C O , USA) as 8-bit grey scale images at resolutions of 75,150, 300 and 600 dpi . The process gave rise to images wi th different sizes such that 75 dpi image was half the size of 150 dpi , in each plane (Table 5, page 52). There were 12 radiographs for each file size. They were organized and taped together as 3 x 4 batches of radiographs. The radiographs were oriented that the direction of the endodontic files was perpendicular to the direction of the scanning, since the precision of the scanner was thought to be more reliable this way. The digitized images were opened in Colorlt Program (MicroFrontier, Inc., Des Moines, IA, USA) and their sizes were enlarged, i.e., 75 x 8, 150 x 4 and 300 x 2, in order to be standardized and have the same size for comparison. Then the images were rotated 90 degrees (because of the orientation of the scanning mentioned previously)/ Then every 3 images of 3 different positions were selected and copied into an area of 80 m m width and 25 m m height and captured in the computer memory, e.g. file 15, 600 dpi , magnification 1 (Figure 25). a - Randomization The data of 192 images, consisting 4 file sizes, 4 digital unsharpnesses, 4 magnifications and 3 positions, were entered into an Excel Program worksheet (Excel, Microsoft Corporation, Redmond, W A , U S A ) . The sequence of the images was first randomized and then organized in 48 groups of four images. Figure 25. A n example of 3 positions of file size 15, wi th the resolution of 600 dpi and magnification 1 (1 m m through, 1 mm short and 3 m m short of the apex). b - Image Presentation The images in each group of four, selected after randomization, were opened in N I H Image Program (Wayne Rasband, National Institutes of Health, U S A ) . Us ing the facilities of the program, a rectangular selection of 200 pixels wid th and 400 pixels height was placed around the root surface (Figure 26). A matte acetate sheet was then placed on the screen and the outline of the root and the selection were traced to serve as a template. This template was used for all the images to help orientate the root in the same way relative to the screen. This selection was then copied into an area of 800 pixels wid th x 400 pixels height, which could incorporate 4 images of each group, using the template. This procedure was performed for all the 48 groups of images. Figure 26. Outline of template superimposed on scanned image of root to enable reproducible positioning of the images. The images were then opened in PowerPoint Program (Microsoft Corporation, Redmond, W A , U S A ) in a black background wi th each group of four fil l ing the computer screen (230 x 170 m m approximately), representing a magnification of 6 .8. c - Visual Analog Scale A visual analog scale was used for the subjective evaluation of image quality. Aldus Super Paint Program (Silicon Beach Software, Inc., San Diego, C A , USA) was used to make a graded bar. A horizontal line of 3 pixels width and 2 inches long was drawn with ticks at each end. The two ends were designated wi th 0 and 10. Zero meant that the image quality was poor and not clear and 10 meant that the image was perfectly sharp and clear. A copy of this visual analog scale was pasted above eacfj image in every group of four (Figure 27). 4.2.2.4 - Observation and Interpretation a - Selection of Observers Since there was no significant difference among different types of observers in the pilot study, it was decided to use only 2 types of observers. Ten observers were selected consisted of 2 groups of 5 dentists and 5 senior dental students. Each observer viewed the images on the computer screen in a room wi th subdued lighting. N o time limits were set, and observers were allowed to vary their distance from the screen, but not to adjust the contrast, brightness, nor any software features. They were given written instructions in which they were told the objective of the study. They were required to mark the end of the file and the end of the root, using the marking facility of the PowerPoint Program. Addit ional ly they recorded, as a 65 Figure 27. A n example of the slides with root containing endodontic files and visual analog scale. mark on the visual analog scale, their subjective impression of the quality of the images (Figure 28). The marked images were captured to memory. b - First Reading A t the first reading, all observers viewed 192 images without interruption. The actual time spent to view the whole images ranged from 37 to 65 minutes with the mean of 48.7 minutes. 66 6 7 c - Second Reading For the second reading, it was decided to select a random sample of 24 images instead of having the observers to review the whole procedure. The 192 images were randomized for the second time and the first 24 images were copied in six groups of four and presented to all observers under the same protocol used previously. The second reading was also performed in one session within at least one week interval. The mean time spent to view the 24 images on this occasion was 5.9 minutes and the range was from 4 to 8 minutes. d - Measurement The marked images were opened in N I H Image Program. The marks at the end of the root and the end of the files, identified by the observers, were four pixels in size, each labeled wi th a pair of x and y coordinates (Figure 29). We used the coordinates of the upper left pixel ( x l , y l ) for all the marks. The x coordinate of each mark on the visual analog scales was also obtained. Coordinates describe the location of a pixel wi th in the image. The marks at the end of the root and the end of the files represented the two corners of a right angle triangle and their coordinates were used to obtain the hypotenuse between them, i.e. the shortest distance between the two points. The distances were measured to the nearest pixel and subsequently converted to millimeters. In order to calibrate these measurements, a millimeter ruler was scanned at 600 dpi . The scanned ruler was opened in PowerPoint Program and a photograph was taken from the screen. This photograph was opened in N I H Image and a 10 millimeter length was measured in pixels (Figure 30). There were 18.7 pixels in each millimeter of the scale and this factor was used for converting the 1 measurements. The accuracy of measurements was 0.05 m m (Toy mm). 68 C o c o ^ .S £ OH TJ s QJ 5 ca o o M 01 -0 e 0) ca CA 0) 4-1 ca g TJ >H o o o D o ^ o OH X cn 2 ce 'rt ca 05 M CU > 0) cfl o o OH ca H o -4-1 o OH G QJ QJ u ca O QJ > QJ bO J-H QJ </> QJ 00 ca fi • P H QJ 69 Measured Distance: Known Distance: Pinel Aspect Ratio: 187J.0 Pinels 10.00 1.0000 Units: Millimeters •{ Scale: 18.700 pinels per mm Cancel) | OK I Figure 30. Measured distance of 10 mm on the scanned image of a ruler. Dialog box identifying the scale (NIH Image). 70 Chapter 5 Analysis and Results 71 5.1 - Data Analysis For analyzing the data, the observers' measurements had to be compared wi th the real distances between the file and the end of the root as our gold standard. The positions of the files in the root were obtained as precisely as possible during the experiment (described in chapter 4). The data of the observers' measurements were entered into a spreadsheet in Microsoft Excel Program and the slides order, which had been randomized, was sorted back to the original order. The errors of the observers' measurements were called differences in distance. These values were obtained by subtracting the measured distances from the gold standard. The gold standard was the actual distance from the end of the file to the end of the root, in the 3 positions, as described in chapter 4. In the case of those images obtained wi th magnifications of 1.1, 1.2 and 1.3, the values of distances were divided by these factors in order to correct for magnification and make the distances comparable. Since we had both positive and negative values for the positions (positive when through the apex and negative when short of the apex), we had to keep the signs in order to get correct distances showing if the endodontic file was observed through the apex or short of the apex. The whole data consisting of the number of the endodontic files (4), positions (3), magnifications (4), digital resolutions (4), observers and observer types were entered into a spreadsheet of SYSTAT Program (Intelligent software, Evanston, IL, USA) . The measured distances for the length (between the file and the apex) and the bar value (visual analog scale) were also entered into the program. There were altogether 1920 data entered for each variable. 72 Since we were using parametric or quantitative scales for our measurements and the distribution of the data was linear, a statistical program in SYSTAT Software, called Multivariate General Linear Hypothesis ( M G L H ) was used to estimate and test the Analysis of Variance ( A N O V A ) . For each of the independent or categorical variables (file size, positions, magnifications, digital resolutions and the observers), A N O V A was performed to compare the differences among the means and standard deviations. The dependent variables were "length", the observers' measurements between the end of the endodontic file and the apex, for the objective assessment of image quality, and "bar", the measurement on the visual analog scale, for the subjective assessment. The interactions between each pair of independent variables were also included in the test. The residuals were saved to test for any trend in the residual plots around the zero value. For testing the pairwise differences between the means of each two variables, for instance endodontic files, positions, magnifications or digital resolutions, the Bonferroni's Post Hoc Tests were performed. The probability values less than 0.05 (p < 0.05) were considered significant. The results were also shown in histograms using Cricket Graph Software Program. 