0099-2399/90/1604-0173/$02.00/0 JOURNAL OF ENDODONTICS Copyright 9 1990 by The American Association of Endodontists
Printed in U.S.A.
VOL. 16, NO. 4, APRIL1990
Digital Subtraction Radiography for Detecting Cortical and Cancellous Bone Changes in the Periapical Region Donald A. Tyndall, DDS, PhD, Stanley F. Kapa, DMD, MS, and Charles P. Bagnell, PhD
angiography system became available in the late 9170's and is now a common imaging modality in medical radiography
Previous studies have found that bony lesions cannot be visualized on conventional radiographs unless there is cortical plate involvement. The aim of this project was to compare the sensitivity of digital subtraction to conventional radiography for detecting periapical changes in cortical and cancellous bone. Using a long source-to-object X-ray technique and E-speed film, serial radiographs of a dry skull mandible were obtained. Two bone lesions per radiograph were simulated using #1 to 8 round burs. Conventional and digitally subtracted images were evaluated for lesion presence by a board of reviewers. The results demonstrated greater sensitivity scores for digitally subtracted images in identifying cortical and cancellous bone changes. The lower limit of detection was less for digitally subtracted images in cortical and cancellous bone as well.
(6). Digital image subtraction in dentistry had been investigated by Webber et al. (5) at the National Institutes of Dental Research where applications were made for the assessment of subtle periodontal bone changes. Results of these studies were first published in the dental literature in 1982 (5). Subsequent investigations and publications by this group and others (notably Grondahl) have focused on the application of digital subtraction radiography in the detection of subtle periodontal bone changes and the evaluation of periodontal treatment (7-
12). Other diagnostic situations occur in dentistry where small changes in bone may not be detected when structured noise prevents adequate detection from conventional radiographs. Such a situation is the evaluation of periapical radiographs for assessing changes in periapical bone resulting from pulpal inflammation. There is currently very little literature on any application of digital subtraction radiography to periapical bone changes. Probably the most significant work in this area was recently performed by Kullendorff et al. (13). The successful use of this technique in the diagnosis of periodontal conditions and the work of Kullendorffet al. indicates that digital subtraction radiography could be useful in the early detection of subtle bony changes in the periapical region. The initial success with this technique in periodontal diagnosis and treatment and the paucity of literature on its applications to periapical changes provided the impetus for this investigation. The aims of this investigation were to apply digital subtraction radiography methods to the detection of subtle bony changes in the periapical region. Specifically, digital subtraction was compared with conventional radiographic techniques in a laboratory model for sensitivity in detecting changes in both cortical and cancellous bone.
In 1961 Bender and Seltzer (1, 2) and, more recently, van der Stelt (3) found that lesions in bone could not be visualized on normal radiographs unless there was sufficient change to, or perforation of, the cortical plate. Damage confined to cancellous bone did not produce a recognizable radiolucency on standard radiographs. It is estimated that 30 to 50% of the mineralized component of bone must be lost before the loss is evident on conventional radiographs (1-3). Therefore, extensive disease could be present and not be detected by a conventional radiographic examination. The object of the digital subtraction technique in radiography is to depict diagnostic information without background noise or those structures with no diagnostic significance (structured noise) (4). An advantage of digital subtraction radiography is that it reduces the sources of error inherent in photographic subtraction by minimizing or eliminating the irrelevant differences produced by film nonlinearity, exposure differences, and processing variations (5). Investigations with computers and digitization of X-ray images from film began in the late 1960's. The first commercial digital subtraction
MATERIALS AND METHODS A long source-to-object head positioning technique as suggested by Jeffcoat et al. (14) was utilized in this study. The photon source was a Franklin skull unit with a rotating anode X ray tube (Machlette Dynamax; Transworld X-Ray Corp., Charlotte, NC). A modified head positioning device and
173
174
Tyndall et al.
