Flat-Panel Display (LCD) Versus High-Resolution Gray-Scale Display (CRT) for Chest Radiography: An Observer Preference Study

Transcription

1 - Balassy et al. Display Preferences for Chest Radiography Csilla Balassy 1 Mathias Prokop 1,2 Michael Weber 1 Johannes Sailer 1 Christian J. Herold 1 Cornelia Schaefer-Prokop 1,3 Received February 12, 2003; accepted after revision August 10, Department of Radiology, University of Vienna, Währinger Gürtel 18-20, Vienna A-1090, Austria. Address correspondence to C. Balassy. 2 Department of Radiology, UMC, Utrecht, The Netherlands. 3 Present address: Department of Radiology, Amsterdam Medish Centrum (AMC), University of Amsterdam, The Netherlands. AJR 2005;184: X/05/ American Roentgen Ray Society Flat-Panel Display (LCD) Versus High-Resolution Gray-Scale Display (CRT) for Chest Radiography: An Observer Preference Study OBJECTIVE. Our objective was to compare cathode ray tube (CRT) display with liquid crystal display (LCD) for soft-copy viewing of chest radiographs in a clinical setting. MATERIALS AND METHODS. We displayed 80 posteroanterior digital chest radiographs side by side on a 5-megapixel CRT display and a 3-megapixel LCD. Gradation characteristics of both monitors were adjusted to DICOM display standards. Using a 4-point scale, seven radiologists ranked overall image quality and visibility of anatomic landmarks. Data analysis included Wilcoxon s rank sum test to assess the significance of preference for the different display modes and calculation of the percentage of images ranked equally by at least five of the seven radiologists. RESULTS. Wilcoxon s rank sum test found significant preferences (p < 0.001) for the CRT display for visualization of structures in low-attenuation areas of the thorax and for the LCD for visualization of structures in high-attenuation areas of the thorax. Overall image quality was ranked equal by at least five radiologists in 70% of cases, whereas for the remaining images a significant preference was found for the CRT display. CONCLUSION. We conclude that, under subdued ambient lighting conditions and without use of windowing, for most images the overall quality is equal with high-resolution CRT display and LCD. In images judged preferentially, we found a significant superiority for LCD for delineating mediastinal structures and for CRT display for delineating structures in the lung. ith the rapid development of WPACS in radiology, digital image interpretation is being introduced at an ever-growing number of hospitals. The technical, financial, and practical advantages of digital radiology can be exploited to their full potential only when image evaluation relies solely on soft-copy display. Results of several previous studies in various areas of diagnostic imaging have shown that soft-copy image quality is at least comparable to conventional hard-copy imaging and that primary soft-copy evaluation is feasible [1 6]. Most of these studies used cathode ray tube (CRT) displays and focused on the impact of technical parameters, such as luminescence, spatial resolution or data processing, and diagnostic performance [7 9]. Liquid crystal displays (LCDs) are common in consumer electronics but only recently have been introduced for soft-copy interpretations in radiology. Compared with CRT displays, LCDs are characterized by a lower matrix size but a higher small-spot contrast ratio and larger dynamic range [10 12]. Although initial studies described a relatively high number of missing pixels ( black holes ) per panel, most recent publications reported excellent spatial resolution and a high uniformity and almost complete elimination of veiling glare [12]. Experience with LCDs for diagnostic evaluation is limited. A questionnaire that evaluated keyboard usage, acceptability of monitor screen, and display size did not find significant differences between the LCDs and the CRT displays [10]. No further evaluations of diagnostic performance or more specific assessments of image quality in a clinical setting have been performed, to our knowledge. The purpose of our study was, therefore, to compare the image quality of CRT displays and LCDs in a clinical setting by subjective assessment of the quality of side-by-side images. Materials and Methods Study Group We retrospectively selected 80 upright posteroanterior chest radiographs for the study. The study group included 40 radiographs with normal findings and 40 radiographs with abnormal findings, for which lesion conspicuity was classified as being low 752 AJR:184, March 2005