5.2 - Results The density values of the test objects (unexposed film and the penetrameter), processed before start, between each batch of 12 and at the end, were measured by the densitometer (described on page 60). The density ranges, recorded in 4 areas of the unexposed films (base plus fog), varied from 0.23 to 0.25 wi th the mean value of 0.24 and standard deviation of 0.004. The density values of the darkest part of the penetrameter varied from 1.63 to 1.73 wi th the mean value of 1.68 and standard deviation of 0.04. This confirms the consistency of the processing throughout the entire time. . 73 1 The density values of the unattenuated part of the 48 radiographs, after processing, also measured using the densitometer, varied from 1.76 to 1.98 with the mean value of 1.87 and standard deviation of 0.05. Figure 31 shows the density of the radiographs plotted against the processing order. There was no significant variation among the processing order. Figure 31. axis). Density of the radiographs (y-axis) plotted against processing order (x-74 Effects of Geometric and Digital Unsharpnesses on Observers' Objective and Subjective Evaluation Figure 32 shows the mean difference in distance recorded between the end of the files and the apex of the root for different digital resolutions. The mean magnification (geometric unsharpness) is considered for each dpi . The mean differences or observers' errors for the size 15 file are very small. Figure 33 shows the observers' assessment of image quality for the same values. It looks like a mirror image of the previous figure, wi th the number 15 file having the highest assessments. Figure 34 indicates the mean differences in distances between the file and the apex for different geometric unsharpnesses wi th pooled digital resolutions, and Figure 35 shows the range of observers' assessments for the same values. Figures 36 and 37 show the effect of digital and geometric unsharpness on observers' objective evaluation (mean differences in distance) for all the file sizes. The same scaling has been used in these illustrations to facilitate comparison. Figures 38 and 39 show these effects on observers' subjective evaluations for all the file sizes. Table 6-a gives A N O V A for the differences in distance, and Table 6-b the Bonferroni p-values to separate the different effects of the four digital resolutions and the four geometric unsharpnesses. As the figures show in Table 6-b, there is significant difference between the digital resolution of 75 dpi and all the others for all the file sizes. A s the file size becomes larger, there is no significant difference between different resolutions other than 75 dpi . There is also a significant difference between the magnification of 1.3 and all the others for all the file sizes. Table 7-a gives A N O V A for the observers' assessment of image quality, and table 7-b the Bonferroni p-values to separate the different effects of the four digital resolutions and the four geometric unsharpnesses. Figure 40 shows the effect of geometric unsharpnesses and digital resolutions of file size 15 wi th 75 dpi and 1.3 magnifications being removed. • f i le6 El f i le8 • f i le lO H f i le15 75 150 300 600 DPI Bonferroni p-values: Difference in Distance. Fi le 6 8 10 8 0.03 10 <.00001 <.00001 15 <.00001 <.00001 <.00001 DPI 75 150 300 150 <.00001 300 <.00001 0.11 600 <.00001 >.9999 0.61 Figure 32. Mean difference in distance for each digital resolution (pooled magnification). Bars represent standard errors. 76 7 5 1 5 0 3 0 0 DPI 6 0 0 • fi le6 ^ f i le8 • f i le lO M f i le15 Bonferroni p-values: Observers' Assessment of Image Quality. Fi le 6 8 10 8 >.9999 10 <.00001 <.00001 15 <.00001 <.00001 <.00001 DPI 75 150 300 150 <.00001 300 <.00001 <.00001 600 <.00001 <.00001 >.9999 Figure 33. Observers' assessments for image quality for each digital resolution (pooled magnification). 77 • file6 Ij file8 • filelO H file15 M a g n i f i c a t i o n Bonferroni p-values: Difference in Distance. Fi le 6 8 10 8 0.15 10 <.00001 <.00001 15 <.00001 <.00001 <.00001 M a g 1.0 1.1 1.2 1.1 >.9999 1.2 >.9999 >.9999 1.3 0.001 0.0002 0.0008 Figure 34. Mean difference in distances for each magnification (pooled digital resolution). 78 C (0 V 3 E O (0 E co — O o • file6 Wk file8 • filelO P J file15 Bonferroni p-values: Observers' Assessment of Image Quality. Fi le 6 8 10 8 >.9999 10 <.00001 0.00003 15 <.00001 <.00001 <.00001 M a g 1.0 1.1 1.2 1.1 0.27 1.2 <.00001 <.00001 1.3 <.00001 <.00001 <.00001 Figure 35. Observers' assessment of image quality for each magnification (pooled digital resolution). DPI 600 • Magi W Mag2 • Mag3 M Mag4 Figure 36. Mean difference in distances for each digital resolution and each magnification, Files 6 and 8. o £ c »- 0 0> o H - C 5 2 • Magi ^ Mag2 • Mag3 H Mag4 a> E Q ro gs o 5 c 2 H o-i— File 15 I _JHHBKJ^L 7 5 150 300 DPI • Magi M Mag2 • Mag3 M Mag4 600 Figure 37. Mean difference in distances for each digital resolution and each magnification, files 10 and 15. 10 File 6 8H Figure 38. Observers' assessments of image quality for each digital resolution and each magnification, files 6 and 8. £ o CO (D w Si, < I CO — O o Magi Mag2 Mag3 Mag4 600 cu CO E 6 CO 0) » & CO O) < I CO — O o 7 5 600 • Magi W Mag2 • Mag3 M Mag4 150 300 DPI Figure 39. Observers' assessments of image quality for each digital resolution and each magnification, files 10 and 15. 83 LO r—< CJ •T-I P H P H O T-H • i-H P H 0 ) C ca -*-» CO Q •01 u C 4> JH QJ cy C ra ca > ca ca H t / j 'ca C < P H 0 0 r - H P H VO P H P H Cu P H o- 6 QJ ^ 3 ft CU 60 <u 01 1-4 o r-t r-H T-H T-H o o o o o o o T-H o o o o o o o o o o o o d d d d V V V V T—1 LO LO VO T-H LO CN T-H LO rH CN LO CO vd 0 0 T-H CO T~< o o LO o o CO o OV p p r-H p o o o d d d V VO VO 0 0 0 0 p T-H o o vO cri CN* CN T - i o CN T-H LO CN T-H T-H o o o o o o o o o o o o o o o o o p p o d o d d d V V V vO t N , VO vO Ov CN T-H CM CO T—i 0 0 vb 0 0 T-H T-H T-H T-H T-H T-H T-H o o o o o o o o o o o o o o o p o p o p o o d d d V V V V V 0 0 r>s vO VD L\ CO Ov CO CO 0 0 d CO LO T-H T-H T-H T-H CO CO • CN ON Ov P H D c Vt-> ca u C • C O ca s o O P H O* > >H 4 1 c/j O Ov vO CO 0 0 CC h—I P H Q o I H ) H pq 'in i—i o o C O o LO in O N O N O N O N A O N O N O N O N O N O N O N O N A A O O O o o o o o o o o o V V V o I—I o o C O o in O N O N O N O N A O N O N O N O N O N O N O N O N A A o o o o o o o o o o o o V V V 00 P H o o C O o in m C N o o O N °^ -O N C V O N O A O O O O o o o o o o o o o V V V N O JO o o C O o m O N O N O N O N A V O C N O o o o o o o o o o o o o o o o o V V V C N o o o o P H Q o o o m o o r—I C O N O 00 O N o O N o O N o O N p A o C N O N o o O N o o O N o p O N o o A C N O O O N O N O N O N r - H O N O A O N O N O N 00 O N O N o O N O N o O N A o A C N I - J 00 o T - H O N o O o N O O N o O O N o O N H o T - H O N C N p A O * o C N N O o o o O N O O N O O N O O N O A V r - H O N o O N o O N C N o O N o A O V r - H C N C O r-5 r - i T - H ca a o> bo RS 6 o> s CO CO 0> cn CO < "cn ?H CU > 5H 0> cn 8 cu u C fS • r H >H ca > cn CU j^~> ce 5 LO rH J H • r H PH PH O T -H JO 00 JO a, PH OH PH OH PH ° fi 60 OJ nj ^H Oi u JH o r-H r-H r H r H O o o o o o " t f o o o o o o o p o o o p o o o o o V V V V VO T-H o 00 VO -tf o NO 00 CN LO LO tv o LO VO r H "tf r-H r-H r-H T"H r H o o o o o o o o o o o o o o o p , p p p p o o o o o V V V V V 00 VO o ON CN VO o r H tv o o IV.' o ON LO r H T-H H r-H r-H r H o o o o o o o o o o o o o o o p p p p p o o o o o V V V V V (N CN tv r H CN vq o LO LO VO T-H o vd ON vo CO CN r H T-H r-H r-H r-H r H r H o o o o o o o o o o o o o o o o p p p o o o o o o V V V V V 00 o CO "tf 00 CO tv "tf 00 "tf "tf 00 - -tf CN* VO H CN CO CN CN r-H CO CO CN ON ON 6 o • H -t-> c3 O bO ca s o cn O PH <U > JH 0) cn 6 ON VO CO bC ca X r H PH Q O i-i >-i P- l 86 a QJ bC cs s QJ B CO CO OJ CO CO < " c o IH QJ > IH QJ. CO § CO QJ j 3 "c3 > OH -9 o cC O H pa LO r-H PH1 o r - H CO r S VO ft O o CO o LO LO IN d o o o o o o o o V V o o o o o o o o o o o o V V V o o CO o LO CO CO d o o o o o o o o V V o o o o o o o o o o o o V V V o o CO o LO Ov o\ ON ON A o o o o o o o o V V o o o o o o o o o o o o V V V o o CO o LO LO vO o o o o o o o o o V V o o o o o o o o o o o o V V V s o o o LO o o T—i CO vO CN o o o o V CN T-H O o o o o o p o d V T-H T-H O N o o O N o o O N o o O N o o A V V CN O o o o V r - H o T-H O o o O O d v CO o o o o o o o o p o o d v V CN CN O o o o o o o o o o o V V* o o o o CO o o CO o o d v V CN o o o o V o o o o o o o o V V o o o o LO O O LO O O d v V CN CO 87 Figure 40. Removing dpi , 75 and magnification of 1.3 from the graph. The mean difference in distance for file 15 is less than 0.2 mm. Table 8 shows the Bonferroni p-values for the difference in distance from the total study which confirms the significant differences among the digital resolution of 75 dpi and the others. The same goes for the magnification of 1.3. Figure 41 shows the residual plots which are more or less around zero wi th no up or down trend. Figures 42 and 43 show the objective and subjective evaluations between the first and second readings for 5 of the repeat images. Since the 24 repeat images were a random combination of poor and good quality images, it was decided to select 5 of them wi th the range of poor to good digital resolutions and magnifications. There is no significant difference between the first and second reading for either of objective and subjective evaluations, implying consistent intraobserver variation. Figure 44 shows the objective and subjective evaluations of the best images (mag 1, dig 600) for the 3 different positions. There is a significant difference between the position of the file size 10 and the rest of the files. The mean difference in distance for the best images considered as the silver standard were used in order to compare the positions between different file sizes. A N O V A was performed to test the difference among the means. There was no significant differences among the 3 positions for any of the file sizes either wi th gold or silver standards, other than file #10, gold standard (Figure 45). Table 9 shows the A N O V A for the two observer types. It shows no significant difference between the dentists and the students. 