oral and maxillofacial radiology Editor: ALLAN G. FARMAN, SDS, PhD (Odont), MBA Division of Radiology and Imaging Sciences Department of Biology and Physical Sciences School of Dentistry, IJniversity of Louisville Louisville, Kentucky 40292

Preliminary evaluation of a digital system for rotational panoramic radiography S. Brent Dove, DOS, MS,a William D. McDavid, PhD,” Uu Welander, DDS, PhD,C and Gunilla Tronje, DDS, PhD,d San Antonio, Tex., and Stockholm, Sweden DEPARTMENT CENTER

OF DENTAL

AT SAN

ANTONIO,

DIAGNOSTIC AND

THE

SCIENCE, DEPARTMENT

UNIVERSITY OF ORAL

OF TEXAS RADIOLOGY,

HEALTH

SCIENCE

KAROLINSKA

INSTITUTET

A prototype system for direct digital panoramic radiography has been evaluated with respect to density, contrast, magnification, distortion, resolution, and overall image quality. Density and contrast depend on detector calibration and may be modified by the display system or by digital processing of the captured image. Variation of magnification in the horizontal and vertical dimensions gives rise to distortion phenomena that are identical to those encountered in film-based systems. Resolution in the vertical dimensnon is determined by the pitch of the detector elements. In the horizontal dimension, resolution is limited by the effective width of the detector elements. To evaluate the clinical acceptability of the images, radiologists and general practice residents were asked to assess the perceptibility of important radiographic landmarks in film-based and digital images of both a radiographic phantom anld a patient. The digital system performed on a par with film in the representation of normal morphologic structures of the clinical human subject whereas more differences were appa.rent in the phantom images. The general practice residents consistently rated the digital images higher than their radiologist counterparts did. No consistent trends were found to indicate any inherent deficiencies of the digital system in the depiction of any one area. The results indicate the promise of direct digital acquisition as a method of panoramic imaging. (ORAL SURG ORAIL MED ORAL PATHOL

1992;73:623-32)

I

n a previous publication,’ the components and operation of a system for digital pano’ramic radiography have been briefly outlined. In this system, the filmThis project was supported by grants from the University of Texas Higher Education Coordinating Board Advanced Technology Program (2224 and 003659-015) to the University of Texas Health Science Center and by a grant from the Naltional Institute of Dental Research (IR43DE08930) to Biomedical Development Corporation. aAssistant Professor, Department of Dental Diagnostic Science, University of Texas Health Science Center. bProfessor, Department of Dental Diagnostic Science, University of Texas Health Science Center. “Professor and Chairman, Department of Oral Radiology, Karolinska Institutet. dDocent, Department of Oral Radiology, Karolinska Institutet. 7116134799

drive mechanism of an Orthopantomograph model OP 10 (Instrumentarium Imaging, Helsinki, Finland) has been replaced by an electronic sensor consisting of a one-dimensional photodiode array coupled to a rare earth scintillator screen. The signal from the photodiode array is digitized, corrected for offset and gain, and delivered to a computer for storage in digital format. The images are stored on an optical disk and displayed on a high-resolution video monitor. The system uses a custom-designed solid-state linear x-ray image sensor (Thomson-CSF, Boulogne, France). This sensor is composed of a linear silicon photodiode array covered by a scintillator that detects x-rays that pass through the object under examination. The signal from each photodiode is read out 623

24

Dove et al.

