computer-aided analysis of intraoral ulcerative lesions

Computer-aided analysis of intraoral ulcerative lesions Jane Wilson, DDS,” Richard Firnhaber, DDS,’ and Mark J. Kutcher, DDS, MS,b Chapel Hill, N.C.

UNIVERSITY OF NORTH CAROLINA SCHOOL OF DENTISTRY

At the present time there is no objective method to evaluate oral ulcerative lesions for degree of inflammation, surface area, or other parameters of healing. The capability of infrared photography to uniquely image the inflammatory process associated with oral lesions and the use of image-processing technology for parameter analysis was investigated. Lesions were compared cross-sectionally and over time for area of tissue degeneration, active inflammation, and intensity of inflammation with respect to adjacent tissue. Eight subjects were available for study. Standard black and white, color, and black and white infrared photographs were taken in accord with study protocols. A reflectance standard was included in each photograph to normalize the background intensity. A video camera was used to enter the data into a microcomputer image processing system. Quantitative data obtained included area of ulceration and erythematous halo, inflammatory intensity changes over time, and intensity with respect to adjacent tissue. This computer-aided photographic technique was able to quantify healing progression and intensity parameters associated with intraoral inflammatory lesions. (ORAL SURC ORAL MED ORAL PATHOL 1992;74:393-9)

T he purpose of this study was threefold: first, to develop a prototype for an objective evaluation of inflammatory lesions with the use of available computer technology; second, to compare film types for their usefulness in imaging inflammatory lesions; and third, to evaluate the progression and regression of an inflammatory lesion by means of the prototype developed.

Computer digitalization techniques have been applied to dental radiographs.*-4 We applied this technology by using 35 mm film negatives for data input rather than radiographic film. We selected film types that we believed to be best suited for evaluation of the inflammatory parameters under study. In this study, black and white infrared (IR) and standard black and white (BW) film negatives were used as data input images. Color photographs were used for comparison documentation purposes.

Supported by NIH Grant No. S07RR05333. aPrivate practice. bAssociate Professor and Chairman, Department of Diagnostic Sciences. 7/17/37315

IR photography has been used since the 1930s as an aid in the identification and imaging of pathologic conditions on or near the skin surface.5 The advantage of IR media over visible light media is the ability of IR radiation to penetrate the skin surface more deeply before reflection.6 As a result, subsurface phenomena may be imaged. Venous patterns behind normal skin and scabs were demonstrated by Massopust7 and Haxthausen.5 Subsequently, attempts were made to use IR for breast cancer screening. These develop- ments led to the evolution of thermography.

In thermography, the direct radiant heat of the skin is recorded with the use of an IR-sensitive camera that scans the area of interest.8 Early applications of this technology to dentistry were not promising. Crandell and Hill9 could not distinguish vital from nonvital teeth by temperature. Soffin et al.tO were able to identify only the involved from the uninvolved side in eleven patients with various oral inflammatory conditions. White et al., I1 however, who used a more sophisticated thermal video system and microproces- sor, were able to record the temperature of lesions to 0.2” C and demonstrated that intraoral ulcers varied in temperature from their surroundings and across

393

394 Wilson. Firnhaber. and Kulcher mat SLIK(l OKt\l Mt 0 OK\L. P-zlHOl September I992

Fig. 1. Imaging setup with light box, camera lens, and monitor with display of computer-processed ulcerative lesion.

their surface. They suggested that the technology was sensitive enough to be used in anti-inflammatory agent studies.

There are problems with the use of thermography as a convenient and accurate diagnostic method. Be- cause thermography has generally been replaced by more sophisticated imaging systems such as magnetic resonance, computer tomography, etc., availability and the impetus for further innovation has decreased. The cost has also kept it confined to larger medical centers. In addition, the actual clinical environment must be carefully controlled for such variables as room temperature. Perhaps the greatest disadvantage of thermography is the limitation to superficial phenomena since thermography cannot image deeper tissue structures.”