89 File 6 8 10 8 0.006 10 <.00001 <.00001 15 <.00001 <.00001 <.00001 Posi t ion 1 2 2 >.9999 3 0.008 0.0006 Magnification 1.0 1.1 1.2 1.1 >.9999 1.2 >.9999 >.9999 1.3 <.00001 <.00001 <.00001 DPI 75 • 150 300 <.00001 <.00001 0.04 <.00001 >.9999 0.35 150 300 600 Table 8. Bonferroni p-values for difference in distance from the total study. Figure 41. Studentized residuals plotted against estimates for the four file sizes Difference between 1st & 2nd Readings 7 5 1 5 0 3 0 0 6 0 0 6 0 0 1 0 6 6 1 0 1 5 P1 P3 P3 P3 P2 M1 M2 M3 M4 M2 A N O V A for 5 repeat images: Difference in Distance. Source Degrees of F-Ratio P Freedom Reading 1 1.22 0.28 Obs 9 1.30 0.27 Image 4 12.67 <.00001 Obs x Reading 9 1.92 0.08 Error 36 Figure 42. Difference in distance for first and second readings for 5 of the repeat images. Dif ference between 1st & 2nd Readings 92 1st 2nd 7 5 1 5 0 3 0 0 6 0 0 6 0 0 1 0 6 6 1 0 1 5 P1 P3 P3 P3 P2 M1 M2 M3 M4 M2 A N O V A for 5 repeat images: Observers' Assessment of Image Quality. Source Degrees of Freedom F-Ratio Reading 1 2.62 0.11 Obs 9 8.45 <.00001 Image 4 136.10 <.00001 Obs x Reading 9 3.29 0.005 Error 36 Figure 43. Observers' assessment of image quality for first and second readings for 5 of the repeat images. 93 Mag 1, Dig 600 P o s i t i o n Figure 44. Objective and subjective evaluation of the best images (mag 1, dig 600) for 3 different positions. Mag 1, Dig 600 0.30-> (/> 0.20 -• P o s i t i o n ANOVA for Position (Mag=l, Dig=600) Analyses of Variance: Difference in Distance. File 6 File 8 File 10 File 15 Source Degrees o f F p F p F p F P Freedom Gold Standard-Position 2 1.89 0.17 2.43 0.11 6.05 0.007 0.72 0.50 Error 27 Silver Standard Position 2 2.45 0.11 1.15 0.33 1.31 0.29 0.57 0.57 Error 27 Figure 45. Objective evaluation of the positions using silver standard. Analyses of Variance: Observers' Type (l=Dentist 2=Student). Source Degrees of Freedom F-Ratio P Obs Type 1 0.92 0.34 File x Obs Type 3 0.98 0.40 Posit x Obs Type 2 0.66 0.52 M a g x Obs Type 3 0.70 0.55 Dig x Obs Type 3 1.47 0.22 Error 1851 Table 9. A N O V A for the two observer types. Chapter 6 Discussion and Conclusions 97 6.1 - Discussion and Conclusions This study has shown that the endodontic file model functions wel l for comparison of geometric and digital unsharpnesses. We found that little clinically relevant information is gained by scanning radiographs wi th resolutions greater than 300 dpi. Furthermore, we have shown that geometric unsharpness has very little or no influence on diagnostic accuracy in clinical radiography. The design of the study was planned wi th care to minimize any k ind of bias or inconsistency. The randomization of the sequence of processing order was performed to avoid a systematic error. The observation sequences were also randomized to make the study blind and avoid any repeating pattern. Base plus fog (the density ranges, recorded in 4 areas of the unexposed films) varied from 0.23 to 0.25 wi th the mean value of 0.24 and standard deviation of 0.004. Svenson et al, 1990, studied the influence of film fog on radiographic caries diagnosis. They concluded that fog up to a level of 0.6 optical density units had no influence on diagnostic accuracy in the absence of any compensation for the increased density caused by fog (Svenson et al, 1990). The density values of the darkest part of the penetrameter wi th the mean value of 1.68 and standard deviation of 0.04 confirmed the consistency of the processing throughout the entire time. The digital resolution of 75 dpi gave much larger errors than other digital resolutions but there was no significant difference between the resolutions of 300 dpi and 600 dpi , which was quite surprising. For the file size 8, however, except for the magnification 1, 300 dpi was better than 600. This was an unexpected result and considered to be a random error. File 6 position was unexpectedly wel l determined by the observers. Surprisingly, the measurement errors for file 10 for the positions 98 using the "gold standard" and the best image quality (mag 1, dig 600) were significantly different from all the other files. This might be a random error due to the difference between the radiographic view and the actual view, or a systematic error due to malpositioning the file in the root. Comparing the mean difference in distance using the silver standard for position (explained in Chapter 5, page 88) showed no significant difference for any of the file sizes (Figure 45). For the file size 15, other than 75 dpi and magnification of 1.3, the errors were less than 0.2 mm. This amount of error is quite trivial and would have no clinical relevance. 75 dpi gives poor quality intraoral radiography images and it was included in order to have a more extended range of the digital resolutions and for comparison of the results. Other studies have mostly looked at resolution wi th measuring l p / m m (described in chapter 3). The relationship between the pixel size and resolution is shown in Table 10 (Vrijens and Strommer, 1997). Pixel /s ize (|iim) l p / m m 67 7.5 100.5 5.0 134 3.7 167.5 3.0 201 2.5 Table 10. The relationship between pixel size and resolution. Figure 47 shows four digital images of group 18 of the line pair plate (10 lp /mm) digitized wi th different resolutions. The image of the whole line pair plate is shown in chapter 3, page 26, wi th line pairs ranging from 1 to 10 l p / m m . As Figure 46 shows, the resolution of 600 dpi gives lines which can be wel l defined from each other. Wi th lower resolutions (fewer and larger size pixels), the definition between the lines becomes harder. Studies that have explored the resolution of digital images include Benz and Mouyen, 1991, who indicated the resolving power of the R V G system as 11 l p / m m compared wi th a maximum of 14 l p / m m for intraoral film (Benz and Mouyen , 1991). Huda et al, 1997, found that the l imit ing spatial resolution of a PSP system was approximately 6.5 l p / m m whereas for film it was in the range of 11 to 20 l p / m m (Huda et al, 1997). Farman A G et al, 1998, found that maximum spatial resolution of panoramic film approached 5 l p / m m whereas wi th the C C D it was just above 4 l p / m m (Farman and Farman, 1998). 600 300 150 75 D P I D P I D P I D P I Figure 46. Four digital images of group # 18 (10 l p / m m ) of the line pair plate shown in Figure 1, page 26, at the four resolutions used in this study. The magnification of 1.3 is also beyond the range of geometric unsharpnesses that are clinically relevant, and the same reasons apply for including it in the study. Based on observers' objective evaluation, there was no significant difference between the magnifications other than 1.3, for any of the file sizes. However, their subjective assessment based on the visual analog scale indicates that, regardless of 1 00 the highest magnification (1.3), there were significant differences between the magnification of 1.0 and 1.2 and also 1.1 and 1.2. This shows the difference between objective and subjective evaluations. There are very few studies that have used parametric data as an objective evaluation to compare the results. This is the only way that certain common statistical tests that calculate means and variances can be applied and the results are not merely based on observers' subjective rating of the image quality. This study shows how subjective and objective assessments can be different for some image quality tasks. As was mentioned, our nu l l hypothesis was based on the theory that higher digital resolutions have better image quality. Lack of significant difference between the resolution of 300 and 600 dpi for the file size 15, which is the min imum file size for accurate radiographic interpretation (Glassman and Serota, 1997), rejects the nul l hypothesis and implies no benefit in having digital resolutions greater than 300 dpi to improve image quality. Lack of significant difference between any of the magnifications other than 1.3 (which is beyond clinical relevance) also proves that geometric unsharpness had no influence on diagnostic information of clinical radiographs for this purpose. The only study that has looked at the effect of geometric unsharpness, due to focal spot finite size, is by Platin et al., 1996 (Platin et al., 1996). They also concluded that although focal spot size is theoretically important in image production, its variation plays little importance in clinical practice. However, they selected focal spots wi th relatively little difference, did not design the image geometry to maximize the unsharpness, and used a caries model, which