2 Display Preferences for Chest Radiography in 12 cases, moderate in 13 cases, and high in 15 cases. Radiographs were selected to encompass a variety of pulmonary pathology and to represent a wide range of patient constitution (Table 1). Classification of normal and abnormal findings was based on a consensus interpretation by two experienced chest radiologists who were not otherwise involved in the study. All chest radiographs were clinically indicated, having been requested by clinicians as part of the patient s diagnostic workup regardless of this study. At the time our study was conducted, our ethics committee did not require its approval or patient-informed consent for a study of this type, because patient reports had been based on hard-copy interpretation, and softcopy interpretation was not yet routinely used in our department for projection radiography. The study setup was in agreement with the Helsinki declaration, according to which all patient-related information (such as name or identification number) was obscured during interpretation. Image Acquisition and Processing We obtained the posteroanterior chest radiographs with a flat-panel digital radiography system TABLE 1 Summary of Pathologic Findings Finding No. Focal pulmonary disease 14 Atelectasis 2 Infection 8 Granuloma 2 Metastasis 1 Cyst 1 Diffuse pulmonary disease 11 Congestion 5 Emphysema 2 Fibrosis 3 Chronic bronchitis 3 Displacement of the mediastinal contours 12 because of Cardiomegaly 10 Lymphadenopathy 1 Struma (goiter) 1 Chest wall abnormality 9 Pleural effusion 5 Rib fracture 1 Adhesion 1 Chest wall resection 1 Pneumothorax 1 (Digital Diagnost, Philips Medical Systems). Exposures were taken with 125 kvp, an integrated 12:1 grid, and a 2-m detector tube distance using automatic exposure control that was adjusted to a system sensitivity of 400. Images were processed with the same algorithm that is routinely used in our institution for processing hard copies of chest films and is based on unsharp mask filtering (lungs adjusted to an optical density of 1.8, a gamma of 2.6, detail contrast enhancement of 0.4, and noise reduction of 0.8). The image matrix was 2,941 3,021 pixels, with a pixel size of mm. Monitor Technology Images were evaluated on a high-resolution 5- megapixel monochrome (gray-scale) CRT display (HB 2183, Agfa) and a 3-megapixel LCD (C3, Dome). Actual matrix resolutions were 2,048 1,536 for the LCD and 2,560 2,048 for the CRT display. The theoretic maximum brightness was 515 candelas (cd)/m 2 for the LCD and 600 cd/m 2 for the CRT display. The CRT faceplate was covered by an antireflective coating. The two displays had equal screen sizes (30 40 cm). Display functions agreed with the DICOM standard based on the Barten model [13] to ensure a consistent image appearance with both devices. Maximum luminance was adjusted to 300 cd/m 2 for both displays; minimum luminance was set as low as possible ( 0.3 cd/m 2 ). Maximum and minimum luminance values were set under the same ambient lighting conditions while the interpretation took place. Image Evaluation and Interpretation Methodology The two monitors were side by side on a standard table so that lighting conditions were identical for both. The ambient lighting in the room was subdued (< 20 lux). No additional image processing was applied for the monitor displays. No online processing, such as magnification or windowing, was available. Viewing distance could be adjusted to individual preference; viewing angle was consistently close to 90 with both displays to eliminate the effects of off-angle viewing. The images were evaluated by seven radiologists, two of whom were senior chest radiologists, one with more than 15 years of experience in digital radiology. The other five were senior residents (fourth and fifth year). All radiologists were familiar with soft-copy interpretation of cross-sectional images on a CRT display but had variable experience with soft-copy interpretation of radiographic studies. All radiologists had only limited experience with LCDs. Loading of images from the PACS workstation onto either monitor took less than 2 sec. Evaluation time per image was unlimited. The posteroanterior chest radiographs of the 80 patients were presented in a random order that was different for each radiologist. The two monitors displayed the same radiograph simultaneously for direct comparison. Radiologists were asked to grade, subjectively, the