Image Display and Perception

Transcription

1 Image Display and Perception J. Anthony Seibert, Ph.D. Department of Radiology UC Davis Medical Center Sacramento, California, USA Image acquisition, display, & interpretation X-rays kvp mas Tube filtration Collimation Patient Size Restraints Exam type ESE, dose Detector Efficiency Resolution Scatter grid DQE Technologist/engineering domain Computer Digitization Preprocessing Postprocessing Configuration Physicist & Radiologist domain PACS Data delivery Data display Data storage Workflow PACS admin domain Human Radiologist Physician Experience Condition 1

2 Outline of Presentation Human visual response Monitor pixel size Types of monitors Luminance, Illuminance and ambient lighting Pan and zoom DICOM GSDF and monitor calibration Viewing pitfalls Quality control Visual Accommodation Determines optimal field of view and monitor dimensions 2

3 Visual Accommodation Average resting point of accommodation 31 inches (75 cm), or arms-length viewing distance Field of view for a specific viewing distance Diagonal dimension should be ~80% of distance Normal viewing at 31 (75 cm) : about 21 (50 cm) From Mike Flynn Henry Ford Hospital Human visual response Frequency variations at fixed contrast Contrast variations at fixed spatial frequency 3

4 Contrast Sensitivity 10/26/2012 Spatial Acuity 60 cm viewing distance 100 cd/m 2 Luminance 21 mm 2 test pattern # pixels / degree and viewing distance determines pixel dimension and matrix size required mm mm % max Spatial Frequency, mm -1 Corresponds to a pixel dimension of ~200 m Monitor pixel size? 4

5 Retina Display Pixel density is such that the human eye is unable to notice pixelation at a normal viewing distance Portable imaging displays Calgary Scientific Resolution MD FDA approved for use iphone, ipad, web 5

6 Types of Monitors Medical Grade Higher luminance (>300 cd/m 2 ), good warranty > 3y Larger bit depth for display, bits Built-in calibration, typically dedicated video card High pixel density, smaller pixels $$$$ Professional / Pro-sumer Commodity/Consumer Grade Lower luminance, longevity/warranty limited Typical 8 bit performance Calibration external $$ Diagnostic Monitors 5 MP (2500 x 2000) 600 Cd/m 2 Monochrome 3 MP (2000 x 1500) 400 Cd/m 2 Monochrome or Color 6

7 Intermediate pro-sumer 76 cm diagonal 2560 x 1600 ~ 0.21 mm/pixel 4 MP = 2 x 2 MP >300 cd/m 2 <$1200/monitor Dual-link DVI video card ($100) Dual monitor setup Technologist QC monitors 2 MP and lower 1600x x x768 Often suboptimal --insufficient resolution --insufficient luminance 7

8 Technologist QC monitor Insist on high quality Mammography: 3 MP, >250 cd/m 2 Technologist now able to discern motion at point of care. Medical or Consumer Grade? Medical: Higher luminance (up to 1000 cd/m 2 ), higher pixel count, higher up-front costs, longer warranty, self-calibration Consumer: Lower luminance, much lower cost, reduced matrix size, external calibration. wide range of models Overall long-term cost? Perhaps similar 8

9 Considerations for Displaying Medical Images on LCD panels LCD glass varies in quality Types of LC switching TN (Twisted Nematic): low grade, severe angular brightness loss PVA (patterned vertical array): intermediate grade, much better viewing performance IPS (In-Plane-Switching): high grade, good angular response, good speed Unpolarized fluorescent light source Horizontal Polarizing Filter LC layer Vertical Polarizing Filter Off Dark Pixel Output Bright On Horizontally Polarized Light Vertically Polarized Light 9

10 Considerations for Displaying Medical Images The matrix size of a radiographic image is often much greater than display matrix size In order to view the whole projection image field of view (FOV), spatial down-sampling (interpolation) is required In order to view the image at its native resolution, a one-toone image-to-monitor pixel mapping is necessary For large images, a sub-image can be displayed at native resolution, and panning the image is necessary This takes more time for the radiologist s diagnosis Typical consumer-grade display monitor Typical radiograph Choices: 1. Display the full image 2. Display full resolution 3. Other (in-between or mag) 10

11 1760 Image 1200 x 1600, full resolution Zoom and Pan 2184 Display with full image resolution 1200 Display x 1600, partial (interpolated) resolution 2560 x 3328 image displayed in an 850 x 1100 display matrix 66% reduction Can see full image FOV 50% reduction in FOV provides better resolution Resolution test chart 11

12 25% reduction in FOV provides enhanced resolution True size (1 monitor pixel = 1 image pixel) 50% magnification (pixel replication) does not increase visible resolution, but does present expanded information for a better match to the human visual system response (you can view the monitor at a further distance) Magnification with bilinear interpolation 12