we regard as a poor evaluator of resolution, because of the ill-defined edges. Edge resolution is affected by geometric unsharpness, but dental caries does not have a defined edge between the lesion and the sound area. In dental caries or in periodontal disease the radiolucency of the affected area, on radiographs, gradually 101 fades out without showing any sharply defined edge. The models that do have defined margins and could be good for studying geometric unsharpness and resolution include, endodontic files, fractures in teeth, lamina dura and periodontal ligament. Reduction of observers' errors for the larger files, file size 15 versus 10, 8 and 6, and significant differences between the file sizes does not reject our nu l l hypothesis that larger files show better in radiographs. There was no significant difference between the two types of observers, dentists and senior dental students, which implies these tasks are not dependent on observers' skills. For the 5 repeat images with the range of poor to best image quality, there was no significant difference between the 1st and 2nd reading. There was little intraobserver variation for the 24 repeat images between the first and second reading. The standard deviation of the mean difference in distance for the 10 observers for the best image quality (mag 1 and dig 600) ranged from 0.05 to 0.18 mm. This implies a minimal interobserver variation for the good images. Since we deliberately had a large variation of poor and good image qualities wi th different size of endodontic files, the interobserver variation for the whole study was significant ( p <0.00001 to p = 0.005 for the difference in distance and p < 0.00001 for the observers' assessment of image quality) which was not surprising. In a study by Eckerbom et al., which evaluated interobserver variation in radiographic examination of endodontic variables, it was concluded that the interobserver variation may be reduced by calibration of observers and by establishing strict criteria for evaluation (Eckerbom et al, 1986). In another study by Reit and Hollender, regarding radiographic evaluation of endodontic therapy and the influence of inter-and intraobserver variation, they found large variability (Reit and Hollender, 1983). It was concluded that the variation is due to difficulties in defining and maintaining 1 02 criteria for radiological evidence of periapical disease. Their findings were however, based on subjective evaluation and most of the other studies also regarding these issues have not explored the objective parametric evaluation of the differences. It was decided to avoid the addition of a tissue equivalent scattering medium in order to reduce complexities that would compromise the precision of determination of file position. However, the comparison of the effects of geometric and digital unsharpnesses have been explored more reliably by this design and the findings are applicable to clinical situations where more noise is present. In a study by Eckerbom et ah, in 1997, estimation of the technical quality of endodontic treatment, using conventional buccolingual radiographic projection and the effect of surrounding tissues on these evaluations were explored (Eckerbom and Magnusson, 1997). They found that the length of endodontic root fil l ing, in vivo and in vitro, was interpreted to be the same. In other words, the amount of tissue surrounding the teeth d id not make a difference in interpretation of the length of the root filling. Price and Ergi i l , 1997, also found that the effect of scattering medium in their caries model was insignificant (Price and Ergi i l , 1997). Regarding the method of measurement, a few other studies have performed similar objective methods using endodontic file models for difference purposes. Versteeg et al, 1997, evaluated estimation of distances from the tip of an endodontic file to the radiographic apex of teeth on digital images (which are much larger than actual size) and conventional radiographs in order to compare observers' estimates (Versteeg et al, 1997). Cederberg et al, 1998, performed a clinical study using an endodontic file model, to compare measurements of file lengths and root lengths and their differences on digital images and conventional radiographs (Cederberg et ah, 1998). The purpose of the study was to compare PSP luminescence imaging versus radiographic fi lm. 1 03 6.2 - Different aspects of Measurement 6.2.1 - Resolution. H o w fine a detail can be measured? The details could be measured to the nearest pixel as a whole pixel size. As explained in chapter 4, page 67, there were 18.7 pixels in each m m of the scale. We used this ratio for converting our measurement from pixels to millimeters. The accuracy of the measurement, of course, depends on the location of the line on the ruler that we selected for the measurement (Figure 30, page 69). However, it was consistent wi th a few repeated measurements. The resolution of the distance measurements, after converting pixels to mm, was 0.05mm. 6.2.2 - Precision. H o w close repeated measurements of the same quantity are to each other? Repeating the measurements of the x and y coordinates in pixel size by the same operator, resulted in the same values. The reports of some measurements by a different operator resulted in slightly different values, approximately ±1 pixel (0.05mm), which were ignorable. Since all the measurements were performed by the same operator, they were consistent and precise. 6.2.3 - Accuracy. Do the measurements include the true value? "The criterion for accuracy is the absence of bias." (Brunette, 1996) We can discuss the accuracy in regard to our gold standard or the true measurement between the file and the root. We used a microscope and stage micrometer along wi th 2mm stops for determining the positions. The level of accuracy was 0.01 millimeter for 1 the stage micrometer. The 2mm stops were measured wi th a micrometer to an accuracy of 0.025mm. Some additional error might have occurred due to limitations in the fit of the stops to the root face and the files. The analysis of variance for positions using the "gold standard" and the best quality images (magnification 1 and 600 dpi) showed a significant difference for the position of file 10 (p=0.007) from all the other files. We considered this difference: 1) a random error due to the difference between the radiographic view and the actual view, or 2) a systematic error due to malpositioning the file #10 in the root. The latter was confirmed by comparing the mean difference in distance using the "silver standard" (the mean of the best image quality of observers' measurements) for the position, which showed no significant difference for any of the file sizes (Figure 45, page 94). This indicated that the error was due to malpositioning the file for this file size. 6.2.4 - Validity. Does it really measure what it claims to measure? H o w wel l the distance between the end of the file to the tip of the root can be determined depends on the image quality and the observers' skills. Our aim regarding this measurement was to assess the effect of geometric unsharpness and digital resolutions in clinical dental radiography. When image quality is evaluated, the choice of the model has considerable bearing on the relative weighting of these factors. The interpretation of dental caries is a common and relevant task, but, because of the ill-defined edge of the lesions, it depends more on broad area contrast than on resolution. Therefore the validity of dental caries models for evaluating resolution is questionable. We chose endodontic files, which due to their sharply defined edges and small diameters, formed a good model for our purpose. 1 05 Correlation validity compares the results of a new measuring method to a highly validated measurement (gold standard) (Brunette, 1996). In our study, because of the combination of good and poor images, intentional to study the effects, the correlation between observers' measurements and the gold standard would be a meaningless task. Another argument is that the observers' measurements are in the form of continuous data, whereas our gold standard is discrete. Therefore, we cannot correlate the two factors. Instead we decided to select the high quality images (mag 1 and 600 dpi) and calculated the standard deviation of measurement errors (Difference in Distance of our observers' measurements using the gold standard). This standard deviation was 0.1mm which is quite a small amount and validates the method of the measurement. Standard deviation of measurement errors using the silver standard was even smaller (0.07mm) which can be explained by the error of malpositioning file #10 as was discussed before. Performing A N O V A for the best images also shows no significant difference among the observers. 