Masel Precision (Masel Orthodontics Inc., Philadelphia, PA) intraoral film-holding device were used to hold sections of human mandible (model 24-7020; Carolina Biological Supply, Burlington, NC). The head positioning device has angle measurements for precise repositioning of the human skull and mandibular phantoms. The X-ray source-to-object distance was 84 inches, This cephalometrically related method was used for two reasons. Theoretically, the greater the sourceto-object distance the greater the image sharpness. This is because as the distance from the X-ray source-to-object increases, a smaller penumbra is produced, resulting in a sharper image. In theory, an infinite distance would result in parallel X-rays but this technique was selected because there is a practical distance beyond which the human eye can no longer statistically perceive any improvement in the image sharpness. Also, the nearer parallel the X-rays are the greater the reduction of film, patient, and X-ray source misregistration errors (14). Second, this technique was selected because this type of methodology will most likely be used in the clinical situation as opposed to the more costly and time-consuming stent method (14). The latter method requires fabrication of a filmholding stent registered to the patient's occlusion. A dry mandible with a removable cortical "window" of bone (thus exposing cancellous bone and apices) was used as the object to be imaged (Fig. 1). An experimental protocol related to that used previously by Bender and Seltzer (1 2) and Van der Steit (3) was utilized. Artificial lesions were simulated at two separate sites by #1, 2, 4, 6, and 8 round dental burs. These burs correspond to the following sizes, respectively (as measured by a micrometer gauge), 0.64, .67, 1.28, 1.41, and 1.93 ram. Two lesions were placed in mandibular interradicular and periapical cortical bone for each lesion size. Each lesion was created by the entire depth of the bur. In a separate procedure the cortical bone window was removed and two lesions were placed in cancellous bone in periapical and interradicular bone for each bur size. The interradicular lesion was placed posterior to the
Journal of Endodontics
periapical lesion for both cortical and cancellous bone lesions. The window was then secured in place. Images were obtained before lesion simulation and after each bony change created by progressively larger dental burs. The reader is reminded that the lesion sites were constant since each bur removed bone from the same two locations. E-speed film (Eastman Kodak Inc., Rochester, NY) was used along with the previously described long source-to-object imaging system. Processing of the film was performed according to manufacturer's recommendations under strict quality control conditions. Three films of each bur hole size lesion were obtained for each ~ t of experimental lesions including the unaltered mandible. This yielded a total of two lesions per film with three images of each altered mandible for five bur sizes, generating a total of 15 images to evaluate for bone loss. Imaging parameters for each technique were set as follows: 75 kVp, 400 MAS with 2.7 mm aluminum equivalent filtration. Image acquisition was accomplished by a CCD camera (Panasonic WV-C050; Panosonic Industrial Co., Secaucus, N J). t-ach film was placed on a masked viewbox of uniform intensity. Image processing was performed on a Digital Equipment Corporation computer (Digital Equipment Corporation DEC LSI 11/73 with an R T I I operating system; Digital Equipment Corp., Maynard, MA) with an Imaging Technology (Imaging Technology 5 i 2 System with ITEX/Q Software; Imaging Technology Inc., Woburn, MA) digital processing system. Figure 2 demonstrates the entire image processing system. Digital images of each radiograph were acquired by averaging 128 video frames from the CCD camera, each frame consisting of a 512 x 512 pixel matrix. Images were processed through two-frame grabbers of 8 and 16 bits in depth. The final images contained 256 gray levels. A gamma correction program developed by Ruttimann et al. (15) was used to correct for contrast differences between film sets. Digital image subtraction was achieved as follows. The unaltered mandibular radiograph was stored in the computer's memory and displayed on the RGB monitor. The radiograph to be subtracted was imaged on the same monitor by
, 23-;
F~G 1. The mandible with a "bone window" which was utilized in the investigation. Cancellous bone lesions were made in the cancellous bone which was rendered accessible by the opened window such that no cortical bone was involved. Cortical bone lesions were made on the facial aspect of the cortical plate of bone on the contralateral side.
FIG 2. The digital subtraction image processing system used for this investigation. The CCD camera and viewbox are on the left. images were viewed by all observers on the RGB monitor seen on the right.
Digital Subtraction Radiography
Vol. 16, No. 4, April 1990
placing it on the viewbox and under the CCD camera and manipulated such that its real time image was superimposed on the stored image of the original radiograph. A real-time subtraction program was implemented such that when superimposition occurred, both images cancelled each other out and no anatomical structures were discernible. The second (postlesion simulation) image was acquired, gamma corrected, and subtracted from the first image. Registration of images was checked by examining standard deviations associated with first order histograms of the subtracted images. The images were then subtracted. Subtracted images were then enhanced by a linear stretch algorithm which spread the narrow peak of gray levels (centered around 128) across all 256 gray levels. Figure 3 demonstrates an image set consisting of the unaltered and postaltered mandibular radiographs for cortical (top) and cancellous bone (bottom). Figure 4 reveals the first order histogram of the original image, the linear stretch histogram for lesion enhancement, and the final enhanced subtracted images depicting cortical and cancellous bone changes. All images (conventional and enhanced subtracted images) were coded and stored on hard disc memory. Three observers (all radiologists) independently evaluated each image for simulated pathosis twice. Image viewing was accomplished under darkened room conditions. Each image was randomly called up on the screen by one of the investigators while the observer/radiologist answered a simple questionnaire regarding probable or definite presence of each (posterior or anterior) simulated lesion. These procedures were repeated for each image yielding two observer responses for each image. In summary, since there were two simulated lesions on each film, 15 total films (three of each bur hole size), and three observers with two responses each, there were a total of 36 responses per bur size for cortical and cancellous bone, respectively. A sensitivity calculation was carried out for each type of
~!'~ I > . :~ ':
175
~ i
f~ ,
?