IR film photography, on the other hand, can penetrate the skin to a depth of up to 3 mm.12 Subsurface red blood cells demonstrate a unique color when photographed with IR color film.t3 Marshall extensively investigated how IR data could be obtained and stan- dardized. It was found that IR photography could image neuroanatomic details in spinal defects be- neath membranes that previously required dissec- tion.14 Marshall’s objective method could differenti- ate malignant melanoma from benign lesions more accurately than experienced clinicians but not quite as accurately as experts in the field.15, I6 He suggested that computer digitalization would increase the diagnostic utility of IR photographs.‘7 Because this computer technology is now available, we chose to evalu-

ate the usefulness of IR photography as a possible data input source for our analysis of oral lesions.

We also wished to develop a method to objectively quantify the size of an inflammatory lesion. For this we used standard BW film. We chose not to use color film for data input at this time because of the more complex computer programming that would be required. However, color films were taken throughout the study for documentation purposes and to record the clinical appearance of the lesions studied.

METHODS

Potential subjects were recruited when they pre- sented with an intraoral ulcer. The subjects were seen within the first day or two of the onset of their lesion. All eight subjects were a part of the University of North Carolina at Chapel Hill Dental School popu- lation, and all were white. The subjects in the study were healthy and ranged in age from 21 to 26 years. All lesions imaged were diagnosed as aphthous ulcers. The protocol was slightly refined and modified after a review of the data from the first patients. Camera settings, lighting conditions, and film processing variables were adjusted to provide the most ideal IR image of the lesions.

A 35 mm camera (Nikon 2000, Nikon, Inc., Melville, N.Y.) with ring flash and macro lens set at a magnification ratio of 1: 1 was used for all of the photography. Film types used included Kodak Pana- tomic-X 32 BW film (Eastman Kodak Co., Roches- ter, N.Y.) Ektachrome 100 color film, and high-speed

Volume 74 Number 3

Computer analysis of oral ulcers 395

HORIZONTAL IMAGE CALIBRATION VERTICAL ADJUSTMENT - - ADJUSTMENT

NEGATIVE IMAGE CONVERTED TO POSITIVE

I I BRIGHTNESS

STANDARDIZATION HISTOGRAM

IMAGE ANALYSIS IMAGE TRACING

AREA/BRIGHTNESS RATIO DETERMINED

SEED PIXEL/COLOR - SELECTED AREAS COLORED SELECTED

2. Flowchart details imaging protocol from photographic negative input to calibrated, colored, lesional image.

BW and IR film. As a reflectance and size standard, we used a 6 mm circle cut from a Kodak chromato- gram sheet (silica gel without fluorescent indicator). The dot was placed near the lesion in each view. This material was chosen because it uniformly reflects the wavelengths of interest and can be related to a universal magnesium oxide reference.i2 The dot also provided a standard to serve as a reflectance/brightness and size control in the computerization process for each film type and between films on the same patient. For the IR photographs, a red opaque Kodak 88A filter was placed over the camera lens. Since the focal length in IR photography differs from that of visible light-sensitive film, and since the best aperture and speed for intraoral IR photography varies, a series of shots was taken to frame these factors. The best IR settings were determined by three observers who independently viewed the resultant negatives. All observers found that the speed and exposure settings of 400 at f-32 or 800 at f-22 provided the best image vi- sualization. The settings for the standard BW and color films followed the manufacturer’s recommen- dations. BW films and IR films were developed by the investigators who used HCl 10 dilution B developer (Eastman Kodak Co., Rochester, N.Y.) at 68” for 6 minutes with manual agitation. The color slides were

developed by a professional photography laboratory. After the film protocol had been determined as

aforementioned, the following clinical protocol was followed. The patient was seated in a standard dental chair, the lesion was identified, and the area was dried with a cotton swab. The standardization dot was placed on the mucosa in the vicinity of the lesion. A color photograph was taken of the lesion for documentation purposes. The color film was replaced by IR film. A series of IR films were taken by an investigator who set the camera and focused on the lesion. As soon as the camera settings were established, a second investigator placed the IR filter over the lens and the film was exposed. The IR photographs were taken at the aforementioned speeds and exposures. IR pictures at similar settings were then taken of the contralateral, clinically normal side of the oral cavity to demonstrate IR imaging of the lesion. The IR film was replaced by BW film in total darkness. Color and BW films followed the same protocol. The film was developed as aforementioned, at which point the negatives from the BW and IR films were used for opti- cal input data to the digitizer.