6.2.5 - Reliability. The extent to which the method of measurement performs consistently. To determine the reliability of our measurements, we asked all our observers to view a sample of 24 re-randomized images for the second time wi th at least one week interval. To obtain the intraobserver variation for the 10 observers, a paired t-test was performed for each observer. There was no significant difference between 1st and 2nd readings for any of the observers (t = 0.14 to 1.50 and p = 0.15 to 0.89 for the difference in distance) other than one of the dentists (t = 2.75 and p = 0.01). This was considered a high reliability for the nine observers. The problem is, of course, 1 06 that the values attained may be related more to the individuals selected for the study than to methods themselves. "In clinical studies, reliability is often considered relative to the total variability in the sample. The reliability is a measure of the proportion of the variability in scores, which is due to true difference between individuals." (Brunette, 1996) To determine the reliability among the 10 observers for each repeat image, we also performed a paired t-test for each image. We found significant differences among the observers, for mostly the good quality images (t = 0.01 to 42.44 and p = <0.0001 to 0.04 for the differences in distance). This can be explained by the fact that good quality images have less variability, therefore the t-test is able to show slight differences. Whereas for the poor images, the variance is large which tends to produce a small t value, making it harder to detect differences between groups. We also calculated the correlation coefficient for our data, that are in the form of continuous measurements, to assess reliability. The Pearson correlation coefficient between the 1st and 2nd reading of length (mean difference in distance) and bar (observers' assessment of image quality) for the 24 images showed the correlation coefficient r = 0.873 for the length and r = 0.978 for the bar. Correlation coefficient values for both bar and length are high which indicates a high reliability of the measuring method. Figure 47 shows a positive correlation between first and second readings. 1 07 a) b) Figure 47. a) Scatter plot relating the differences in distance for the 1st and 2nd readings (Length 1, Length 2). b) Scatter plot relating the differences in observers' subjective assessments (Bar 1, Bar 2). 1 08 Estimating reliability by the approach of analysis of variance (Brunette, 1996) In this approach, the variance in a population of measurements is affected by various components of the process including the effects of images, observers and random error. The calculation of the error term involves determining how much each individual score deviates from its expected value. The reliability coefficient is defined as the ratio of variance among images of the total of error variance plus variance among images and variance among observers. 2 . s images R = ~ 2 2 2 s images + s observers + s error 39.13812 ~ 39.13812 + 3.68405 + 0.71406 R=0.90 The data was taken from computer calculation for A N O V A between 1st and 2nd reading for the difference in distance (length). The reliability coefficient is 0.90, which confirms that the measurement reflects the true value for the images. However, it should be remembered in interpreting the reliability coefficient (coefficient is the ratio of subject variability to total variability), tests conducted on heterogeneous populations (which have high variability, e.g. high variability of our images qualities), tend to give high reliability values (Brunette, 1996). 1 09 Few studies have used the parametric measurements for the same purpose as our study. Most of the studies, regarding the assessment of radiographic endodontic images, use subjective or qualitative measurements. Therefore, their analysis is l imited to nonparametric analysis that does not include means and standard deviations. Our study in approaching a method of measurement using parametric data can be considered an advantage over the previous methods of evaluating image quality, using non-parametric techniques, that were more dependent on subjective analysis rather than real differences i n objective evaluation. Our study involves scanning conventional radiographs but the findings can be extrapolated to direct digitally acquired images. Digi t iz ing conventional film images provided a much greater flexibility of image quality than using direct digitally acquired images. In direct digital imaging, the resolution is typically in region of 600 dpi or less. There seems to be a great move at the moment to produce direct digital imaging systems wi th greater resolutions and larger number of grey shades. Our study has raised questions about the validity of these attempts. 1 1 0 Bibliography A D A . (1970). American National Standard, Speed Groups: American Dental Association). Adosh, L . , Vandana, K . L . , and Mehta, D . S. (1997). A n appraisal of periodontal bone loss surgically and by radiovisiography. A comparative study. Indian Journal of Dental Research 8, 27-31. Benz, C., and Mouyen, F. (1991). Evaluation of the new RadioVisioGraphy system image quality. Oral Surg Oral M e d Oral Pathol 72, 627-31. Borg, E. , and Grondahl, H . G . (1996). Endodontic measurements in digital radiographs acquired by a photostimulable, storage phosphor system. Endodontics & Dental Traumatology 12, 20-4. Borg, E., and Grondahl, H . G . (1996). O n the dynamic range of different X-ray photon detectors in intra-oral radiography. A comparison of image quality i n fi lm, charge-coupled device and storage phosphor systems. Dento Maxi l lo Facial Radiology 25, 82-8. Borg, E. , Grondahl, K . , and Grondahl, H . G . (1997). Marginal bone level buccal to mandibular molars in digital radiographs from charge-coupled device and storage phosphor systems. A n in vitro study. Journal of Clinical Periodontology 24, 306-12. Borg, E., Kallqvist, A . , Grondahl, K . , and Grondahl, H . G . (1998). F i l m and digital radiography for detection of simulated root resorption cavities. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 86,110-4. 111 Brunette, D . M . (1996). Critical Thinking, Understanding and Evaluating Dental Research (Chicago: Quintessence). Cederberg, R. A . , Tidwel l , E., Frederiksen, N . L . , and Benson, B. W . (1998). Endodontic working length assessment. Comparison of storage phosphor digital imaging and radiographic film. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 85, 325-8. Christgau, M . , Hi l ler , K . A . , Schmalz, G. , Kolbeck, C , and Wenzel, A . (1998). Accuracy of quantitative digital subtraction radiography for determining changes in calcium mass in mandibular bone: an in vitro study. Journal of Periodontal Research 33, 138-49. Christgau, M . , Hil ler , K . A . , Schmalz, G. , Kolbeck, C , and Wenzel, A . (1998). Quantitative digital subtraction radiography for the determination of small changes in bone thickness: an in vitro study. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 85, 462-72. Conover, G . L . , Hildebolt, C . F., and Yokoyama-Crothers, N . (1996). Comparison of linear measurements made from storage phosphor and dental radiographs. Dento-Maxillo-Facial Radiology 25, 268-73. Cox, V . S., Brown, C , Jr., Bricker, S. L . , and Newton, C. W. (1991). Radiographic interpretation of endodontic file length. Oral Surgery, Oral Medicine, Oral Pathology 72, 340-4. 1 1 2 Curry III, T. S., Dowdey, J. E., and Mur ry Jr, R. C. (1990). Christensen's Physics of Diagnostic Radiology, 4th Edition (Philadelphia, London: Lea & Febiger). Delano, E. O., Tyndall , D. , Ludlow, J. B., Trope, M . , and Lost, C. (1998). Quantitative radiographic follow-up of apical surgery: a radiometric and histologic correlation. Journal of Endodontics 24, 420-6. Eckerbom, M . , Andersson, J. E. , and Magnusson, T. (1986). Interobserver variation in radiographic examination of endodontic variables. Endodontics & Dental Traumatology 2, 243-6. Eckerbom, M . , and Magnusson, T. (1997). Evaluation of technical quality of endodontic treatment—reliability of intraoral radiographs. Endodontics & Dental Traumatology 13, 259-64. Ellingsen, M . A . , Harrington, G . W., and Hollender, L . G . (1995). Radiovisiography versus conventional radiography for detection of small instruments in endodontic length determination. Part 1. In vitro evaluation. Journal of Endodontics 21, 326-31. Ellingsen, M . A . , Hollender, L . G. , and Harrington, G . W. (1995). Radiovisiography versus conventional radiography for detection of small instruments in endodontic length determination. Part 2. In vivo evaluation. Journal of Endodontics 21, 516-520. El lwood, R. P., Davies, R. M . , and Worthington, H . V . (1997). Evaluation of a dental subtraction radiography system. Journal of Periodontal Research 32, 241-8. Farman, A . G. , Avant, S. L . , Scarfe, W. C , Farman, T. T., and Green, D . B. (1998). In vivo comparison of Visualix-2 and Ektaspeed Plus in the assessment of 1 1 3 periradicular lesion dimensions. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 85, 203-9. Farman, A . G. , and Farman, T. T. (1998). Panoramic dental radiography using a charge-coupled device receptor. Journal of Digital Imaging 11,166-8. Farman, T. T., and Farman, A . G . (1998). Cl inical trial of panoramic dental radiography using a C C D receptor. Journal of Digital Imaging 11,169-71. Farman, T. T., Farman, A . G. , Kel ly , M . S., Firriolo, F. J., Yancey, J. M . , and Stewart, A . V . (1998). Charge-coupled device panoramic radiography: effect of beam energy on radiation exposure. Dento-Maxillo-Facial Radiology 27, 36-40. Farman, T. T., Farman, A . G. , Scarfe, W. C , and Goldsmith, L . J. (1996). Optical densities of dental resin composites: a comparison of C C D , storage phosphor, and Ektaspeed plus radiographic film. General Dentistry 44, 532-7. Fitzgerald (1947). Dental Roentgenography, 1: A n investigation in adumbration, or the factors that control Geometric Unsharpness. The Journal of the American Dental Association 34,1-20. Garlock, J. A . , Scarfe, W. C , Kamer, K . R., and Farman, A . G . (1996). Measurement algorithm accuracy of the R V G - P C i in vertical and diagonal assessments at various beam energies. Journal of Endodontics 22, 646-50. Glassman, G . D . , and Serota, K . S. (1997). Endodontics in Transition: Evolution of the Rotary Revolution. Oral Health, 37-40. Goaz, P. W. , and White, S. C. (1982). Oral Radiology, Principles and Interpretation (Toronto: Mosby). Goaz, P. W. , and White, S. C. (1994). Oral Radiology, Principles and Practice, third Edit ion (Toronto: Mosby). Goshima, T., Goshima, Y. , Scarfe, W. C. , and Farman, A . G . (1996). Sensitometric response of the Sens-A-Ray, a charge-coupled imaging device, to changes in beam energy. Dento-Maxillo-Facial Radiology 25, 17-8. Graft, B. M . , Sickles, E. A . , and Nguyen, N . T. (1979). Dental xeroradiography for endodontics: a rapid X-ray system that produces high-quality images. Journal of Endodontics 5, 266-70. Hayakawa, Y . , Farman, A . G. , Scarfe, W. C., Kuroyanagi, K . , Rumack, P. M . , and Schick, D . B. (1996). Opt imum exposure ranges for computed dental radiography. Dento-Maxillo-Facial Radiology 25, 71-5. Hayakawa, Y . , Shibuya, H . , Ota, Y. , and Kuroyanagi, K . (1997). Radiation dosage reduction in general dental practice using digital intraoral radiographic systems. Bulletin of Tokyo Dental College 38, 21-5. Hildebolt, C. F., Fletcher, G. , Yokoyama-Crothers, N . , Conover, G . L . , and Vannier, M . W. (1997). A comparison of the response of storage phosphor and film radiography to small variations in X-ray exposure. Dento-Maxillo-Facial Radiology 26,147-51. 1 1 5 Huda , W. , R i l l , L . N . , Berrn, D . K . , and Pettigrew, J. C. (1997). Comparison of a photostimulable phosphor system with film for dental radiology. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 83, 725-31. Huysmans, M . C , Hintze, H . , and Wenzel, A . (1997). Effect of exposure time on in vitro caries diagnosis using the Digora system. European Journal of Oral Sciences 105,15-20. ISO. (1981). International Standard, Processing. Jones, G . A . , Schuman, N . J., and Woods, M . A . (1998). Estimated skin exposure as an indicator for comparing radiovisiography (RVG) versus conventional Ektaspeed Plus dental radiography. Journal of Clinical Pediatric Dentistry 22, 121-3. Kang, B. C , Farman, A . G. , Scarfe, W. C , and Goldsmith, L . J. (1996). Observer differentiation of proximal enamel mechanical defects versus natural proximal dental caries wi th computed dental radiography. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 82, 459-65. Kassebaum, D . K. , McDav id , W. D. , Dove, S. B., and Waggener, R. G . (1989). Spatial resolution requirements for digitizing dental radiographs. Oral Surgery, Oral Medicine, Oral Pathology 67, 760-9. Kircos, L . T., Staninec, M . , and Chou, L . S. (1989). Effect of developer temperature changes on the sensitometric properties of direct exposure and screen-film imaging systems. Dento Maxi l lo Facial Radiology 18,11-4. 1 1 6 Kircos, L . T., Staninec, M . , and Chou, L . S. (1989). Rare earth filters for intraoral radiography: exposure reduction as a function of kV(p) wi th comparisons of image quality. Journal of the American Dental Association 118, 605-9. Kullendorff, B., and Nilsson, M . (1996). Diagnostic accuracy of direct digital dental radiography for the detection of periapical bone lesions. II. Effects on diagnostic accuracy after application of image processing. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 82, 585-9. Kullendorff, B., Nilsson, M . , and Rohlin, M . (1996). Diagnostic accuracy of direct digital dental radiography for the detection of periapical bone lesions: overall comparison between conventional and direct digital radiography. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 82, 344-50. Kullendorff, B., Petersson, K . , and Rohlin, M . (1997). Direct digital radiography for the detection of periapical bone lesions: a clinical study. Endodontics & Dental Traumatology 13, 183-9. Langen, H . J., Kle in , H . M . , Wein, B., Stargardt, A . , and Gunther, R. W. (1993). Comparative evaluation of digital radiography versus conventional radiography of fractured skulls. Investigative Radiology 28, 686-9. Lawrence, H . P., Benn, D . K . , and Sheiham, A . (1997). Digital radiographic measurement of approximal caries progression in fluoridated and non-fluoridated areas of Rio de Janeiro, Brazil . Community Dentistry & Oral Epidemiology 25, 412-8. Leddy, B. J., Miles, D . A . , Newton, C. W., and Brown, C , Jr. (1994). Interpretation of endodontic file lengths using RadioVisiography. Journal of Endodontics 20, 542-5. 1 1 7 L i m , K . F., Loh , E. E. , and Hong, Y . H . (1996). Intra-oral computed radiography—an in vitro evaluation. Journal of Dentistry 24, 359-64. Meredith, W . J., and Massey, J. B. (1968). Fundamental Physics of Radiology (Bristol: John Wright & Sons LTD.) . Mistak, E. J., Loushine, R. J., Primack, P. D. , West, L . A . , and Runyan, D . A . (1998). Interpretation of periapical lesions comparing conventional, direct digital, and telephonically transmitted radiographic images. Journal of Endodontics 24, 262-6. Mondou , D. , Bonnet, E., Coudert, J. L. , Jourlin, M . , Molteni , R., and Pachod, V . (1996). Criteria for the assessment of intrinsic performances of digital radiographic intraoral sensors. Academic Radiology 3, 751-7. Moystad, A . , Svanaes, D . B., Risnes, S., Larheim, T. A . , and Grondahl, H . G . (1996). Detection of approximal caries wi th a storage phosphor system. A comparison of enhanced digital images wi th dental X-ray film. Dento Maxi l lo Facial Radiology 25, 202-6. Nair , M . K . , Tyndall , D . A . , Ludlow, J. B., and May, K . (1998). Tuned aperture computed tomography and detection of recurrent caries. Caries Research 32, 23-30. Naitoh, M . , Yuasa, H . , Toyama, M . , Shiojima, M . , Nakamura, M . , Ushida, M . , Iida, H . , Hayashi, M . , and Ar i j i , E. (1998). Observer agreement in the detection of proximal caries wi th direct digital intraoral radiography. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 85, 107-12. 1 1 8 Nielsen, L . L . , Hoernoe, M . , and Wenzel, A . (1996). Radiographic detection of cavitation in approximal surfaces of primary teeth using a digital storage phosphor system and conventional fi lm, and the relationship between cavitation and * radiographic lesion depth: an in vitro study. International Journal of Paediatric Dentistry 6,167-72. Ohki , M . , Okano, T., and Nakamura, T. (1994). Factors determining the diagnostic accuracy of digitized conventional intraoral radiographs. Dento Maxi l lo Facial Radiology 23, 77-82. Okano, T., Wiebe, J. D. , Webber, R. L . , and Wagner, R. F. (1983). Effective exposure level and diagnostic performance in endodontic radiography. Oral Surgery, Oral Medicine, Oral Pathology 55, 527-36. Platin, E. , Mauriello, S., and Ludlow, J. B. (1996). Effects of focal spot size on caries diagnosis wi th D and E speed images. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 81, 235-239. Price, C . (1995). Sensitometric evaluation of a new E-speed dental radiographic film. Dento-Maxillo-Facial Radiography 24, 30-36. Price, C , and Ergi i l , N . (1997). A comparison of a film-based and a direct digital dental radiographic system using a proximal caries model. Dento Maxi l lo Facial Radiology 26, 45-52. Price, C , and McDonnel l , D . (1990). Effects of niobium filtration and constant potential on the sensitometric responses of dental radiographic films. Dento-Maxillo-Facial Radiology 20, 11-16. 1 1 9 Price, C , and Radan, E. (1997). A n endodontic file model to compare geometric and digital unsharpnesses in dental radiographs. In Advances in Maxillofacial Imaging, A . Farman, G , ed. (Amsterdam: Elsevier), pp. 271-278. Reit, C., and Hollender, L . (1983). Radiographic evaluation of endodontic therapy and the influence of observer variation. Scandinavian Journal of Dental Research 91,205-12. Reukers, E., Sanderink, G. , Kuijpers-Jagtman, A . M . , and van't Hof, M . (1998). Assessment of apical root resorption using digital reconstruction. Dento-Maxillo-Facial Radiology 27, 25-9. Richards, A . G. , and Colquitt, W. N . (1981). Reduction in dental X-ray exposures during the past 60 years. Journal of the American Dental Association 103, 713-718. Rontgen, W . C . (1896). O n a N e w K i n d of Rays. Translated by Stanton, A . Nature 53, 274-277. Sanderink, G . C , Huiskens, R., van der Stelt, P. F., Welander, U . S., and Stheeman, S. E. (1994). Image quality of direct digital intraoral x-ray sensors in assessing root canal length. The RadioVisioGraphy, V i s u a l i x / V I X A , Sens-A-Ray, and Flash Dent systems compared wi th Ektaspeed films. Oral Surgery, Oral Medicine, Oral Pathology 78,125-32. Scarfe, W. C , Fana, C , Jr., and Farman, A . G . (1995). Radiographic detection of accessory/lateral canals: use of Radio VisioGraphy and Hypaque. Journal of Endodontics 21,185-90. 120 Scarfe, W. C. , Farman, A . G. , Brand, J. W. , and Kel ly , M . S. (1997). Tissue radiation dosages using the R V G - S wi th and without niobium filtration. Austral ian Dental Journal 42, 335-42. Scarfe, W . C., Norton, S., and Farman, A . G . (1995). Measurement accuracy: a comparison of two intra-oral digital radiographic systems, RadioVisiography-S and FlashDent, wi th analog film. Dento-Maxillo-Facial Radiology 24, 215-20. Shearer, A . C. , Horner, K . , and Wilson, N . H . (1990). Radiovisiography for imaging root canals: an in vitro comparison wi th conventional radiography. Quintessence International 21, 789-94. Shearer, A . C. , Horner, K . , and Wilson, N . H . (1991). Radiovisiography for length estimation in root canal treatment: an in-vitro comparison wi th conventional radiography. International Endodontic Journal 24, 233-9. Svanaes, D . B., Moystad, A . , Risnes, S., Larheim, T. A . , and Grondahl, H . G . (1996). Intraoral storage phosphor radiography for approximal caries detection and effect of image magnification: comparison wi th conventional radiography. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 82, 94-100. Svenson, B., Grondahl, H . G. , and Lindval l , A . M . (1990). Influence of film fog on radiographic caries diagnosis. Dento Maxil lo Facial Radiology 19,105-8. Tirrell, B. C., Miles, D . A . , Brown, C. E., Jr., and Legan, J. J. (1996). Interpretation of chemically created lesions using direct digital imaging. Journal of Endodontics 22, 74-8. 121 Tyndall , D . A . , Clifton, T. L . , Webber, R. L . , Ludlow, J. B., and Horton, R. A . (1997). T A C T imaging of primary caries. Oral Surgery, Oral Medicine, Ora l Pathology, Oral Radiology, & Endodontics 84, 214-25. Tyndall , D . A . , Lud low, J. B., Platin, E., and Nair , M . (1998). A comparison of Kodak Ektaspeed Plus fi lm and the Siemens Sidexis digital imaging system for caries detection using receiver operating characteristic analysis. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 85,113-8. Vande Voorde, H . E., and Bjorndahl, A . M . (1969). Estimating endodontic "working length" wi th paralleling radiographs. Oral Surgery, Oral Medicine, Oral Pathology 27,106-10. Velders, X . L . , Sanderink, G . C , and van der Stelt, P. F. (1996). Dose reduction of two digital sensor systems measuring file lengths. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 81, 607-12. Versteeg, C. H . , Sanderink, G . C , Geraets, W. G. , and van der Stelt, P. F. (1997). Impact of scale standardization on images of digital radiography systems. Dento-Maxillo-Facial Radiology 26, 337-43. Versteeg, C. H . , Sanderink, G . C., Lobach, S. R., and van der Stelt, P. F. (1998). Reduction in size of digital images: does it lead to less detectability or loss of diagnostic information? Dento-Maxillo-Facial Radiology 27, 93-6. 1 2 2 Versteeg, C . H . , Sanderink, G . C., and van der Stelt, P. F. (1997). Efficacy of digital intra-oral radiography in clinical dentistry.. [Review] [63 refs]. Journal of Dentistry 25, 215-24. Versteeg, C . H . , Sanderink, G . C., van Ginkel , F. C., and van der Stelt, P. F. (1998). Effects of calibration and automatic greyscale adjustment on detectability of simulated bone lesions using a storage phosphor system. Dento-Maxillo-Facial Radiology 27, 240-4. Versteeg, C . H . , Sanderink, G . C., van Ginkel , F. C., and van der Stelt, P. F. (1998). A n evaluation of periapical radiography wi th a charge-coupled device. Dento-Maxillo-Facial Radiology 27, 97-101. Versteeg, K . H . , Sanderink, G . C., van Ginkel , F. C., and van der Stelt, P. F. (1997). Estimating distances on direct digital images and conventional radiographs. Journal of the American Dental Association 128, 439-43. Versteeg, K . H . , Sanderink, G . C., Velders, X . L. , van Ginkel , F. C. , and van der Stelt, P. F. (1997). In v ivo study of approximal caries depth on storage phosphor plate images compared wi th dental x-ray film. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 84, 210-3. Visser, H . , and Kruger, W . (1997). Can dentists recognize manipulated digital radiographs? Dento-Maxillo-Facial Radiology 26, 67-9. Vrijens, N . M . P., and Strommer, P. (1997). Physical evaluation of the Planmeca digital panoramic system. Dento-Maxillo-Facial Radiology 26, 270-271. 1 23 Webber, R. L . , Horton, R. A . , Tyndall , D . A . , and Ludlow, J. B. (1997). Tuned-aperture computed tomography (TACT). Theory and application for three-dimensional dento-alveolar imaging. Dento Maxi l lo Facial Radiology 26, 53-62. Webber, R. L . , Horton, R. A . , Underbi l l , T. E., Ludlow, J. B., and Tyndall , D . A . (1996). Comparison of fi lm, direct digital, and tuned-aperture computed tomography images to identify the location of crestal defects around endosseous titanium implants. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 81, 480-90. Wenzel, A . (1998). Digital radiography and caries diagnosis. Dento-Maxillo-Facial Radiology 27, 3-11. Wenzel, A . , Borg, E., Hintze, H . , and Grondahl, H . G . (1995). Accuracy of caries diagnosis in digital images from charge-coupled device and storage phosphor systems: an in vitro study. Dento-Maxillo-Facial Radiology 24, 250-4. Wenzel, A . , Gotfredsen, E., Borg, E., and Grondahl, H . G . (1996). Impact of lossy image compression on accuracy of caries detection in digital images taken wi th a storage phosphor system. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics 81, 351-5. Wenzel, A . , and Hintze, H . (1993). Perception of image quality in direct digital radiography after application of various image treatment filters for detectability of dental disease. Dento Maxi l lo Facial Radiology 22, 131-4. Whaites, E. , and Brown, J. (1998). A n update on dental imaging. British Dental Journal 185, 166-72. 1 24 White, S. C , and Yoon, D . C. (1997). Comparative performance of digital and conventional images for detecting proximal surface caries. Dento Maxi l lo Facial Radiology 26, 32-8. Yoshioka, T., Kobayashi, C , Suda, H . , and Sasaki, T. (1997). Quantitative subtraction wi th direct digital dental radiography. Dento-Maxillo-Facial Radiology 26, 286-94. Yoshiura, K . , Stamatakis, H . , Shi, X . Q., Welander, U . , M c D a v i d , W . D. , Kristoffersen, J., and Tronje, G . (1998). The perceptibility curve test applied to direct digital dental radiography. Dento-Maxillo-Facial Radiology 27,131-5. 1 25 Appendix Price, C , and Radan, E v "An endodontic file model to compare geometric and digital unsharpness in dental radiographs." In "Advances in Maxillofacial Imaging" (Amsterdam, Elsevier, 1997). 271-8 1 26 An endodontic file model to compare geometric and digital unsharpness in dental radiographs Colin Price and Elham Radan Faculty of Dentistry, University of B.C. 2199 Wesbrook Mall , Vancouver, B.C., V6T 1Z3, Canada 1. INTRODUCTION It is recognized that image quality is dependent on three factors: contrast, resolution, and noise [1], When image quality is evaluated, the choice of model has considerable bearing on the relative weighting of these factors. Dental caries models have been used by many to compare different imaging systems [2-51. In most cases these models have failed to show significant differences between systems. Wenzel and others [6] showed surprisingly little difference in the diagnostic accuracy of caries with severely compressed digital image files. The interpretation of dental caries is a common and relevant task, but, because of the ill-defined edge of the lesions, it depends more on broad area contrast than on resolution. Focal spot characteristics and geometric unsharpness have been given scant attention in the field of oral radiology, in spite of their obvious relevance to image geometry and quality. Platin and others [7] recently studied the effects of focal spot size on diagnostic accuracy, and failed to show any relevance. However, they selected focal spots with relatively little difference, did not design the image geometry to maximize the unsharpness, and used a caries model, which we regard as a poor evaluator of resolution. Models using endodontic files have been used on several occasions [8-11]. We decided to explore the use of such a model to study geometric unsharpness and digital image degradation. This work was intended to be a pilot study, a precursor to a more detailed investigation, but several findings are of interest, and we offer them in this preliminary report. 127 2. MATERIALS AND METHODS 2.1. Test Objects We placed endodontic files of sizes 8, 10, and 15 in the root canal of an extracted lower incisor tooth. The number 8 file protruded beyond the apex and the other two files were short of the apex. No further attempts were made to establish the precise positions of the files. The tooth was mounted on a piece of vacuum-form splint material 2 mm thick. A series of radiographs was produced with the number 8 file before instrumentation with the other files. 2.2. Radiography A test column, consisting of three vertical steel rods and four horizontal aluminium platforms, served to standardize the relationship of the x-ray source, object, and film. Two of the platforms were adjustable, and their positions could be established, by a Vernier scale and a vertical steel ruler, with a precision of 0.1 mm. A General Electric G E 100 x-ray unit (General Electric Company, Milwaukee, WI, USA) was secured in a reproducible position at the top of the column. We determined the precise position of the focal spot by triangulation, using images of a pewter plate with a grid of 300 pirn holes at 5 mm centres. Ektaspeed Plus films (Kodak Canada Inc., Toronto, Ontario, Canada) were placed 400 mm from the focal spot. One series of radiographs was produced with the root placed as close as possible to the films. This smallest file to film distance was approximately 5 mm, producing a magnification of 1.01, and a negligible geometric unsharpness. The distance between the root and films was altered to give increased magnification and concomitant greater geometric unsharpness. File to film distances of 50 and 100 mm were used to give magnifications of 1.14 and 1.33. Processing was by an A T 2000 processor (Air Techniques Inc., Hicksville, N Y , USA) at 28 °C for a total processing time of 5.5 min. using Kodak Readymatic chemistry. Exposure factors were 90 kVp, 15 mA, 3.5 mm aluminii'm total filtration, and 8 cycles. This produced radiographs with a density of approximately 2.0 in the unattenuated part of the image, and approximately 1.2 in the region of the dentine of the root. 128 23. Digitizing We scanned the 9 radiographs with a Hewlett Packard 3c flatbed scanner and a 4c transparency adapter (Hewlett Packard Company, Greeley, CO, USA) as 8 bit greyscale images at resolutions of 150, 300, and 600 dots per inch (DPI). The digitized images were cropped to areas of 4.2 x 12.7 mm (0.167 x 0.5 in.) and imported into NIH Image Program (Wayne Rasband, National Institutes of Health, USA) in randomly allocated sets of four. The unattenuated parts of the images had pixel values of approximately 200, and areas of dentine of the root, approximately 100. 