IT Guidance Documents Picture archiving and communication systems (PACS) and guidelines on diagnostic display devices Second edition Board of the Faculty of Clinical Radiology The Royal College of Radiologists
This guidance forms part of a series on the developments in information technology in radiology. This is a fast-moving field and developments are occurring rapidly. Consequently, this guidance will be updated regularly and readers should check regularly that they are using the most up-to-date guidance available. 2
These guidelines are aimed at simplifying the essentials relating to medical display devices (including picture archiving and communication system [PACS] monitors), and deal with the issues of spatial and contrast resolutions, and provide a recommended basic specification table. These guidelines should be read in conjunction with those on the ergonomics of a PACS workstation 1 to appreciate the importance particularly of ambient lighting conditions on digital display devices. This document replaces BFCR(08)7 Picture archiving and communication systems (PACS) and guidelines on diagnostic display devices, which has been withdrawn. 1. Classification of display devices 1.1 Medical display devices can be classified as either primary (diagnostic) or secondary (review). Primary display devices are used for the interpretation of medical images. Secondary display devices are used for reviewing medical images, usually in conjunction with the report. It is recommended that all primary display devices undergo acceptance testing and regular performance review. 1.2 Specialist medical liquid crystal display (LCD) devices should be distinguished from common-off-the-shelf (COTS) LCD devices. Medical displays offer improved pixel resolution: a 2048 x 1536 pixel display is typical for a medical display, whereas COTS displays are rarely available with resolutions greater than 1920 x 1200. Medical displays can also offer improved performance in terms of luminance level, uniformity and constancy, contrast ratio, viewing angles and compliance with the Digital Imaging and Communications in Medicine (DICOM) Greyscale Standard Display Function (GSDF). Advanced management features available with medical displays include remote quality assurance and automatic quality assurance (QA) report generation. 1.3 Specialist medical LCD devices are preferred to COTS devices for primary diagnosis. Lower diagnostic performance has been demonstrated on appropriately calibrated COTS devices compared to specialist medical LCD displays. 2 2. Spatial resolution 2.1 When first viewing a digital radiographic image, the image data should be interpolated to display the image fully within the maximum available screen area since it is recognised that pattern recognition of nonspatially limited abnormalities can be overlooked if the image is not viewed as a whole. There is insufficient evidence to quantify the clinical risks associated with viewing images interpolated below their acquisition resolution, but it has been shown that any risk can be minimised by a process of systematic magnification. 2.2 On a medical imaging workstation, magnification should be achieved by using the software zoom, pan and magnification tools, as closer visual inspection will not overcome the effects of image interpolation. Ideally, images should be magnified to their acquisition resolution or to a whole number magnification factor greater (for example, x2 or x3) to avoid the risk of introducing artefacts due to image interpolation. For example, when an image is too large to be displayed fully on screen, it should be displayed at its acquisition resolution (1:1 pixel matching) and the image panned around the screen until the whole of the image has been viewed. Alternatively, the magnifying glass tool can be used to systematically magnify areas of the image to reveal the full image detail. Studies suggest that there is little reduction in the diagnostic power of using these techniques when compared to displaying the whole image at 1:1 on higher resolution screens, but there is an increase in the time taken to make a report. 2.3 When viewing multiple images, it is recommended that serial images are compared directly, side-by-side, to highlight any difference in appearance. Smaller images, such as those obtained in cross-sectional imaging, can be displayed fully side-by-side on a single screen. Larger images, such as from projection radiography (CR and DR), should ideally be viewed on separate displays to optimise the ratio of the display resolution to the acquisition resolution. 2.4 At normal viewing distances, the psychophysical performance of the human eye imposes a limit on the perceived resolution. It has been calculated that at a 60 cm viewing distance, human visual performance is well matched to a screen with 0.25 mm pixels (pixel pitch). This is equivalent to a native screen resolution of 1280 x 1024 (~1.3 MP) on a 42 cm (~17 ) display, or 1600 x 1200 (~1.9 MP) on a 50 cm (~20 ) display. Higher resolution display devices enable the user to view the image in greater detail by closer inspection of the screen, but a similar effect can be achieved with software zoom, pan and magnification tools. High fidelity dual screen displays ( 3 MP) are recommended in radiology and other areas where large numbers of radiographic images are reported, to reduce reporting times and thereby optimise department workflow. 