successively by a group of multiplexers. The multiplexed signal is subsequently processed by an electronic preprocessor that corrects the offset and gain on each channel by digital processing. The corrected signal can then be exploited either for direct display through a frame memory or for image processing. The sensor has 512 photodiodes with a 0.225 mm pitch. The operating range of the sensor is ideal for this application, accepting a photon energy range of 5 to 200 keV. Absorption is 60% to 90% in the energy range of interest. The x-ray dose accumulated by the object to obtain an image (signal/noise ratio of 100: 1) is between 1 and 2 microgray. Under these conditions, it is possible to detect and display 1% variations in the incident x-ray flux. The sensor allows an external signal to trigger the scanning array, which accommodates integration times that vary from 5 to 100 msec.2 A proprietary clock circuit provides all integration times. The integration clock circuit is designed to interface with the preprocessor unit. The detector system output is eight-bit digital image data that is interfaced to the Univision UPX 1000 high-resolution programmable image digitizer. The UPX 1000 digitizer supports inputs from digital scanners at transfer rates of up to 10 million eight-bit pixels per second at a 10 MHz clock rate.3 The UPX 1000 is integrated with the high-resolution display controller (Univision UDC 2600, Univision Technologies, Inc., Burlington, Mass.). These two IBM AT bus compatible boards allow for an interactive frame memory and display system. The UDC 2600 supports 60 Hz noninterlaced monitors. This board provides for eight-bit (256 gray levels) monochrome display with two bits of graphic overlay. Final output display is 1024 X 1024 X 8 bits3 A Philips high-resolution 60 Hz noninterlaced monitor (CRT, Univision Technologies, Inc., Burlington, Mass.) is used for the main output device. Hard copy may be provided by a digital high-resolution paper printer or a laser printer that uses transparent film. To assess the basic imaging characteristics of the system, a series of tests have been performed. The purpose of this article is to describe the results of these tests, which include a preliminary evaluation of the physical characteristics of the system as well as an assessment of the images from a clinical perspective. MATERIAL AND METHODS Physical evaluation

To describe the basic imaging characteristics of the system, density and contrast were studied as well as magnification and resolution at different object depths.

Density and contrast. Density and contrast depend on detector calibration and may be modified by digital processing of the captured image as well as by setting of the video display. The offset and gain calibrations determine the basic gray scale characteristics. The offset calibration is performed in the absence of x-rays and assigns a value of 255 to the white portions of the image. The gain calibration is performed in a uniform x-ray field and assigns a value of 0 to the x-ray level used in the calibration process. This level will appear as black in the resultant images. In practice, a filter is used to reduce the x-ray intensity to a level comparable to that of the beam transmitted through the most radiolucent portions of a patient undergoing panoramic radiography. With an adjustment of the level of x-rays used in the gain calibration, the contrast of the digital image can be altered. To investigate the effect of detector calibration on contrast, an aluminum step-wedge consisting of 11 steps from 5 to 35 mm in thickness, graduated in 3 mm increments, was radiographed with varying thicknesses of aluminum for the gain calibration. Gray scale values for the different steps in various regions of the image were averaged and plotted against step thickness to provide a graphic representation of the contrast scale. Magnification and distortion. The characteristic distortion phenomena associated with panoramic radiography have been discussed elsewhere.4 Objects positioned in the central plane of the image layer are depicted with identical vertical and horizontal magnification factors. This results in a properly proportioned image. For objects positioned outside the central plane, however, the magnification is different in the two dimensions of the image. This results in distortion effects unique to panoramic radiography. Objects positioned between the central plane of the image layer and the image receptor are depicted with a horizontal magnification smaller than the vertical magnification, whereas objects positioned between the rotation center and the central plane are portrayed with a horizontal magnification larger than the vertical magnification. The prototype digital system has been programmed to simulate the geometric properties of an ordinary panoramic radiograph recorded on a moving film exposed in the Orthopantomograph model OPlO. Objects in the central plane of the image layer should be depicted with correct proportions and should have a magnification factor that varies across the image in the same way as the unmodified OP 10. Images of objects positioned at various object depths outside the central plane should exhibit the familiar pattern of distortion seen in panoramic radiographs. To verify

Evaluation of digital panoramic radiography

Volume 73 Number 5

Fig. 1. Digital image of radiographic high-resolution monochrome monitor.

phantom display on

625

Fig. 2. Digital image of radiographic printed film.

by a LaserTechnics

phantom hard copy laser printer on transparent

Table 1. Categories for the image quality

evaluation survey 1. 2. 3. 4. 5. 6.