FiG 4. Representative histograms and final subtracted images generated from this investigation. The top left image demonstrates a typical gray level histogram from a conventional film. The top right shows a linear stretched histogram where the gray levels from a first order histogram of a subtraction were spread out over 256 gray levels. The lower left image demonstrates a subtracted image of bur #6 cortical bone lesions (arrowheads) while the lower right image demonstrates the presence of bur # 6 cancellous bone lesions on a subtracted image (circled areas). Both images have undergone linear stretching enhancement.
imaging system (conventional versus subtraction). Responses from each observer/radiologist were checked against the known presence or absence of a simulated lesion. The number of total observer responses yielded a measure of sensitivity. Here sensitivity is defined as the ability of an imaging system to identify a true positive finding as a positive finding. RESULTS Table 1 demonstrates the overall combined sensitivity scores of images with lesions in cortical and cancellous bone for conventional versus subtracted radiographs. Two categories are presented: a response combining (a) lesion probably present and lesion definitely present and (b) lesion definitely present. Sensitivity scores in this table are limited to total responses in cortical and cancellous bone with no regard to lesion size. Table 2 reveals the sensitivity responses for cortical bone lesions only for each bur size. Once again data are presented in terms of a combined response (probably and definitely present) as well as a single response, definitely present. Table 3 reveals sensitivity responses for cancellous bone using the same format as in Table 2.
FIG 3. The top left image depicts a conventional radiograph of an unaltered mandible before cortical bone lesion simulation. The top right image shows the same portion of mandible with a # 6 bur lesion made in cortical bone in two locations (interradicular and periapical bone). The bottom left image depicts a conventional radiograph of an unaltered mandible before cancellous bone lesion simulation. The image at the lower right shows the same portion of mandible with a # 8 bur ~esionmade in cancellous bone in two locations (interradicular and periapical bone).
DISCUSSION An examination of Tables 1 to 3 reveals several interesting findings. In Table 1 it can be seen that digital subtraction radiography generated greater sensitivity scores (ability to detect true positive finding) than conventional radiography. This finding holds true for both categories of responses. For cortical bone, the overall sensitivity scores were 0.756 to 0.594
176
Tyndall et al.
Journal of Endodontics
TABLE 1. Sensitivity scores for cortical and cancellous bone for both observer response categories (probably and definitely present)
Cortical bone lesions (180 total responses) Probably present and definitely present Definitely present Cancellous bone lesions (180 total responses) Probably present and definitely present Definitely present
Conventional Radiographs
Digital Subtraction Images
0.594 0.322
0.756 0.706
0.106 0.00
0,605 0.344
TABLE 2. Sensitivity scores for cortical bone according to bur size for both observer response categories (38 responses per bur size)
Probably and Definitely Present Bur Size
#1 #2 #4 #6 #8
Conventional
Definitely Present
Digital Subtraction Images
Conventional
Radiographs
Radiographs
Digital Subtraction Images
0.083 0.250 0.667 0.972 1.00
0.083 0.667 1.00 1.00 1.00
0,00 0.056 0.167 0,389 1.00
0.055 0.472 1.00 1.00 1.00
TABLE 3. Sensitivity scores for cancellous bone according to bur size for both observer response categories (36 responses per bur size)
Probably and Definitely Present Bur Size
#1 #2 #4 #6 #8
Conventional
Definitely Present Conventional
Radiographs
Digital Subtraction Images
Radiographs
Digital Subtraction Images
0.083 0.111 0.167 0.00 0.167
0,444 0.444 0.500 0.722 0.917
0.00 0.00 0.00 0.00 0.00
0.166 0_111 0.250 0.583 0.611
when the lesions were judged to be probable or definite for subtraction and conventional radiography, respectively. Interestingly, the subtraction sensitivity scores for cortical bone, remain about the same (0.706) when observer response to lesion presence was definite (versus 0.322 for conventional radiography). That the overall subtraction sensitivity scores remained so high is probably due to the fact that in subtraction radiography structured noise (normal anatomy) is eliminated, thereby revealing the bone change (black) against a rather gray, homogenous background (Fig. 4). This is in contrast to conventional radiography where bone changes are superimposed against the visually variable background of normal anatomy (structured noise). Digital subtraction demonstrated superiority over conventional radiography 0.605 to 0.106, respectively, for cancellous bone sensitivity scores. This is a finding of potential significance in that previous reports have demonstrated the inability of conventional radiographs to detect cancellous bone loss (11, 13). It also agrees with recently published data by Kullendorff et al. (13). It is noteworthy that when the lesions were felt to be definitely present, the digital subtraction score was