To digitize the images, the negatives were placed on a standard light box that had a camera lens (Nikon Micro NIKKOR, Melville, N.Y .) placed over it (Fig.

396 Wilson, Firnhaber, and Kutcher ORAL SUM, ORAI Mto ORAL PATH~L September 1992

Table I. Fully analyzed data for two patients

Patient 6 ..-~~-

Data measured for BWjilm

Contrast ulcer border 1.20 11.61 0.36 11.1 I Contrast periphery 3.78 7.81 1.15 9.29

Changes (%) Changes (%)

Ulcer area/t* Total area/t Ulcer brightness/t Halo brightness/t Affected brightness/t Unaffected brightness/t

*/t = Per time.

-39.9 -60.8 +31.0 n/a +30.0 +20.0

Table II. Comparison of films

Comparisons between IR and BW film show a difference in the area imaged and interarea contrast. IR film also shows a decreased reflectance at the border of ulceration. Area of ulceration imaged by IR: 78% of BW image area Area of affected mucosa imaged by IR: 31% of BW image area Comparison of IR film with BW film (pixels per unit area)

IR film BW film Ulcer border contrast 0.775 11.16 Periphery contrast 2.47 8.55

1). The computer hardware consisted of an Imaging Technology IPI 12 image processing system (Imaging Technology Inc., Woburn, Mass.) with a DEC PDP 1173 computer (Digital Equipment Co., Woburn, Mass.). The software consisted of a FORTRAN program to be described. The lens system fed the photographic input to the computer. At this point, the lens could be focused by means of the computer monitor. AMDEK color monitor, Digital Equipment Co. Woburn, Mass.) The first step was to set the magnification calibration factor. This was accomplished with the use of the known standard dot diameter of 6 mm located in the area of interest and shot at a magnification ratio of 1: 1. The program then acquired a digital image of the film. The image that appeared on the monitor screen at this point was positive rather than negative. A gray-level histogram was obtained and, if necessary, the TV camera lens f-stop was adjusted so that the histogram was centered on the gray scale. Brightness was available at 256 levels and the intensity of the photographs was adjusted so that the average intensity was 128. The operator then outlined the reference dot that appeared on the screen with the use of a mouse device that, in turn, tracked the cursor movement on the screen. The lesion was outlined in the same manner. When the area was outlined, a pixel element located within the area of interest-a seed

-51.2 +186.8 +44.1 -23.5 60.1 -30.1 -2.0 +2.0 -3.0 -6.0 n/a 0 -3.0 f2.0 +8.0 +0.02 -0.9 -6.0

pixel-was selected and the program was instructed to fill the area with a color chosen by the operator. For this study, yellow was chosen to represent the center of a lesion where necrotic tissue was present, blue was used to represent the most erythematous areas, and red was used to represent the outer border of the lesion. The colors were used for comparison purposes and were not necessary for quantitative analysis. The area of the outlined lesion as well as the average intensity or brightness level was then calculated by the program. These numeric data were used to determine such information as the ratio of the average intensity of the lesion to the average density of the reference dot or changes over time (see Fig. 2). To reduce the variability in this outlining procedure between investigators, each photograph was analyzed by the same viewer four times. To evaluate the variability of positioning between films, all of the negatives taken of one lesion, at one point in time, were traced and the variability of positioning was numerically determined.