2.4. Interpretation Five of the 27 images were randomly selected and copied to give a total of 32 images in eight groups of four. These were presented to observers in a PowerPoint Program (Microsoft Corporation, Redmond, W A , USA), with each group of four images filling the computer screen (230 x 170 mm approximately), representing a magnification of approximately x 13. We selected 12 observers from four groups of individuals: three lay persons with little or no knowledge of dentistry, three senior dental students, three dentists, and three "specialists" - one oral radiologist, one endodontist, and one oral radiology graduate student. Each observer viewed the images on the computer screen in a room with subdued lighting. No time limits were set, and observers were allowed to vary their distance from the screen, but not to adjust the contrast, brightness, nor any software features. They were instructed, using a group of four images, to mark the end of the file and the end of the root, with the marking facility of the PowerPoint Program, prior to observing the 32 images. The marked images were captured to memory. 2.5. Analysis The marked images were opened in NIH Image Program. The coordinates of each mark identifying file and root ends were obtained, as well as a direct determination of the distances between the marks. The distances obtained from the program were confirmed with a scanned image of a millimetre ruler in both horizontal and vertical planes. The coordinates were used to obtain the distances between the marks. There was very close agreement between these values and those obtained by direct measurement. In the case of those images obtained with magnifications of 1.14 and 1.33, the values of distance were divided by these factors. Means and standard deviations were derived, and the differences in distance were compared with Analysis of Variance. 129 3. RESULTS 3.1. Focal Spot The left side of Figure 1 shows an image of the focal spot of the x-ray tube used in this investigation, produced with a 300 pim pinhole placed midway between the source and a Sens-A-Ray sensor (Regam Medical Systems, Sundsvall, Sweden). On the right side of this illustration, the corresponding image from a Belmont Phot-X x-ray unit (Belmont Equipment Corp., Somerset, NJ., USA) is shown for comparison. The lower half of the Figure shows images of a brass washer with an aperture of 5 mm placed in the same relationship. In all instances, the source to receptor distance was 1,000 mm. These images show the differences in size and energy distribution of the focal spots, and the different geometric unsharpnesses produced by them. Figure 1. Top: Focal spot images of General Electric G E 100 (left) and Belmont Phot-X (right). (Source to receptor: 1000 mm, 300 pirn pinhole at 500 mm.) Bottom: Images of 5 mm aperture in a brass washer. (Geometry and x-ray units as above.) 3 m m 3.2. Effects of Geometric and Digital Unsharpnesses Figure 2 shows the mean differences in distance recorded between the apex of the root and the end of the files for different digital resolutions. The geometric unsharpnesses (magnifications) have been pooled in this illustration. The bars represent Standard Errors. Figure 3 indicates the same values for the different geometric unsharpnesses, with pooled digital resolutions. In all cases, the mean differences for the number 15 file are very small. Figures 4 and 5 show the effects of digital and geometric unsharpnesses for files 8 and 15 separately. The same scaling has been used in these two illustrations to facilitate comparison. E E co o c CO CD CO -= c Q a to cu Figure 2 Mean difference in distances for each digital resolution (pooled magnifications). co E c £ cu CO CO 2~ o — c 1 Q CO 1 if) s5 CO «3 c F i le 8 F i l e 10 • F i le 15 Figure 3 1 . 1 4 1 , M a g n i f i c a t i o n Mean difference in distances for each magnification (pooled digital resolutions). Bars represent Standard Errors. ~ File 8 1 5 0 Figure 4 • 1.14 • 1.33 3 0 0 DPI X 6 0 0 Mean difference in distances for each digital resolution and each magnification, file 8. 1 5 0 Figure 5 3 0 0 DPI 6 0 0 Mean difference in distances for each digital resolution and each magnification, file 15. Bars represent Standard ETors. Table 1 gives A N O V A for the differences in distance, and Table 2 the Bonferroni p-values to separate the different effects of the three digital resolutions and the three geometric unsharpnesses. Table 1 Analyses of Variance: Difference in Distance. File 8 File 10 File 15 Source Degrees of Freedom F P F P F P DPI 2 22.04 <0.00001 4.19 0.02 0.08 0.92 Mag. 2 86.77 <0.00001 17.11 <0.00001 20.68 <0.00001 Observer 11 7.48 <0.00001 5.23 0.00003 1.12 0.37 DPI x Mag. 4 5.02 0.002 0.72 0.58 3.02 0.03 Mag. x Obs. 22 7.09 <0.00001 1.24 0.27 1.04 0.44 Error 44 Table 2 Bonferroni p-values: Difference in Distance. File 8 File 10 File 15 DPI 150 300 150 300 150 300 300 <0.00001 0.02 >0.99999 600 0.00009 0.26 0.21 0.98 >0.99999 >0.99999 Mag. 1.00 1.14 1.00 1.14 1.00 1.14 1.14 0.61 0.9 0.34 1.33 <0.00001 <0.00001 0.0002 0.00001 <0.00001 0.00011 4. DISCUSSION The model employed in this study functioned well for the comparison of the effects of geometric and digital unsharpnesses, and we are proceeding with a more definitive evaluation, with extended ranges of file sizes, file positions, and geometric and digital unsharpnesses. The relative lack of effect of quite severe unsharpness on the determination of the position of the number 15 file was unexpected. Only the magnification of 1.33 gave a significant difference with this file, and the mean difference in distance was less than 0.2 mm, which would have no clinical relevance. Furthermore, this degree of geometric unsharpness is beyond that encountered in a clinical situation. The smaller files also gave very acceptable precision of measurement with all except this extreme unsharpness. These results cause us to question the need for concern about the effects of geometric unsharpness in clinical dental radiography, and the empirical choice of cone length. The lack of significant difference between digital resolutions of 300 and 600 DPI for any .of the files was a further surprise. Few diagnostic tasks in dentistry are more demanding on resolution, and the quest for greater digital resolutions than 300 DPI appears pointless. One observer in the lay group gave outlying values for the number 10 file, and consequently this group performed significantly inferiorly for that file. Otherwise, there were no significant differences between the groups, which implies that these tasks are not dependent on training or experience. (File 8:- F a 8 4 ) = 1.35, p = 0.26; File 10:- F ( 3 , 8 4 ) = 3.70, p = 0.01; File 15:- F( 3 g4) = 0.63, p = 0.60.) Differences between the first and second readings of the 5 Feplicated images were not significant ( F ^ = 0.64, p = 0.43). Our model avoided superimposition of bone and soft tissues, and the complexities attending the imaging of multiple-rooted teeth, all of which can be expected to reduce the precision of determination of file position. However, we have been able to compare the effects of these unsharpnesses more reliably by this design, and the findings are applicable to clinical situations when more noise is present. The addition of further degrading factors is unlikely to accentuate the effects we have shown, though it might diminish them. 1 3 3 R E F E R E N C E S 1. T. Okano, J.D. Wiebe, R.L. Webber and R.R Wagner, Oral Surg Oral Med Oral Pathol, 55 (1983) 527. 2. A . Wenzel, H. Hintze, L. Mikkelsen and F. Mouyen, Oral Surg Oral Med Oral Pathol, 72 (1991) 621. 3. S.C. White, L. Hollender and B . M . Graft, J Am Dent Assoc, 108 (1984) 755. 4. H. Hintze, A . Wenzel and C. Jones, Caries Res, 28 (1994) 363. 5. D.B. Svanaes, A . Moystad, S. Risnes, T.A. Larheim and H.-G. Grondahl, Oral Surg Oral Med Oral Pathol Oral Radiol Endod, 82 (1996) 94. 6. A . Wenzel, E. Gotfredsen, E. Borg and H.-G. Grondahl, Oral Surg Oral Med Oral Pathol Oral Radiol Endod, 81 (1996) 351. 7. E. Platin, S. Mauriello and J.B. Ludlow, Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics, 81 (1996) 235. 8. T. Okano, H.J. Huang and T. Nakamura, Oral Surg Oral Med Oral Pathol, 59 (1985) 543. 9. X . L . Velders, G.C. Sanderink and P.F. van der Stelt, Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, & Endodontics, 81 (1996) 607. 10. G.C. Sanderink, R. Huiskens, P.F. van der Stelt, U.S. Welander and S.E. Stheeman, Oral Surg Oral Med Oral Pathol, 78 (1994) 125. 11. A .C . Shearer, K. Horner and N.H. Wilson, International Endodontic Journal, 24 (1991) 233. 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0089020/manifest

Comment

Related Items