35-mm film scanner as an intraoral dental radiograph digitizer

35mm film scanner as an intraoral dental radiograph digitizer II: Effects of brightness and contrast adjustments Michael K. Shrout, DMD,a Brad J. Potter, DDS, MS,b H. Michael Yurgalavagel” Charles F. Hildebolt, DDS, PhD,d and Michael W. Vannier, MD,e Augusta, Ga. and St. Louis, MO. MEDICAL

COLLEGE

OF RADIOLOGY,

OF GEORGIA,

WASHINGTON

SCHOOL OF DENTISTRY,

UNIVERSITY

AND MALLINCKRODT

INSTITUTE

SCHOOL OF MEDICINE

Typical 35-mm slide scanners use a photodiode array and software that allows for digital and analog controls that are manually adjustable. The digital controls provide brightness and contrast adjustments, whereas corresponding analog controls adjust the exposure time and black level that determines the clamping level of the charge-coupled device for the maximum black in the image. The objective of this study was to determine the effects of these controls on the radiometric data of intraoral dental radiograph images, to establish recommended settings, and to set specific standard guidelines for the digitization process, Three approaches were used. The results of this study demonstrate that brightness and contrast control alterations on the digitizer produces different optical densities and modulation transfer function values, The impact of these results is unresolved yet must be considered in analysis on quantitative radiometric studies. (ORAL SURC ORAL MED ORAL PATHOL 1993;76:510-8)

As progress is made in digital dental radiographic techniques for subtraction and radiometric image analysis,‘-l’ it has become increasingly important to evaluate the effects of variations in the digitization process. Fundamental to all digital analysis is the conversion of the radiographic image captured on film into digital data. One technique that has become increasingly available to researchers is the solid-state slide/film scanner that provides the potential to increase spatial resolution of images as well as to improve sharpness and resolution.‘2, l3 Several manufacturers offer 35-mm slide scanners that are compatible with a variety of personal computers, including the Apple Macintosh IT (Apple Computer Co., Cupertino, Calif.), Sun SparcStation, and DOSbasedpersonal computers (PCs). These scanners are designed to allow numerous adjustments during the digitization operation, the objective is to optimize the visual quality of the images. In quantitative imaging This study was supported in part by NIDR projects DE08173 and DE10084 and a Medical College of Georgia Research Institute grant. =Associate Professor in Oral Diagnosis and Patient Services, Assistant Professor in Oral Biology, Medical College of Georgia, School of Dentistry. bAssistant Professor in Oral Diagnosis and Patient Services, Medical College of Georgia, School of Dentistry. CResearch Assistant, Medical College of Georgia, School of Dentistry. dAssociate Professor in Radiology, Mallinckrodt Institute of Radiology. “Professor in Radiology, Mallinckrodt Institute of Radiology. Copyright @ 1993 by Mosby-Year Book, Inc. 0030-4220/93/$1.00 + .lO 7/16/48918

HO

studies, however, the objective is rarely to produce a pretty image but to extract quantitative measures from the digital image. A concern in quantitative studies is that both digital and analog functions can be varied during digitization. The digital controls provide brightness and contrast adjustments, whereas the corresponding analog controls adjust the exposure time and black level that controls the maximum black in the image by adjusting up and down from the default black (3.0 optical density units). The principle difference between the two types of adjustments is that the analog adjustments affect the charge-coupled device (CCD) voltage levels and exposure times whereas the digital adjustments affect the data after it has been collected from the CCD.14 It is unclear if or how these adjustment affect the radiometric data found within the radiographic image. The objective of this study was to evaluate the effects of these controls on a digitization system, to make recommendations on their settings, and to establish standard guidelines for the digitization process. MATERIAL AND METHODS

With the use of a range of brightness and contrast settings on the scanner,various images were digitized: (1) a routine periapical radiograph, Fig. I; (2) a serially exposed dental radiograph, Fig. 2; and (3) an operational alignment test pattern provided through the Society of Motion Picture and Television Engineers (SMPTE), Fig. 3. The resulting images were used to evaluate the effect of digital and analog controls on radiometric image values: first, to assessthe

Shrout et al.

ORAL SURGERY ORAIL MEDICINE ORAL PATHOLOGY Volume 76, Number 4

Fig. 1. Periapical radiograph used to produce periapical seriesof images.Thesewere produced by varying brightness and contrast settings in the analog and digital modes.

577

Fig. 2. Variable optical density film used to produce density series of images.