RESULTS

Examples of the BW and IR photographic results of a digitized case are shown in Fig. 3. Numeric data were obtained for the reference marker area, the necrotic area of the ulcer, the area of the erythematous halo, and the area between the edge of the halo and the periphery of the affected area. Brightness values that represent pixel intensity per unit area were recorded for the reference marker, the necrotic area, the halo, the affected area, and an unaffected control area. Because 6 mm reference dots of uniform reflectance appeared in each photograph, the ratio of sample brightness to reference dot brightness could be determined by the computer program, which allowed for comparison of photographs taken at different times or of different subjects. The degree of contrast

Volume 74 Number 3


Fig. 3. Intraoral ulcer analyzed in accord with study protocol. A, BW film shows initial ulcerative lesion on right and reflectance and size standard circle on left. B, BW film shows final ulcerative lesion. C, Com- puter image of initial lesion from BW film data. D, Computer image of final lesion from BW film data. (Key: y = yellow; b = blue; r = red.)

between different areas could then be numerically represented as the difference between the ratio of sample brightness to dot brightness. Contrast was represented by a numeric range from 0 to 256, with 0 as the minimum contrast and 256 as the maximum contrast. Table I shows the numeric findings for the two patients fully evaluated.

DISCUSSION

We found IR film to be more perishable than visible light-sensitive film and to be more affected by changes in light and heat. The IR film provided a consistently smaller and less detailed image than the BW film (Table II). The smaller size and decreased detail of the IR imaged lesion in the quantified examples may be due to the previously documented ability of IR film to image slightly subsurface phenomena. The bowl-shaped nature (larger in area at the surface than the subsurface) of aphthous ulcers may be demonstrated by the smaller IR image when compared with the corresponding BW image. In the IR photographs, a pale white outline adjacent to the necrotic

Table Ill. Average percentage of variability of tracing with cursor

IR film BWjlm (%I (%I

Reference dot 0.68 0.68 Ulcer area 4.50 1.20 Affected area 26.00 18.00

Average percentage of variability because of patient positioning was <l’% greater than for tracing alone. Average percentage of variability of brightness values was 0.92% for IR and BW.

area that could represent the area of healing was ob- served. This outline was not seen in the BW photos. The IR film had more diffuse borders than did the BW film. This was probably related both to the depth of IR penetration and the grainier nature of the IR film.

With the use of this method, we confirmed the need for lesions to be photographed with similar orienta- tion and tension on mobile tissue. The lesion should be in the same plane as the reference dot and parallel to the photographic plate.

398 Wilson, Firnhaber, and Kutchrr OK:\L SL’RG OR4L Mkr> OKI. PAlHO September 1997

Fig. 3, cont’d. Intraoral ulcer analyzed in accord with study protocol. E, Initial lesion taken with IR film. F, Final lesion taken with IR film. G, Computer image of initial lesion from IR film data. H, Computer image of final lesion from IR film data. (Key: y = yellow; b = blue; r = red.)

Numeric data were obtained from the areas outlined by the cursor for each case (Table I). Brightness differences of <l pixel per unit area over a 256-level gray scale could be seen on the film, but not on the image-processing system. The average percentage of tracing error with this software was quite small for lesions with distinct borders but slightly greater when tracing indistinct borders (Table III). Further software refinements may decrease the error. It would be meaningful to develop an edge-finding component to the program. Patient-positioning variability appeared to add less than 1% to the tracing variability. It would be possible to consider a direct video camera input of the lesion at some future time.

The relative brightness was expressed as the ratio of the brightness of the sample area to that of the reference dot. It was expected that the brightness of the unaffected tissue would differ little over time. In one instance, however, it differed by 20%. Brightness measurements may require more careful monitoring of exposure and gray scaling, even when compared with a reference standard. The average percentage of variability of brightness measurements, however, was still less than 1% for both BW and IR film.