General morphologic structures Periodontal membrane space in the apical region Lamina dura Trabecular bone pattern Cementoenamel junction Interdental septal bone level

this result, a chain of metallic spheres 2 mm in diameter was arranged along a curve that represents the central plane of the image layer of the OP 10.5 The test device was mounted horizontally in the machine, with the most anterior spheres aligned with the light beam provided for patient positioning. The resultant radiographs were examined to verify that the images of the spheres were round and that their magnification factors fell within the range predicted by theory. To study the effects of positioning on distortion, a single metallic sphere was positioned at different ‘object depths and radiographed with the digital system. Resolution. Resolution in the vertical dimension is determined by the pitch or height of the detector elements. In the horizontal dimension, resolution is limited by the width of the detector elements. To measure resolution and compare it with that of an ordinary film-based system, commercially available test patterns (Nuclear Associates, Carle Place, New York) made of lead-foil screens sandwiched between plastic plates were exposed in the same machine with the electronic acquisition system and also with the screenfilm combination (Lanex Regular/T-mat G, Eastman Kodak Co., Rochester, N.Y.) ordinarily used in

Fig. 3. Conventional

film image of radiographic

phan-

tom.

the x-ray clinic. For the vertical measurements, a bar pattern with groups of lines that ranged from 0.5 to 10 line pairs/per millimeter was used. For the horizontal measurements, a star-segment pattern was used to provide line pairs up to 20/mm. All measurements were performed with the test patterns placed at the midline in the most anterior portion of the central plane of the OPlO image layer. Image Quality Evaluation

To evaluate image quality, a questionnaire was developed on the basis of the subjective perception of morphologic landmarks by a panel of observers6, lo The use of anatomic criteria to assessthe quality of radio-

26

Dove et ai.

1 &

d B &

1

2

3

4

5

6

7

*

13mmAl

*

7mmAl

*

OmmAl

t3

Step Number

igl. 4. Response curvesfor the TMJ region.

Fig.

5. Digital radiograph

of a chain positioned

graphs has been reviewed extensiveiy by Vucich7 While receiver operating characteristic curve analysis would provide a more direct assessmentof diagnostic accuracy, it is limited by the demands of the experiment design. Adequate numbers of casesthat represent diseaseand nondiseasestates with verifica-

along the central plane of the OPlO pantomograph.

tion of ‘“truth” must be acquired to establish statistical validity. Becauseof these limitations, the present analysis was performed to obtain some preliminary information as to the quality of the images produced with this prototype device. The questionnaire was composed of 35 questions

Evaluation

Volume 73 Number 5

of digital panoramic

radiography

627

Fig. 6. A round object positioned in the central plane of the OP 10 image layer is depicted in its correct proportions (central radiograph). When the object is placed too close to the detector, the image appears too narrow (radiograph on right). When the object is placed too close to the rotation center, the image appears too wide (radiograph on left).

Mean Score - GP - Clinical

Image

?i cm a Laser 3

1

2

3

4

5

6

Film

T

Questions Fig. 7. Ratings of the clinical radiographs by the general practice residents. covering six basic categories (Table I). The respondents were asked to evaluate the indicated morphologic landmarks on the panoramic images on the basis of the following criteria: 1. Inadequate for diagnosis 2. Barely adequate for diagnosis

3. Adequate for diagnosis 4. More than adequate for diagnosis 5. Excellent for diagnosis Two types of subjects were used for the study: a 3M radiographic phantom and a human subject. Conventional film-based images were acquired with the use

628

Dove et al.

core Excellent

5

Adequate

3

Inadequate

1 1

2

iologist

3

-

5

4

6

T

Questions Ratings of the clinical radiographs

by the radiologists.

core - GP - Phantom Excellent

Image

5

CRT Laser Film Adequate

inadequate

1

2

3

5

4

6

T

Questions Fig. 9. Ratings of the phantom radiographs of Kodak Lanex Regular screenswith Kodak T Mat G film, The film wasprocessedwith a Kodak X-Qmat film processorand standard Kodak chemicals (Eastman Kodak Co., Rochester, N.Y.). Exposure parameters were 75 kVp, 6 mA, and I.5 seconds.Direct digital imageswere acquired at 8 1 kVp, 12 mA, and I5

by the general practice residents.

secondsand stored on an optical disk drive for later retrieval and display. The images were presented in three different sessionsto two groups of viewers. One group of viewers was composedof 10 general practice residents from the General Practice Department at the University of

Evaluation

Volume 73 Number 5

Mzan Score - Radiologist

of digital panoramic

radiography

629

- Pharrtom

Exce”ent5 -/

CRT Laser Film Adequate

3 -

Inadequate

1 1

2

3

5

4

6

T

Questions Fig. 10. Ratings of the phantomradiographsby the radiologists.