2.5 Medical displays achieve their higher resolution by the use of a smaller pixel pitch, with most display devices for primary diagnosis being limited to a size of approximately 54 cm (22 ). COTS devices tend to use a larger minimum pixel pitch and high pixel counts are achieved primarily by increasing the physical size of the screen. The use of very large displays (>70 cm/27 ) may make achieving adequate ergonomics difficult, particularly when used in a portrait configuration, or when multiple display devices are used on a single workstation. 3
2.6 Display devices, especially COTS displays, are increasingly available in wide aspect ratios (AR) (for example, 16:9) as opposed to the older aspect ratios (4:3 and 5:4). When comparing the resolution of devices with different aspect ratios, the pixel count may be misleading, as the limiting factor is the absolute number of pixels in the limiting direction. Wide aspect ratios may not be as well suited to most radiographic images with squarer AR. However, a wide monitor with the same number of pixel rows provides a greater screen area, so may help workflow by permitting the display of more images side by side. 3. Contrast resolution 3.1 The contrast resolution of a display system depends on a number of factors, including the maximum and minimum luminance of the display device, the characteristic (luminance response) curve of the display device, the greyscale bit depth, the background ambient illumination, and the use of application software windowing tools (see below). A useful concept is the just noticeable difference (JND) index. Each JND index step corresponds to a perceivable difference in grey scale resolution. In medical image viewing, it is desirable to optimise the number of JND index steps available. 3.2 The contrast ratio of a display system is defined as the ratio of the maximum to minimum luminance that the device is capable of displaying. In general, the higher the contrast ratio of a display device, the higher the number of JND index steps that can be perceived, and the better the contrast resolution. High fidelity medical grade monochrome LCD monitors are brighter than COTS LCD displays, achieving maximum luminance values >500 cd/m2, and contrast ratios >800:1. Higher display luminance is useful in maintaining visible contrast in the presence of ambient illumination. There is some evidence that high brightness display devices with maximum luminance >500 cd/m2 can cause fatigue and impair contrast resolution by their effect on the adaptation level of the human visual system, and the optimum operating level may vary between users. It is common for medical LCDs to be specified with maximum luminance significantly higher than the recommended level. This is advantageous in prolonging the life of the display, as the luminance of the light source declines with aging. LCD contrast ratio can deteriorate significantly at oblique viewing angles, and it is recommended that all reporting is performed with the user perpendicular to the screen. 3.3 The characteristic curve of a display device is a plot of the measured luminance response for each step in the digital driving level (DDL) of the display system. The human visual system has a non-linear response to contrast resolution, with the effect that it is more difficult to perceive contrast differences at low illumination. To compensate for this, a standard curve of luminance versus JND index has been defined within the DICOM Greyscale Standard Display Function (GSDF). 3 Devices calibrated according to the GSDF are said to be perceptually linear in response, and optimised to human visual performance. 3.4 The greyscale bit depth of a display device is the number of levels of grey that can be represented by the digital driving level (DDL) of the display device. An 8-bit (per pixel) greyscale display device can represent 256 levels of grey, while a 10-bit device can represent up to 1024 total levels. For colour monitors displaying monochrome images, 24-bit and 32-bit colours are equivalent to 8-bit greyscale. Whether a user is able to perceive all the greyscale bit-depth levels that are represented depends on the maximum luminance, contrast ratio, and calibration of the display device to the GSDF (see above). In practice, a user will typically only see a fraction of the 256 levels of grey represented by a non-calibrated 8-bit display device as the luminance difference between each step in the DDL will not correspond to a perceivable difference in greyscale contrast (JND index step). By comparison, high contrast 10-bit calibrated display devices are capable of displaying >500 perceivable shades of grey (JND index steps). [Current operating system