CONCLUSIONS

With the use of the prototype developed (as aforementioned and referenced in Tables I, II, and III),

area and brightness measurements for aphthous ulcers may be compared cross-sectionally with a deter- mination of distinct zones within lesions and over time, if photographic and imaging parameters are held constant. Brightness differences across a lesion can be compared numerically. This method could be used to quantitatively study the progression of lesions other than aphthous ulcers. IR film images aphthous ulcers and adjacent normal tissues differently from standard film because it shows a slightly subsurface view. BW film allowed analysis of the superficial area of the lesion and provided more distinct borders than IR film. With further development and standardization, this prototype could be used in research efforts that quantify healing responses to trial medications as well as to study the progression of oral inflammatory lesions objectively and quantitatively.

We thank Jamie Firnhaber for her assistance in proof- reading this article and Eastman Kodak for supplying chromatograph paper for use in this project.

REFERENCES

I. Grondahl H, Grondahl L, Webber R. A digital subtraction technique for dental radiograuhv, ORAL SURG ORAL MED ORAL@ATHOL 1983;55:96-102: .

2. Chow CK. Hilal SK. Hiebuhr KE. X-rav image subtraction bv digital mdans, IBM: J Res Dev 1973;1?:206118.

3. McHenry K, Hausmann E, Wikesjo U, Dunford R, Lyon- Bottenfield E, Christersson L. Methodological aspects and

Volume 74 Number 3


4.

5.

6.

7.

8.

9.

10.

11.

12.

quantitative adjuncts to computerized subtraction radiography. J Periodontal Res 1987;22:125-32. Rethman M, Ruttiman U, G’Neal 0. Diagnosis of bone lesions by subtraction radiography. J Periodontol 1985;56:324-9. Haxthausen H. IR photography of the subcutaneous veins: demonstration of concealed varices and ulcer in eczema of the leg. Br J Dermatol 1933;45:506. Medical IR Photography, Publication N-l, Rochester, New York: Eastman Kodak Medical Infrared Photography, CAT 151 9982, 1973;1-130. Massopust LL. IR photographic study of the changing venous patterns of the superficial veins in a case of human pregnancy. Surg Gynecol Obstet 1963;86. Lawson R. Thermography: new tool in the investigation of breast lesions. Canad Serv Med J 1957;8:517-24. Crandell C, Hill R. Thermography in dentistry: a pilot study. ORAL SURG ORAL MED ORAL PATHOL 1966;21:316-20. Soffin C, Morse D, Seltzer S, Lapayowker M. Thermography and oral inflammatory conditions. ORAL SURG ORAL MED ORAL PATHOL 1983;56:256-62. White B, Lockhart P, Connelly S, Sonis S. The use of IR thermography in the evaluation of oral lesions. J Am Dent Assoc 1986;113:783-6. Marshall R. IR and UV photography in a study of the selec-

13.

14.

15.

16.

17.

tive absorption of radiation by pigmented lesions of the skin. Med Biol Illus 1976;May 26(2):71-84. Gibson H. Further data on the use of IR color film. J Biol Pho- togr Assn 1965;33:155. Marshall RJ. IR recording of neural tube defects. J Audiov Media Med 1979;2:92-4. Marshal R. Evaluation of a diagnostic test based on photographic photometry of IR and UV radiation reflected by pigmented skin lesions. J Audiov Media Med 198O;July 3(3):94-S. Marshall R. IR and UV reflectance measurements as an aid to diagnosis of pigmented skin lesions. J Audiov Media Med 198l;Jan 4(1):11-4. Marshall R. A TV method for measuring IR and UV reflectance of pigmented lesions. J Audiov Media Med 1982;Apr 5(2):51-5.

Reprint requests: Mark J. Kutcher, DDS, MS Associate Professor and Chairman Department of Diagnostic Sciences University of North Carolina at Chapel Hill, School of Dentistry CB #7450, Brauer Hall Chapel Hill, NC 27599-7450

computer-aided analysis of intraoral ulcerative lesions

Documents