delineation of a number of anatomic landmarks that were within high- and low-attenuation areas of the chest radiograph. Within the high-attenuation area, radiologists scored the visibility of the lower trachea, the carina, and retrocardiac vascular structures. Within the low-attenuation areas, they scored the visibility of the peripheral vessels in the 2-cm-wide subpleural space, of the perihilar structures in the 2-cm-wide perihilar space, and of the lung parenchyma, including abnormal densities. We used a subjective scoring system ranging from 1 to 4, with 4 representing very good, 3 good, 2 satisfactory, and 1 insufficient visibility of the structure under evaluation. The radiologists were instructed to rank visibility of structures on the monitor display by coding appreciated differences in image quality with appropriate scores. Equal visibility ranked equally; radiologists were not forced to express a preference for one monitor over the other. Overall image quality also was classified using the same 4-point scale. Radiologists were asked to determine whether they appreciated differences in image quality that would potentially influence the detection or interpretation of abnormalities. However, we did not specifically assess the detection of thoracic abnormalities. Data Analysis A power analysis for the comparison of the two monitor displays was performed using the nquery Adviser program, version 5.0 (Statistical Solutions). The error value was set at 0.05, and the β-error value and the power were calculated. Cohen s definition of effect sizes was applied as follows: A small-effect size ( ) meant a nonoverlap in 0 15%, a medium-effect size ( ) meant a nonoverlap in 21 33%, and a large-effect size ( ) meant a nonoverlap in 38 47% [14]. Wilcoxon s rank sum test was used to assess the significance of differences between ratings at a p level of less than This was performed on a reviewer-by-reviewer basis and after averaging the scores over the seven radiologists. We also calculated the proportion of images (percentage of n = 80) for which multiple radiologists agreed on equal display quality or uniformly preferred one of the two monitor displays. The threshold was arbitrarily determined at a level of five, meaning that five, six, or all seven radiologists agreed on an equal or preferential image quality. A two-sided binomial test was used to assess whether the numbers of images that had been rated uniformly to be superior by at least five AJR:184, March

3 Balassy et al. radiologists were significantly different for the two displays (p < 0.05). Results The results of the power sample size calculation indicated a power of 100%, 86%, and 23% for large-, medium-, and small-effect sizes, respectively (Table 2). Overall image quality was ranked equal by at least five radiologists in 70% of cases. For the remaining images, overall image quality was judged significantly superior with the CRT display (p < 0.001). Agreement on equal display quality was reached by at least five of the seven radiologists for retrocardiac vessels in 85% of cases, for perihilar structures in 75% of cases, and for peripheral vessels in 74% of cases. For the contours of the trachea and for the lung parenchyma, the proportion of images with equal display quality was lower, at 51% and 54%, respectively (Table 3). For the images judged differently, Wilcoxon s rank sum test found significant preferences (p < 0.001) for the CRT display in visualization of structures in low-attenuation areas of the thorax and for the LCD in visualization of structures in high-attenuation areas of the thorax. Delineation of the tracheal contour and retrocardiac vessels was ranked superiorly with LCD, whereas delineation of perihilar structures, peripheral vessels, and the lung parenchyma was ranked superiorly with CRT display. Results of Wilcoxon s rank sum test for averaged interpretations and individual interpretations are summarized in Tables 4 and 5. Discussion The increasing number of PACS installations is inevitably associated with a growing transition from hard-copy evaluation to the exclusive use of soft-copy evaluation [15]. Although soft-copy evaluation readily is accepted for cross-sectional imaging methods, such as CT or MRI, primarily because of organizational issues, acceptance of soft-copy displays for primary evaluation of radiographic images is more tentative. Yet, multiple studies show an equal performance for soft-copy and hard-copy interpretations for different