Mean Score - GP Excellenii

5

1

1

2

:3

4

5

6

T

Questions Fig. 11. Ratings of both radiographsby the generalpractice residents. Texas Health Science Center at San Antonio. The other group was composedof 10 radiologists selected from the Department of Dental Diagnostic Scielnceat the same institution. The digital images were displayed on a high-resolution monochrome monitor (CRT, Model 2064M Philips Electronics Ltd., Scar-

borough, Ontario, Canada) (Fig. 1) or as hard copies printed by a LaserTechnics (Model 300D, Lasertechnits, Albuquerque, N. Mex.) laser printer on transparenl film (Fig. 2). Conventional film images (Fig. 3) and digital laser-printed imageswere presentedon a 14 X 16 inch viewbox that had been masked to

630

Dove et al.

ORALSURG

ORAL

MED

ORALPATHOL

May

Excellent

1992

5

Adequate

3

inadequate

1 1

2

3

4

5

6

3

CRT

2

Laser

a

Film

T

Questions Fig. 12. Ratings of both radiographs by the radiologists.

eliminate any extraneous light. Only the right side of the panoramic images was presented to the observers to prevent comparisons with the contralateral side. To prevent direct comparisons by the observers, the order of presentation was (1) digital CRT, (2) digital laser-printer film, and (3) conventional film. A week was allowed to pass between viewing sessions. The resultant data consisted of 4200 responses (20 viewers X 35 questions X 3 types of images X 2 types of subjects). The data were first transformed into individual scores based on the six categories covered. In addition, a total overall score was developed with the mean score from all questions. These scores were used in an analysis of variance of each category. If a significance was found, both Fisher’s Protected LSD and Duncan New Multiple Range post hoc tests were performed to indicate which image type was rated significantly higher. A p value of < 0.05 was used to define significance in all statistical tests. The data were subdivided by viewer group (general practice resident-radiologist) and subject type (phantomclinical). RESULTS Physical evaluation Density and contrast. Fig. 4 shows a series of response curves for the TMJ region. The inverted logarithmic response curve provided by the manufacturer was consistently used.’ The use of large thick-

nesses of aluminum for the gain calibration increases the contrast of the images as indicated by the slope of the response curves. To arrive at a suitable calibration for clinical purposes, it was necessary to experiment with various combinations of kVp, mA, and filtration used for the calibration and for the actual radiograph. A tissue-equivalent phantom was used for these tests, and an exposure protocol that resulted in clinically suitable images was selected from the available choices. The programmed combinations of kilovoltage and milliamperage provided with the OP 10 equipment somewhat restricted the available choices. The calibration that appeared to give optimum contrast was a combination of 75 kVp, 6 mA, and 12 mm added aluminum filter. With this calibration, a suitable density for the tissue equivalent phantom was obtained at 8 1 kVp, 12 mA and the ordinary filtration present in the OPlO. Magnification and distortion. Fig. 5 shows a digital radiograph of the chain positioned along the central plane of the OPlO. To avoid superimposition of ghost images, the chain was placed only along one half of the central plane. The roundness of the spheres indicates that the image geometry of the film-based system is correctly simulated by the digital system. With different programming of the acquisition system, the position and/or shape of the correctly proportioned plane can be shifted. With the programming described above, a round