Fig. 3. SMPTE 35 mm operational alignment (TV2-OA) test pattern. White bars and black background were usedin testsfor noise. These areasand spatial frequency bursts were also usedin determinations of MTF. overall effects of the various control settings, cumulative percent histograms (CPH) from a standardized region of interest. (ROI) drawn on each periapical image were calculated and compared; next, to consider how dissimilar film densities are affected by changes in the control settings, CPHs were calculated and compared from a multiple optical density film; and finally, to determine if image resolution is affected by changes in the digital and analog controls, the modulation transfer function (MTF) was calculated for the series of images of the SMPTE test pattern. In addition, the MTF was calculated for two images of the test pattern that were digitized through

Table I. Control settings used for periapical and test pa.ttern film digitization Digital settings (Brightness, contrast) -s,5 -S,O -s,-5 %5 0s 0,-S 5,5 5s 5,-5

!

Analog settings (Exposure, blackness) 255,15 255,0 255,-15 150,15 150,o 150,-15 50,15 50,o 50,-15

512

Shrout

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

et al.

October

I993

4 ‘.*

100

150 Gray

4

200

250

Level

Fig. 4. Typical histograms for periapical image series. Plots are gray-scale values (horizontal axis) against frequency or number of pixels per bin (vertical axis).

a 386 IBM compatible PC (Computer Sales Professional, Somerset, N.J.) using the default analog and digital settings. images A periapical film was exposed with a commercially available positioning device (Kwik-Bite, Hawe-Neos Dental, Gentilino, Switzerland). Exposures were made at 70 kVp, 10 mA, and 0.4 seconds with a longcone technique using E-speed film (EP-21, Eastman

Kodak Co., Rochester, N. Y.) and a GE-1000 X-ray machine (Gendex Corp., Milwaukee, Wise.). The radiograph was developed the same day with an automatic film processor (Xonics, Allied Photo Products, Norcross, Ga.). The variable optical density film was produced by dividing a D-speed film (DF-57, Eastman Kodak Co., Rochester, N. Y.) into six equal parts. A half-inch steel plate was placed on all but the first one-sixth of the film. The GE- 1000 X-ray machine, set at 70 kVp

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY Volume 76, Number 4

Shrout et al.

Cumulative Percentage

Fig. 5. CPH of the histograms in Fig. 4.

60

80

100

120

Gray

140

160

180

200

Level

Fig. 15. Digital mode adjustments demonstrate graphically the effect of brightness control (first number in legend) anld contrast control (the second number in legend) on CPH.

513

514

ORAL SURGERY ORAL MEDKINE ORAL PATHOLOGY October I993

Shrout et al.

20

40

so

60

100

140

120

160

Gray Level

Fig. 7. Analog mode adjustments demonstrate graphically the effect of exposure control (first number in legend) and blackness control (second number in legend) on CPH.

Multiple optical density fii

q

-5,0

* -4,0 q

-3,0

* -2,0 m 40

0 0

50

100

1.50

200

250

Gray Level Fig. 8. Effects of digital adjustments on CPHs of multiple optical density film.

Shrout et al.

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY Volume 76, Number 4

515

Multiple opticaldensityfilm

a 40,o l

50,o

m 65,O *

SO,0

-

95,o

Gray Level Fig. 9. Effects of analog adjustments on CPHs of multiple optical density film. 2.0

1

lb-

l.O-

0.5 -

Fig. 10. Effects of digital adjustments on MTF of SMPTE test pattern. and 10 mA, was placed 1 meter away from the film and exposed for 0.17 seconds. The second one-sixth of the film was removed from under the block and irradiated. This was continued until the last sixth was ex-

posed for 0.17 seconds. The radiograph was developed the same day with the automatic film processor. Film optical densities were determined by measuring each section three times with a densitometer (Model NO.

516

Shrout et al.

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

October 1993

50

100

150

200

250

300

a50

400

450

500

550

600

650

Frequency

Fig. 11. Effects of analog adjustments on MTF of SMPTE test pattern.

Brightness

Contrast (Greater Effect)

Digital

Analog

Q (Greater

Effect)

Fig. 12. Effects of adjustments, recommended setting ranges for digital, and analog brightness and contrast controls with Nikon scanner for dental radiographs. 301, X-Rite Co., Grandviile, Mich.) and calculating a mean value for each section. The test pattern, purchased from the SMPTE, is an operational alignment test pattern (TV2-OA, RP27.1). Each radiograph was mounted in a Bantam 35-mm double glass slide mount (H.P. Marketing Corp., Cedar Grove, N. J.). These mounts hold the radiographs in fixed positions and prevent film bending, which

could cause depth of field distortions. The slide scanner consisted of a Nikon model LS-3510AF CCD linear photodiode array solid state scanner interfaced through Photoshop software (Adobe Systems Incorporated, Mountain View, Calif.) to a Macintosh IIcx computer, with an 8-bit color display. In accordance with the manufacturer and software directions, the radiographs and the test pattern were digitized, converting the optical densities into 256 gray levels. In

ORAL SURGERYORAL MEDICINE ORAL PATHOLOGY

Shrout et al.