indications and specific imaging requirements [1, 6, 16 19]. Some studies, using CRT displays, concentrated on the impact of technical parameters such as spatial resolution or brightness [10, 20, 21], the availability of windowing [19, 22], or the impact of ambient lighting [22, 23]. The most recently introduced active matrix LCDs offer some ergonomic, financial, and display-related advantages over the traditional curved-surface CRT displays. The advantages of flat-panel LCDs include elimination of distortion artifacts, lower susceptibility to light reflections, and a shallower and lighter monitor [10]. Despite the slightly lower matrix size of LCDs, most recent evaluations underline the excellent performance of LCDs based on physical parameters [12]. Experience with LCDs for diagnostic evaluation is limited. Therefore, the purpose of our study was to test the image quality of both monitor types for the evaluation of chest radiographs in a clinical setting. The setup was a subjective preference study using direct side-by-side comparison. The selected study group of 80 images showed a broad range of patient constitution and a variety of pulmonary abnormalities that were thought to be representative of a clinical setting. To evaluate the significance of difference for a medium- or large-effect size, the power of the study setup was sufficiently high, at 86% and 100%, respectively. Wilcoxon s rank sum test was applied to assess the significance of preferences. Results TABLE 3 TABLE 2 Parameter Results of the Power Analysis Small Medium Large D value a Nonoverlap (%) Power (%) Note. Power analysis was performed using nquery Advisor, version 5.0 (Statistical Solutions). a size, which is difference of means divided by SD. suggested that the LCD was preferred (p < 0.001) for the delineation of structures in high-attenuation areas such as the mediastinum. The CRT display was preferred (p < 0.001) for the delineation of structures in lowattenuation areas such as the lung. These differences likely would have lost importance had online windowing been available. Based on the physical properties of CRT displays, their slightly higher matrix may have contributed to the radiologists preference of them for displaying lung parenchyma, and similarly, the higher small-spot contrast ratio of the LCD may have contributed to the superior delineation of the contrast differences in the high-attenuation area of the mediastinum Comparison of CRT and LCD Image Quality, Based on Ratings of at Least Five of Seven Radiologists Difference a Mediastinum 41 (51) 6 (7.25) Retrocardiac vessels 68 (85) Not significant Perihilar structures 60 (75) 1 (1.25) Not significant Peripheral vessels 59 (74) 2 (2.5) Not significant Lung parenchyma 43 (54) 4 (5) Not significant Overall image quality 56 (70) Note. Data are numbers of images (n = 80), with percentages in parentheses. Dash ( ) indicates no agreement on superior image quality for either display type by at least five of the seven radiologists. CRT = cathode ray tube, LCD = liquid crystal displays. TABLE 4 Comparison of CRT and LCD Image Quality, Based on Ratings Averaged over Seven Radiologists Difference Mediastinum < Retrocardiac vessels < Perihilar structures < Peripheral vessels < Lung parenchyma < Overall image quality < Note. All differences were significant using Wilcoxon s rank sum test (p < 0.05). Preferred display is marked in bold. CRT = cathode ray tube, LCD = liquid crystal displays. 754 AJR:184, March 2005

4 Display Preferences for Chest Radiography TABLE 5 Comparison of CRT and LCD Image Quality, Based on Individual Ratings by Seven Radiologists Difference Mediastinum < < < < < < < Retrocardiac vessels < < < < < < < Perihilar structures < < < < < < < Peripheral vessels < < < < < < < Lung parenchyma < < < < < < < Overall image quality < < < < < < < Note. differences was tested using Wilcoxon s rank sum test (p < 0.05). Preferred display is marked in bold. CRT = cathode ray tube, LCD = liquid crystal displays. [24]. Although the image appearance of both display devices closely complied with DICOM standards, it is problematic how much the slightly different gamma characteristics of the displays influenced these results. The lesser familiarity of the radiologists with LCDs also may have influenced the results. All radiologists confirmed that, according to their subjective impression, the appreciated differences in image quality had no relevant impact on diagnostic performance, although this factor