Volume 73 Number 5

object positioned in the central plane of the OPlO image layer is depicted in its correct proportions, whereas objects positioned outside are portrayed with distorted proportions (Fig. 6). Identical distortion effects are obtained in conventional panoramic images. One difference, however, between the digital and conventional images is readily apparent. In conventional panoramic radiography, objects positioned outside the central plane of the image layer are not only distorted but are also blurred during the projection process. In the present digital system this is not the case. The images remain sharp over a wide range of object depths because of the one-dimensional detector used in the present system for image registration. Resolution. In the vertical dimension, the bar pattern is visible up to approximately 2.9 line pairs per millimeter. This is close to the value that might be predicted by projecting the Nyquist frequency8 that corresponds to the 0.225 detector pitch to the object plane using the 1.27 magnification factor in the anterior region of the OPlO. When the same test pattern was radiographed with the Lanex Regular/T-mat G screen-film combinati.on, the bar pattern was 1ik:ewise visible to approximately 2.9 line pairs per millimeter. In the horizontal dimension, the test pattern disappears at approximately 2.25 line pairs per millimeter in the digital images as compared with 2.5 line pairs per millimeter in the ordinary panoramic radiographs. The limiting factor for the digital system is the width of the detector elements.* In the present prototype, this width is 0.5 mm. Image Quality Evaluation

For the clinical subject, there were no significant differences (p > 0.05) between the three images as rated by the general practice residents (Fig. 7). Although there was no statistical significance (JJ > O.OS), it is of interest to note that these residents rated the digital CRT image higher than conventional film-based or digital laser-printed images in four of the six areas surveyed and in the overall score. Even though the radiologists rated the: conventional filmbased image higher than either of the digital representations of the clinical image, no significance (JJ > 0.05) could be demonstrated except for the trabecular bone pattern representation (Fig. 8). In the case of the radiographic phantom, the general practice residents rated the conventional filmbased image superior to the digital CRT image only in the depiction of the area of the cementoenamel junction. In all other categories there was no significant difference (p > 0.05) in their rating. The laserprinted digital image was rated inferior to both the CRT and the conventional film-based images in two

Evaluation of digital panoramic radiography

631

of the six categories, as well as in the overall score (Fig. 9). The radiologists again rated the conventional film-based image superior to either of the digital images in all categories with the exception of periodontal membrane space in the apical region and the interdental septal bone level (Fig. 10). When images for both subject types (clinical and phantom) were evaluated, the ratings of the general practice residents were not significantly different (p > 0.05) between the conventional film-based images and the digital CRT images in any category (Fig. 11). The conventional film-based images were rated superior to the laser-printed digital images in the category of general morphologic landmarks and the total score over all questions. The radiologists rated the conventional film-based images significantly (p < 0.05) higher than either the CRT or laserprinted digital images (Fig. 12). DISCUSSION

Film-based panoramic systems are relatively inflexible with respect to the selection of radiographic contrast. The shape of the characteristic curve is a fixed property of the screen-film combination, and the adjustment of contrast by means of kilovoltage is limited by the wide spectral distribution of the x-ray beam and by the inability of most panoramic systems to vary kilovoltage and milliamperage independently. A digital system, on the other hand, has a great deal of flexibility in regard to contrast. Calibration protocol and look-up tables used to specify the gray scale response curve, as well as the contrast and brightness settings of the display monitor, can all be changed to adjust the scale of contrast to satisfy the preference of a particular clinician or the requirements of a particular diagnostic task. In addition, numeric processing of the captured image permits an additional degree of flexibility. The resolution in both the vertical and the horizontal dimensions appears to be only marginally inferior to present screen/film combinations used in clinical practice. These findings are especially encouraging, in light of the substantial dose reductions that may be possible with direct digital acquisition. The limit of the possible resolution is dependent on the size, sensitivity, and signal/noise ratio of the detector elements. New detector and display technology that will allow even better resolution than is presently possible may become available in the near future. The digital image has magnification and distortion properties similar to those seen with conventional film-based panoramic radiography. For this reason, positioning is still an important consideration. A unique difference is that this technique is not layer-

632

Dove et al.