software has limited support for handling 10-bit greyscale data. Display of 10-bit data requires the image display application to be written such that it communicates directly with a 10-bit capable graphics card, and that there is a 10-bit capable connection between the graphics card and display device. Even with an 8-bit greyscale input, a 10-bit display device can outperform an 8-bit display device by ensuring all 256 input greyscale levels are represented as perceivable differences in greyscale to the end user.] 3.5 Application software windowing tools control the number of simultaneous greyscale values in an image that are presented for display. A DICOM image can encode up to 16 bits greyscale per pixel, but only a fraction of these are shown at any time. By changing the centre (level) and range (width) of the greyscale values presented, it should be possible to demonstrate all the greyscale data represented in the image. The minimum specification of a display device in terms of contrast resolution parameters is therefore somewhat arbitrary, and depends on how the windowing tools are used during normal workflow. High fidelity display devices are recommended in radiology and other areas where large numbers of images are reported to reduce requirements for windowing images, and thus assist in reporting workflow. The luminance of the light source can be reduced electronically to reach the recommended value, but once the light source has degraded and is unable to maintain the recommended luminance, the display has reached the end of its working life. 4
4 Use of colour displays 4.1 Historically, the performance of colour display technology was insufficient to meet the recommended greyscale performance targets for primary diagnosis. Advances in technology have led to the development of colour display devices that are capable of achieving these standards. Previously, this may have required a segregation of reporting work: dedicated workstations with high performance monochrome displays for high-throughput reporting of projection radiography, and workstations with colour displays for reporting of nuclear medicine or other advanced modality imaging. 4.2 The accurate reproduction of colour images is technically complex and beyond the scope of this document. Unlike the multiple national standards for medical display greyscale luminance and contrast, there are not yet similar standards for colour medical image reproduction. However, commonly used (display and graphics) industry standards for colour calibration and response are incompatible with the DICOM monochrome luminance response meaning that a display cannot be calibrated to both standards simultaneously. In the context of current radiology practice, where foreseeable use of colour is limited to false-colour images and/or coloured annotations, the risk associated with incorrect colour rendering is likely to be lower than that obtained by incorrect greyscale rendering. Modern computer operating system software (eg, Windows 7 and Mac OS X) provide a solution by the use of International Color Consortium (ICC) profiles, which inform the system software of the display device s colour response. Colour management capable software can process image colour data, so that the data sent to the display is corrected for the display s response function, and hence should provide near-optimal display regardless of the monitor s calibration (provided that the monitor s response is known, or has been measured). Alternative softwareindependent solutions to this problem have included display devices that automatically detect monochrome and coloured images, and apply the appropriate calibration to their respective screen regions. Where non-standard solutions such as this are used, their performance should be verified as part of acceptance testing. 5
5. Recommended specification 5.1 Table 1 shows the PACS and Imaging Informatics Group minimum and recommended specification for primary diagnostic display devices used for clinical image interpretation. This guidance applies to all workstations where CR, DR, fluoroscopy, ultrasound, CT, MR, nuclear medicine and PET images are viewed (excluding mammography). Table 1. Minimum and recommended specification for primary diagnostic display devices used for clinical image interpretation Screen resolution b (Native pixel array) Minimum a 1280 x 1024 c (~1.3 megapixels) Recommended a 1500 x 2000 d (~3 megapixels) Screen size (viewable diagonal) 42 cm (~17 ) 50 cm (~20 ) Maximum luminance e >170 cd/m2 f 500 cd/m2 g Luminance contrast ratio (maximum/minimum) 250:1 f,h 500:1 Greyscale calibration Within 10% GSDF h Calibrated to GSDF e Greyscale bit depth Video display interface Pixel defects j ISO 13406-2 classification 8-bit greyscale (24-bit colour) i Digital (Analogue not recommended for new installations) Class 2 (2 per million) 10-bit greyscale Digital (eg, DVI, displayport) Class 1 (0 defects) Notes a. The minimum and recommended specifications for diagnostic display devices are only appropriate if clinical image viewing is performed