517

Volume 76, Number 4 addition, two images of the test pattern were digitized through a 386 PC with the use of PhotoStyler software (Aldus Corp., Seattle, Wash.) using the default analog and digital settings. The periapical film and the test pattern were digitized. Each film was used to produce two series of nine images. One varied digital settings and the other varied analog settings. The digital settings ranged from -5 to +5 for brightness and -5 to +5 for contrast The analog settings ranged from 0 to 255 for exposure and - 15 ‘to + 15 for blackness. Adjustments were made to include the full range permitted, see Table I. The range of the multiple density film image series was further limited to the parameters we recommend specifically for dental radiographs. The digital settings were varied from 0 to 5 for brightness and - 1 to + 1 for contrast, the analog from 40 to 95 for the exposure time (brightness) and - 1 to f 1 for the contrast (blackness). Analysis The series of images produced by the periapical and multiple optical density films were evaluated by calculating frequency distributions (Fig. 4) and comparing them. The histograms were produced from a ROI that was kept constant for each image in the series. To help compare hist0gra.m shapes, the original histograms were transformed by a two-step process. First, a percentage was calculated by dividing the frequencies for each of the 256 bins by the sum of the frequencies for all bins and multiplying by 100. Next, a running sum was computed by replacing each percentage with a summation of the percentages up to and including the bin being replaced. A plot of these values is called a cumulative percent histogram (Fig. 5). The CPH for the two series of images was plotted. An overall mea.sure of image contrast, sharpness, and resolution is provided by the MTF. The operational alignment test pattern was used to determine the MTF for each digitization setting. ROIs of the black background, the white bar pattern, and the spat.ial frequency bursts (bars) in the SMPTE test pattern (Fig. 3) were drawn and evaluated. An ideal digitizer would convert the transitions between the white and black areas of the bars to abrupt changes in gray levels and, would assign respectively uniform gray-level values to the white and black areas of the bar patterns. A perfect digitizer with no loss of image quality would have a MTF of 1. To determine the MTFs at different spatial frequencies, the methods of Droege and Morin15, l6 were used as described in part I of this series of articles.12 The resulting MTF values were plotted against fre-

quencies that for this study are functions of bar widths that vary from 0.17% to 1.OO%of test pattern height. RESULTS The CPH shown in Fig. 6 graphically demonstrate the effect of the digital mode brightness and contrast controls of the Nikon system during the digitization process. The brightness control (the first number in the legend) shifts the histograms along the X-axis, whereas the contrast control (the second number in the legend) tends to change the slope of histogram at a given brightness setting. The analog mode adjustments data shown in Fig. 7 demonstrate that the exposure settings affect the CPHs to a much larger degree then do the contrast settings, which affected the data the least of the four combinations. The exposure time control appears both to shift the histogram along the X-axis and to change the slope. The multiple optical density film has mean optical densities of 0.45, 1.0, 1.51, 1.99, 2.36, and 2.7. The filrn was used to generate a series of images designed to evaluate how variations in the digitization process affect the gray-scale values of the digital images. The CPHs for these images are shown in Figs. 8 and 9. The digital-mode histograms (Fig. 8) vary along the X-axis. The analog-mode histograms (Fig. 9) also demonstrate variation. ‘The MTF data shown in Figs. 10 and 11 demonstrate that the resolution is affected by the digital and analog mode settings. Although there appears to be less difference in the MTF between the various analog settings, MTFs for the digital mode appear to be slightly better as a group than the analog. The images processed through the PC have different resolution values than those processed through the Macintosh. DISCUSSION Figs. 6 and 7 indicate that it is possible to shift the mean gray level of an image up or down the scale while digitizing dental radiographs with the Nikon scanner. Cl.ose examination also reveals that the histograms are not merely translated along the x-axis but also undergo shape changes. Simultaneous adjustment of both the brightness and contrast control will surely change the histogram as the look of the image changes. Figs. 8 and 9 demonstrate that an adjustment that results in a 50 gray-scale shift in the 25 to 100 graylevel range produces a much smaller shift farther up the scale, particularly above 200. This may be caused by the nonlinear nature of the H-D curve of the radiographic film in addition to the digitizing system.