forming. As a consequence, the digital image remains sharp over a wide range of object depths. Clinically, this has advantages and disadvantages. On the positive side, the system appears to be less sensitive to positioning errors. Leite et a1.9 have demonstrated a similar effect with narrow-beam collimation used with a conventional film-based panoramic system. They demonstrated that the widened image layer produced by the narrow beam provides for increased diagnostic accuracy of lesion detection in the anterior region over conventional imaging technique when malpositioning of the patient occurs. A disadvantage is the superimposition of structures outside the region of interest that may obscure diagnostic features. In the present system, this is particularly noticeable in the anterior region, where the image of the cervical spine tends to be superimposed over the structures of clinical interest. A radiographic image is built up from image elements and image details that together constitute structural details. They form a complex structural entity that creates a representation of anatomic information,10 With this the case, anatomic criteria for the evaluation of radiographs have often been used to supplement physical measurements usually performed when radiographic systems are evaluated. 11, t2 In this study, the visualization of specific anatomic landmarks was chosen as the criterion of image quality. The diagnosis of pathologic conditions is made possible only when this condition has caused perceptible changes in the normal anatomic appearance. Thus it is logical to assume that structural details that are relevant in the structure recognition are also relevant in a diagnostic situation. Overall, the results of the image quality evaluation indicate that the digital imaging modalities performed on a par with the conventional film-based modality in the representation of normal anatomic structures of the clinical human subject. There appeared to be more differences in the representation of these features in the phantom images. The general practice residents consistently rated the digital images higher than did their radiologist counterparts. This may be because of the more critical evaluation and familiarity with panoramic images of the radiologists. No consistent trends were found to indicate any inherent deficiencies in the depiction of any one area by the digital panoramic x-ray device.

ORALSURGORAL

4iiED ORAL PATH01 May 1992

While these preliminary results indicate the promise of direct digital acquisition as a method of panoramic imaging, the central test of any diagnostic system is the accurate and reliable representation of the state of the object under evaluation. Further studies are necessary to evaluate the diagnostic accuracy of the digital panoramic x-ray device in order to determine the clinical value of this new technology. REFERENCES WD, Dove SB, Welander U, Tronje 6. An electronic 1. McDavid system for the digital acquisition of rotational panoramic radiographs. ORAL SIJRG ORAL MED OFWL PATHOL 199131: 499-502. 2. Munier B, Roziere G, Prieur P, Rougeot H. Solid state 1024 pixel linear x-ray detector. Boulogne, France: Thomson-CSF Electron Tube Division, Technical reference, 1987:2-3. 3. Univision Technologies, Inc. Technical product reference, 1988. U, McDavid WD, Morris CR. Imaging 4. Tronje G, Welander characteristics of seven panoramic x-ray units, Part IV. Horizontal and vertical magnification. Dentomaxillofac Radio1 1985;8(suppl):29-34. 5. McDavid WD, Tronje G, Welander U, Morris CR.. Imaging characteristics of seven panoramic x-ray units. Part II. the image layer. Dentomaxillofac Radio1 1985;8(suppl):13-19. 6. Hurlburt CE. Faster screen/film combinations for cephalometric radiography. ORAL SURG ORAL MED ORAL PATWOL 1981;52:661-5. I. Vucich JJ. The role of anatomic criteria in the evaluation of radiographic images. In: Haus AG, ed. The physics of medical imaging: recording system measurement and techniques. New York: American Institute of Physics, 1979:573-87. 8. Tesic MM, Mattson RA, Barnes GT, Sones RA, Buckney JB. Digital radiography of the chest: design features and considerations for a prototype unit. Radiology 1983;148:259-64. of increased 9. Leite LP, Weems RA, Webber RL. Evaluation slit-beam collimation in panoramic radiography. J Dent Res 1990;69:344. 10. Dahlin H, Welander U, Wilbrand H. Clinical comparison between monochrome color film and black-and-white film. Acta Radio1 1978;(supp1)356. of laboratory and 11. Sickles EA, Genant HK, Doi K. Comparison clinical evaluations of mammographic screen-film systems, In: Grey JE, Hendee WR, eds. App!ication of optical instrumentation in medicine. VI. Boston. Society of Photo-Optical Instrumentation Engineers, 1977:30-35. 12. Proto AV Lane EF. 350 kVp chest radiography: review and comparison with 120 kVp. Am J Roentgen01 1978;130:859-66. Reprint requests: S. Brent Dove, DDS, MS Department of Dental Diagnostic Science University of Texas Health Science Center 7703 Floyd Curl Drive San Antonio, TX 78284