13 DICOM image quality GSDF (DICOM part 14): identifies a method to calibrate the luminance response of monitors according to just noticeable differences and perceptual linearization Concept: equal change in input pixel value (grayscale) results in equally perceived change in output luminance Barten s model fits, for a given luminance level, the change in luminance that an observer would just notice Each JND step: a single observable change in luminance DICOM monitor calibration Why not just calibrate to a linear response? Human contrast sensitivity varies with luminance Low levels very sensitive Perception difference of Cd/m 2 at ~0.05 Cd/m 2 High levels not as sensitive Perception difference of ~25 Cd/m 2 at ~4000 Cd/m 2 13

14 Luminance (Cd/m 2 ) 10/26/2012 DICOM GSDF (part 14) Perceptual Linearity Grayscale Standard Display Function Film CRT, LCD Displays Just Noticeable Difference (JND) Index Display Calibration Qualitative (simple, fast check) SMPTE pattern Grayscale Standard Display Function Quantitative adjustment of monitor Optimize human visual response to JND s AAPM TG-18 protocols 14

15 SMPTE pattern Pre Calibration Post Calibration (GSDF) AAPM TG 18: Monitor Calibration Defines methodology for acceptance testing and quality control of workstation monitors Provides test images (beyond the SMPTE image) with instructions for use Information is available on AAPM website: aapm.org/pubs/reports/or_03/pdf 15

16 General Purpose QC pattern 0% / 5% 100% / 95% QUALITY CONTROL 16

17 TG-18 Image test patterns.. and a lot more Perceptual Contrast Contrast-Luminance Bit-depth, continuous grayscale Pattern Descriptions / Uses 17

18 Methods / Analysis Detailed in executive summary of TG-18 Useful for initial comparison of monitors and acceptance test procedures.. 18

19 Open Source Calibration Software 19

20 Open Source: Henry Ford Hospital Measure GSDF for particular types of consumer-grade monitors Allow LUTS to be distributed to the enterprise (5000+ monitors..) Quantitative and qualitative verification tools are available Photometers.. $2000+ $300. Major difference: Speed Accuracy +/- 1% 20

21 Note: It s more than DICOM GSDF.. Max Luminance (L max ) to Min Luminance (L min ) Ratio: ~350 Many monitors have L max /L min > 800; add offset to L min Calibration software / photometer Drives the display and interfaces to a photometer for measurement Loads Look-Up-Table for video card to implement the Barten curve response Usable on most consumer-grade monitors and displays 21

22 GSDF calibration Pre-calibration Purchase photometer Download software from Internet Post-calibration GSDF conformance 22

23 Viewing and Environmental Considerations Display considerations Must assure uniform appearance at all calibrated display devices Uncalibrated QC monitors Images seen on PACS don t look right Tech or PACS get the blame for the poor monitor Viewing condition precautions Room illuminance, ambient lighting Angular luminance variations of LCD s Small, low quality Tech QC monitor Ambient lighting Maximum performance achieved at lux Higher illuminance allowed with higher luminance monitor 23

24 Incompatible lighting issues Mammography reading area Mammo Diagnostic Workstation PACS Workstation Breast MRI Workstation Film Viewer Ambient lighting.. 20 to 40 lux Internet Access Workstation Brenna, PC, McEntee M, Evanoff, M, et al. Ambient lighting: effect of illumination on soft-copy viewing of radiographs of the wrist. AJR 2007;188:W

25 Viewing Pitfalls Viewing angle dependence of brightness and contrast Displays are non-lambertian Molecular orientation symmetry within LC layer Some (expensive) LCD monitors correct this: Birefringent filter layers Multidomain pixels In-Plane Switching Combinations of the above Viewing pitfalls Angular luminance variation of LCD monitors Appearance of image 25

26 Image presentation might not be a calibration problem! After contrast adjustment Perception Reality on PACS workstation Compliments of Jeff Shepard, M.D. Anderson Cancer Center Quality Improvement LCD is not optimal as a QC monitor unless: The monitor is calibrated to DICOM PS3.14 (GSDF) and The angular dependence of brightness and contrast is adequately corrected (high quality monitor) Techs should be trained to view LCD s from a position that ensures a viewing angle that is normal to the plane of the display surface 26

27 Display Monitor Quality Control Who? IT staff, PACS administrator, mammo QC tech Vendor Physicist Periodic measurements? Primary: Semi-annually (weekly, mammo) Secondary: Annually Automated GSDF: Monthly What? AAPM TG-18 documentation Simple cleaning chores, weekly? At a minimum a qualitative check of calibration 27

28 Statement on Quality Control requirements Any facility using a digital image data management system must have documented policies and procedures for monitoring and evaluating the effective management, safety, and proper performance of acquisition, digitization, processing, compression, transmission, display, archiving, and retrieval functions of the system. The quality control program should be designed to maximize the quality and accessibility of diagnostic information. Summary Display monitors provide the human / image interface Careful consideration of monitor types, viewing conditions and continued quality control and consistency are crucial! Calibration is necessary for optimal information transfer The weak link in an electronic network is often the display 28