noticeably lower (0.344) although still greater than conventional radiography (0.00). This is probably due to the fact that cancellous bone changes revealed in subtraction images are more subtle and difficult to identify than cortical bone changes (Fig. 4). Tables 2 and 3 break down the sensitivity scores according to the size bur used to make the bone lesion. The reader is reminded that the lesion was generated by placing the bur up to its shank into the bone. From these tables it is easier to identify the estimated lower limits of detection (LED) for both imaging systems. For lesions in cortical bone (Table 2), it can be seen that the subtraction images yielded superior scores to conventional radiographs in most bur sizes except where the lesions were so large that either system detected them with t00% sensitivity. The smallest bur size lesions detectable, when both observer responses are combined, were between #2 and #4 and #4 and #6 for digital subtraction and conventional radiographs, respectively. When the observers felt that the lesions were definitely present, the digital subtraction scores remained about the same while the conventional radiograph scores
Vol. 16, No. 4, April 1990
dropped precipitously. The LLD for this category was between the #2 and #4 burs for subtraction and at the #8 bur size for conventional radiography. It should be noted that the #1 bur lesions appear below the estimated LLD for both systems as the sensitivity scores were approximately the same for both imaging systems. For lesions in cancellous bone (Table 3) digital subtraction's superior performance over conventional radiography is even more pronounced. The LLD for digital subtraction, when both observer responses are combined, is between the #4 and #6 bur. Certainly, at the #8 bur hole size the sensitivity scores indicate that cancellous bone lesions can be identified with digital subtraction radiography. This bone loss is less than the amount detected in conventional radiography by Van der Stelt (3). When the category is "lesion definitely present," the subtraction scores drop noticeably, again reflecting their more subtle appearance against the subtraction background (Fig. 4). It is noteworthy that the observers essentially could not detect cancellous bone loss from conventional films no matter which category of observer response was used, a finding partially substantiated by the literature (1-3). There are a few troublesome data points in Table 3 that should be carefully addressed. For the #1 size bur in the "probable and definite" category, digital subtraction scored 0.444, which was greater than that for cortical bone loss. Also, the #6 bur conventional radiograph score in the same observer response category was 0.00, lower than the smaller bur sizes. Finally, in the "definitely present" category for digital subtraction, the #2 bur lesion produced a lower sensitivity score than the #1 bur lesion. All of these apparent discrepancies are probably due to statistical fluctuations. Presumably, the difficulty in identifying cancellous bone changes (Fig. 4) would probably not appear with a larger sample size. It is not surprising that digital subtraction was found to be superior to conventional radiography. This finding is consistent with the many investigations designed to detect cortical bone loss in periodontal patients (7, 10, 16). That cancellous bone loss was detected is an important finding, considering the difficulty of doing so with conventional radiography (13). Hence, the information from this investigation may be beneficial to the clinical practice of general dentistry and endodontics. Clinically, this may mean that bone changes associated with pulpal pathosis may be detectable at a much earlier time. This may have potential treatment and/or prognosis implications. There are other clinical situations where digital subtraction may be useful in the practice of endodontics. These include the monitoring of bone healing (since "subtraction" can detect bone additions subsequent to endodontic therapy), the evaluation of internal and external resorption, and the evaluation of the so-called and diagnostically confusing "periapical scars," present many years after successful endodontic therapy. The advantages of digital subtraction appear to lie in its ability to remove structured noise, thereby allowing detection of changes that the human eye cannot see on conventional radiographs (3, 5). There are several disadvantages to digital subtraction that need to be considered also. The most critical component of a subtraction system is geometric reproducibility of the X-ray source-to-object relationship (5, 17). This is very difficult to achieve with intraoral stents and dental Xray tubes, A new, long source-to-object cephalometric head
Digital Subtraction Radiography
177