according to image viewing guidelines. All diagnostic image interpretation should be performed on DICOM images, making use of the application software zoom, pan, magnification and windowing tools to optimise spatial and contrast resolution. b. LCD devices should be run at their native resolution to ensure there is a 1:1 match between screen pixels and screen resolution, and therefore no loss of image quality due to screen interpolation. Displays should be configured such that the correct aspect ratio is maintained to avoid distortion of the image; this will automatically be the case is a 1:1 pixel match configuration is used. c. Where the majority of reporting performed on a diagnostic workstation is of cross-sectional imaging, lower resolution landscape style displays ( 1.3 megapixels) are considered adequate, providing larger (e.g. direct radiographic) images are interpreted with the aid of systematic magnification. d. High fidelity ( 3 megapixels) portrait style displays are recommended in radiology and other areas where large numbers of plain radiographic images are reported to reduce requirements for systematic magnification, and thus reduce image interpretation and reporting times. e. Display devices may be set initially to operate at a fraction of the maximum luminance in the manufacturer s specification. This can be adjusted to compensate for the decline in performance of the back-light over time while maintaining greyscale calibration. f. American Association of Physicists in Medicine TG18 recommendation. 4 g. High luminance displays can increase the number of perceivable greyscale levels (JND index steps) but may have a detrimental effect in user performance through fatigue and the human visual adaptation response. The optimum operating luminance level may vary between users. h. Institute of Physics and Engineering in Medicine 91 recommendation. 5 i. 24-bit and 32-bit colours are equivalent to 8-bit monochrome greyscale. Colour display devices are recommended for displaying colour images, but they generally perform less well than monochrome display devices in terms of maximum luminance and contrast ratio. j. The number of permissible pixel defects per million is defined by the ISO 13406-2 standard. Class 1 panels should have no defects. Class 2 panels should be replaced if they have >2 whole pixel defects per million. Appropriate use of application software zoom, pan and magnification tools can negate the effect of pixel defects in clinical practice. Approved by the Board of the Faculty of Clinical Radiology: 15 June 2012 6
References 1. The Royal College of Radiologists. Ergonomics. London: The Royal College of Radiologists, 2012. 2. Krupinski E. Medical grade vs off-the-shelf color displays: influence on observer performance and visual search. J Digit Imaging 2009; 22(4): 363 368. 3. DICOM Standard Status. 3.14. Grayscale Standard Display Function. http://www.dclunie.com/dicomstatus/status.html (last accessed 27/11/12) 4. American Association of Physicists in Medicine. Assessment of display performance for medical imaging systems. http://www.aapm.org/pubs/reports/or_03.pdf (last accessed 27/11/12) 5. Institute of Physics and Engineering in Medicine. Report 91. Recommended Standards for the Routine Performance Testing of Diagnostic X-Ray Imaging Systems. London: IPEM, 2005. http://www.ipem.ac.uk/publications/ipemreports/pages/recommendedstandardsfortheroutineperformance TestingofDiagnosticX-RayImagingSystems.aspx (last accessed 27/11/12) 7
The Royal College of Radiologists 38 Portland Place London W1B 1JQ Tel +44 (0)20 7636 4432 Fax +44 (0)20 7323 3100 Email enquiries@rcr.ac.uk www.rcr.ac.uk A Charity registered with the Charity Commission No. 211540 Citation details: The Royal College of Radiologists. Picture archiving and communication systems (PACS) and guidelines on diagnostic display devices. London: The Royal College of Radiologists, 2012. Ref No. BFCR(12)16 The Royal College of Radiologists, November 2012 For permission to reproduce any of the content contained herein, please email: permissions@rcr.ac.uk This material has been produced by The Royal College of Radiologists (RCR) for use internally within the specialties of clinical oncology and clinical radiology in the United Kingdom. It is provided for use by appropriately qualified professionals, and the making of any decision regarding the applicability and suitability of the material in any particular circumstance is subject to the user s professional judgement. While every reasonable care has been taken to ensure the accuracy of the material, RCR cannot accept any responsibility for any action taken, or not taken, on the basis of it. As publisher, RCR shall not be liable to any person for any loss or damage, which may arise from the use of any of the material. The RCR does not exclude or limit liability for death or personal injury to the extent only that the same arises as a result of the negligence of RCR, its employees, Officers, members and Fellows, or any other person contributing to the formulation of the material.