was not specifically tested. A recently published study that evaluated the detection rate of intrapulmonary nodules using receiver-operating-characteristic methodology also reported no significant performance differences [25]. Two other studies from our department, which tested the ability to detect catheter fragments and simulated subtle pulmonary lesions with identical soft-copy interpretation hardware, similarly found no significant performance differences between the two display types. Although the subjective preferences for the two monitor types were statistically significant using the nonparametric Wilcoxon s rank sum test, these preferences were based on relatively few images, whereas for most images, delineation of structures was judged to be equal. In 70% of the images, at least five radiologists agreed that overall display quality was equal. For no examination was a clear preference (by at least five of seven radiologists) seen for either of the two monitors with regard to overall image quality. For various anatomic areas, the rates of images with equal display quality were even higher (85% for retrocardiac vessels, 75% for perihilar structures, and 74% for peripheral vessels). The mediastinum and the lung parenchyma represented those anatomic areas that were subject to lower agreement among radiologists and for which the rates of equal display quality were lower (e.g., 51% and 54%, respectively). Potential disadvantages of the CRT display include distortion artifacts caused by the curved front surface and its increased susceptibility to light reflections. Our results do not suggest a systematic deterioration of image quality in the lung because of distortion artifacts, although we did not specifically address this problem. Other clinical indications, such as stereotactic biopsies or orthopedic applications, may be more prone to these artifacts than is chest radiography [11]. Because interpretation conditions were kept constant and equal for both displays, no comment can be made about the susceptibility to ambient AJR:184, March

5 Balassy et al. lighting. A previous study found LCD to be superior to CRT display in discriminating gray-scale differences under bright ambient light [22]. The monitors tested in that study had relatively low luminance values (100 cd/ m 2 for the CRT display vs 250 cd/m 2 for the LCD) and a rather low spatial resolution, so that the results may not be applicable readily to chest radiography. However, our study had some limitations: First, a potential preference bias could not be eliminated because both monitor types could be identified readily. Second, online processing, such as magnification or windowing, which represents an essential part of soft-copy interpretation, was not included in our study setup. This exclusion was to facilitate the study design and to reduce the influence of confounding factors such as individual interpretation habits and variable familiarity with the use of the workstation, all of which affect detection performance but are difficult to control. A third limitation was that we did not specifically address the issue of ability to visualize abnormalities on the images. Addressing this issue would have eliminated the impact of the individual radiologist s detection performance on our results, but our aim was rather to focus the radiologists attention purely on the delineation of anatomic landmarks. Although the delineation of these structures allows a good estimate of image quality, further studies are required to evaluate the ability to detect pathologic structures. A fourth limitation was that the interpretation sessions were rather short (< 45 min), and other aspects such as eye strain and overall fatigue, which have been reported to be greater for CRT displays, were not studied. Finally, another aspect worth considering is the impact of off-angle viewing. Optimal performance depends on a rectangular viewing angle more for the flat-panel display than for the curved-surface monitor. Luminance and contrast of LCDs correlate strongly with viewing direction. Although in this study setting, the effect of off-angle viewing effectively was eliminated by correct placement of the radiologist (directly on-axis in front of the monitor), this issue could become more important in a busy radiologic department. We conclude from our data that, under ideal viewing conditions (subdued ambient lighting, no off-angle viewing) and without online windowing, the overall visibility of