holder approach developed by Jeffcoat et al. (14) holds promise for clinically relevant subtraction methodology. Once geometric reproducibility is assured, a video camera, personal computer, and the appropriate software are all that will be needed to produce subtraction images. This would make digital subtraction a viable imaging method in institutions or large practices, if not in smaller private practices. There are several limitations to this study that bear mentioning. One is that these images were obtained under rigid laboratory-controlled conditions in which geometric reproducibility was not a significant issue. This was also true for earlier work investigating periodontal lesions. However, clinical applications of subtraction methods have proved useful in assessment of clinical periodontal bone change (loss or gain). It can be assumed that minor adaptations to periapical films (the periodontal studies utilized easier to place bitewing films) will make these laboratory methods useful in clinical endodontic studies. Second, the method of analysis for comparing the two imaging systems of this study only evaluated sensitivity. Future work should also include more rigorous forms of assessing image system performance such as receiver operating characteristics analysis complete with positive and negative predictive values (18). Consequently, any conclusions drawn from this study can relate only to the potential superiority of digital subtraction over conventional radiography for the detection of periapical lesions. For a complete assessment, specificity (ability to detect Iack of disease presence) should be included in further evaluations of digital subtraction radiography. This investigation has demonstrated the potential merits of digital subtraction radiography over conventional imaging in detecting changes in both cortical and cancellous bone. That cancellous bone changes were detectable by digital subtraction imaging is an important finding in itself and holds promise for enhancing the diagnostic methods available to the practitioner for the detection and assessment ofsubtle bony changes in the periapical region. This project was supported in part by NIH Grant 2-507-RR0533. Dr. Tyndall is affiliated with the Department of Diagnostic Sciences, University of North Carolina at Chapel Hill, School of Dentistry, Chapel Hill, NC. Dr. Kapa is affiliated with the Department of Diagnostic Services, University of Pittsburgh School of Dentistry, Pittsburgh, PA. Dr. Bagnell is affiliated with the Department of Pathology, University of North Carolina at Chapel Hill, School of Medicine.
References
1. Bender IB, Seltzer S. Roentgenographic and direct observation of experimental lesions in bone: I. J Am Dent Assoc 1961 ;62:152-60. 2. Bender IB, Seltzer S. Roentgenographic and direct observation of experimental lesions in bone: I1. J Am Dent Assoc 1961 ;61:708-16. 3. Van der Stelt PF. Experimentally produced bone lesions. Oral Surg 1985;59:306-12. 4. Hardstedt CH, Welander U. Photographic subtraction I. Theory of the subtraction image. Acta Radio11975;16:559-64. 5. Webber RL, Ruttiman UE, Grondahl HG. X-ray image subtraction as a basis for assessment of periodontal changes J Periodont Res 1982;17:50911. 6. James AE, Anderson JH, Higgins CB. Digital image processing in radiology. Baltimore: Williams & Witkins, 1985. 7. Grondahl HG, Grondahl K. Subtraction radiography for the diagnosis of periodontal bone lesions. Oral Surg 1983;55:208-13. 8. Grondahl HG, Grondahl K, Webber RL. A digital subtraction technique for dental radiography. Oral Surg 1983;55:96-102. 9. Grondahl K, Grondahl HG, Webber RL. Digital subtraction radiography
178
Tyndall et al.
for diagnosis of periodontal bone lesions with simulated high-speed systems. Oral Surg 1983;55:313-8. 10. Rethman M, Ruttimann U, O'Neal R, et al. Diagnosis of bone lesions by subtraction radiography. J Periodontol 1985;56:324-9. 11. Van der Stelt PF, van der Linden LWJ, Geraets WGM, Alons CL. Digitized image processing and pattern recognition in dental radiographs with emphasis on the interdental bone. J Clin Periodonto11985;12:815-21. 12. Vos MH, Janssen PTM, van Aken J, Heethaar RM. Quantitative measurement of periodontal bone changes by digital subtraction. J Periodont Res 1986;21:583-91. 13. Kullendorff B, Grondahl K, Rohlin M, Henrikson CO. Subtraction radiography for the diagnosis of periapical bone lesions. Endod Dent Traumatol 1988;4:253-9.
Journal of Endodontics 14. Jeffcoat MK, Reddy MS, Webber RL, Williams RC, Ruttimann UE. Extraoral control of geometry for digital subtraction radiography. J Periodont Res 1987;22:396-402. 15. Ruttimann UE, Webber RL, Schmidt E. A robust digital method for film contrast correction in subtraction radiography. J Periodont Res 1986;21:48695. 16. Ruttimann UE, Webber RL. Volumetry of localized bone lesions by subtraction radiography. J Periodontal Res 1987;22:215-6. 17. Rudolph DJ, White SC, Mankovich NJ. Influence of geometric distortion and exposure parameters on sensitivity of digital subtraction radiography. Oral Surg 1987;64:631-7. 18. Swets J. ROC analysis applied to the evaluation of medical imaging techniques. Invest Radiol 1979;14:109-21.