518 Shrout et al.

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

October I993

The implication is that if a radiographic image is judged overexposed (dark) and a correction is attempted during digitization, the radiometric data may be different from an otherwise identical image with a satisfactory optical density. Figs. 10 and 11 demonstrate that adjustments of the brightness and contrast controls affect the resolution of images. The importance of this effect remains to be studied. The least expected result of the resolution study is the difference in resolution between the images digitized by the Macintosh and the PC. Although the differences are not extensive and their importance is hard to judge, these results are noteworthy. On the basis of the results of this study and experience using the Nikon scanner, we recommend the ranges shown in Fig. 12 when using the system for digitizing dental radiographs. CONCLUSlON

The Nikon slide scanner doesprovide the ability to adjust image brightness and contrast during the digitization process.These adjustments, however, affect the nature of the radiometric data in the radiograph. It is suggestedthat extreme settings should be avoided and that the settings should not be changed during a study. The nature of the digital images produced from different systemsneedsmore study, but it is safeto say that video versus slide scanner based,Macintosh versus PC, direct versus indirect digitization will all result in different images. In quantitative digital studies, the mode and settings of the digitization process should be fully reported to insure reproducible results. REFERENCES 1. Grondahl K, Kullendorff B, Strid KG, et al. Delectability of artificial marginal bone lesions as a function of lesion depth: a comparison between subtraction radiography and conventional radiographic technique. J Clin Periodontol 1988;lS: 156-62.

2 Nummikoski PV, Martinez TS, Matteson SR, McDavid WD, Dove SE. Digital subtraction radiography in artificial recurrent caries detection. Dentomaxillofac Radiol; 1992;21:59-64. 3. Hausemann E, Christenson L, Dunford R; Wikesjo U, Phyo J, Genco RJ. Usefulness of subtraction radiography in the evaluation of periodontal therapy. J Clin Periodontol 1985;56:4-7. 4. Engelke W, Ruttimann UE, Tsuchimochi M, Bather JD. An experimental study of new diagnostic methods for the examination of osseous lesions in the temporomandibular joint. ORAL SC’RG ORAL MED ORAL PATHOL

1992;73:348-59.

5. Fujita M, Kodera Y, Ogawa M, Wada T, Doi K. Digital image processing of periapical radiographs. ORAL SURG ORAL MED ORAL PATHOL

1988;765:490-4.

6. Mol A, Dunn SM, van der Stilt PF. Diagnosing periapical bone lesion on radiographs by means of texture analysis. ORAL SURG ORAL MED ORAL PATHOL

1992;73:746-50.

7. Southard KA, Southard TE. Quantitative features of digitized radiographic bone profiles. ORAL SURG ORAL MED ORAL PATHOL

1992;73:751-9.

8. Bragger U, Pasquali LA, Weber H, Kornman KD. Computerassisted densitometric image analvsis _ (CADIA) for the assessment of alveolar bone dinsity changes in furcations. J Clin Periodontol 1989;16:46-52. 9. Bragger U, Pasquali LA, Rylander H, Carnes D, Kornman KS. Computer-assisted densitometric image analysis in periodontal radiographv J Clin Periodontol 1988;15:27-37. 10. Mouyen F, Benz C, Sonnabend E, Lodter JP. Presentation and uhvsical evaluation of RadioVisiGraohv. ORAL SURG ORAL ‘M& ORAL PATHOL 1989;68:238-42.’ ’ 11. Razzano MR, Bonner PJ. RadioVisioGraphy: video imaging alters traditional approach to radiography. Compend Cont Ed Dent 199O;XI:398-400. 12. Shrout MK, BJ Potter BJ, Yurgalavage H, Hildebolt, CF Vannier MW. A 35-mm film scanner as an intraoral dental radiograph digitizer. I: a quantitative evaluation. ORAL SURG ORAL MED ORAL PATHOL

1993;76:502-9.

13. Hildebolt CF, Vannier MW, Pilgram TK. Shrout MK. Quantitative evaluation of digital dental radiographic imaging systems. ORAL SURG ORAL MED ORAL PATH~L 1990;70:661-8. 14. Nikon Driver Software, Nikon Electronic Imaging. 1991;51 l-5-12. 15 Droege RT, Morin RL. A practical method to measure the MTF of CT scanners. Med Phys 1982;9:758-60. 16. Droege RT. A practical method to routinely monitor resolution in digital images. Med Phys 1983;10:337-43. Reprint requests. Michael K. Shrout, DMD Associate Professor in Oral Diagnosis and Patient Services Medical College of Georgia School of Dentistry Augusta, GA 30912-1241