anatomic landmarks is equal with high-resolution CRT display and LCD for most images. If differences are appreciated, the LCD appears significantly superior for delineating structures in the mediastinum, and the CRT display appears significantly superior for delineating structures in the lung. References 1. Brill PW, Winchester P, Cahill P. Computed radiography in neonatal and pediatric intensive care units: a comparison of 2.5 K 2 K soft-copy images vs digital hard-copy film. Pediatr Radiol 1996;26: Fajardo LL, Hillman BJ, Pond GD, Carmody RF, Johnson JE, Ferrell WR. Detection of pneumothorax: comparison of digital and conventional chest imaging. AJR 1989;152: Kehler M, Albrechtsson U, Andresdottir A, Larusdottir H, Lundin A. Accuracy of digital radiography using stimulable phosphor for diagnosis of pneumothorax. Acta Radiol 1990;31: Don S, Cohen MD, Kruger RA. Volume detection threshold: quantitative comparison of computed radiography and screen-film radiography in detection of pneumothoraces in animal model that simulates the neonate. Radiology 1995;194: O Connor PJ, Davies AG, Fowler RC. Reporting requirements for skeletal digital radiography: comparison of soft-copy and hard-copy presentation. Radiology 1998;207: Kundel HL, Polansky M, Dalinka MK. Reliability of soft-copy versus hard-copy interpretation of emergency department radiographs: a prototype study. AJR 2001;177: MacMahon H, Metz CE, Doi K, Kim T, Giger ML, Chan HP. Digital chest radiography: effect on diagnostic accuracy of hard copy, conventional video, and reversed gray scale video display formats. Radiology 1988;168: Goodman LR, Foley WD, Wilson CR, Rimm AA, Lawson TL. Digital and conventional chest images: observer performance with Film Digital Radiography Systems. Radiology 1986;158: Krupinski EA, Maloney K, Bessen SC. Receiver operating characteristic evaluation of computer display of adult portable chest radiographs. Invest Radiol 1994;29: Pavlicek W, Owen JM, Peter MB. Active matrix liquid crystal displays for clinical imaging: comparison with cathode ray tube displays. J Digit Imaging 2000;13[2 suppl 1]: Roehrig H, Krupinski EA, Furukawa T. Evaluation of a flat CRT monitor for use in radiology. J Digit Imaging 2001;14: Roehrig H, Fau J, Krupinski EA. LCDs versus CRTs: a comparative performance evaluation. (abstr) Radiology 2002;225(P): Barten PGJ. Physical model for the contrast sensitivity of the human eye. Proc SPIE 1992;1666: Cohen J. Statistical power analysis for the behavioral sciences, 2nd ed. Hillsdale, NJ: Lawrence, Earlbaum Associates, Mertelmeier T. Why and how is soft copy reading possible in clinical practice? J Digit Imaging 1999;12: Steckel RJ, Batra P, Johnson S. Comparison of hard- and soft-copy digital chest images with different matrix sizes for managing coronary care unit patients. AJR 1995;164: Razavi M, Sayre JW, Taira RK. Receiver operating characteristic study of chest radiographs in children: digital hard-copy film vs 2K 2K softcopy images. AJR 1992;158: Cohen MD, Katz BP, Kalasiniski LA, White SJ, Smith JA, Long B. Digital imaging with a photostimulable phosphor in the chest of newborns. Radiology 1991;181: Kronemer KA, Don S, Luker GD, Hildebolt C. Soft-copy versus hard-copy interpretation of voiding cystoureterography in neonates, infants, and children. AJR 1999;172: Otto D, Bernhardt TM, Rapp-Bernhardt U. Subtle pulmonary abnormalities: detection on monitors with varying spatial resolutions and maximum luminance levels compared with detection on storage phosphor radiographic hard copies. Radiology 1998;207: Graf B, Simon U, Eickmeyer F, Fiedler V. 1K versus 2K monitor: a clinical alternative free-response receiver operating characteristic study of observer performance using pulmonary nodules. AJR 2000;174: Fuchsjaeger MH, Eisenhuber E, Strasser GR, Funovics MA, Prokop M, Schaefer-Prokop CM. Impact of ambient light and windowing on detectability of catheters in soft copy display of bedside chest studies. (abstr) Radiology 2001;217(P): Haak R, Wicht MJ, Hellmich M, Nowak G, Noack MJ. Influence of room lighting on grey-scale perception with a CRT- and a TFT monitor display. Dentomaxillofac Radiol 2002;31: Badano A, Flynn MJ, Kanicki J. Accurate small-spot luminance measurements. Displays 2002;23: Siegel EL, Reiner BJ. Comparison of the clinical performance of a high resolution active matrix LCD to a CRT monitor in the assessment of lung nodules using computed radiographic images. (abstr) Radiology 2002;225(P): AJR:184, March 2005