Room Lighting and Display Luminance in Digital Mammography – What is Needed for Quality Image Display AAPM 2004 Annual Meeting CE: Mammography Physics and Technology – 7 Jerry A. Thomas, M.S., DABR, CHP, DABSNM Assistant Professor of Radiology Department of Radiology and Radiological Sciences Uniformed Services University Bethesda, MD 20814

Overview • Monitor design/types • Monitor performance characteristics • DICOM Part 14 Grayscale Standard Display Function • Flat Panel Calibration • SMPTE Pattern Evaluation • AAPM TG-18

1

Medical Display Choices

1280 x 1024 55 fL / 200 cd/m2

1600 x 1200 35 fL / 120 cd/m2

2048 x 1536 (3 MP) 55 fL / 200 cd/m2

3840 x 2400 (9.2 MP) 80 fL / 270 cd/m2

1024 x 1280 (1 MP) 35 fL / 120 cd/m2

1280 x 1024 35 fL / 120 cd/m2

1200 x 1600 (2 MP) 75 fL / 255 cd/m2

1200 x 1600

1600 x 1200

1600 x 1200

120 fL / 400 cd/m2

235 fL / 800 cd/m2

235 fL / 800 cd/m2

2048 x 2560 (5 MP) 175 fL / 600 cd/m2

Graphic boards • BarcoMed 5MP and 2MP – 2048 x 2560 x 8 or 10 (e.g. 256 or 1024 SOG)

• Planar – 2MP, 4MP, 5 MP and 8MP – Up to 2560 x 3200

• • • •

Siemens Matrox NVIDIA – Quadra Pro Series and FX series Other Commercial Graphics Controllers – for Flat Panels

2

Monitor Performance Characteristics • Luminance Uniformity – Fixed at time of manufacture – Should be < 18% (center to edges) – Best monitors – 8% – 12%

• Contrast Ratio – Lmax/Lmin

• Phosphor Type – P45 – P104

• Anti-reflective coating – Will add color to monitor

• Glass color • Monitor gamma ~ 2.2

• Focal Spot Size • Input impedance – 50 or 75 ohms

Contrast Ratio • Depends on – – – –

Maximum luminance Glass Transmission Ambient Light White areas of the image

• German DIN standard – 100:1 for DR – 40:1 for CT, MRI

3

Phosphors • P45 – – – –

One component phosphor Low aging Short decay time after 2 ms No color variation over time

• P104 – Dual component phosphor – High efficiency – Can be run at luminance levels > 100 fL

Glass Envelope Construction • 27% total transmission • Ambient Light attenuation (93%) • Anti-reflective/antistatic coating • Same refractive index throughout

4

Resolution vs Luminance

Monitor Lifetime • Can last up to 50,000 hr • Phosphor and glass decrease • Cut-off control • Control of Gamma function

5

Long Term Reliability & Conformance of Displays The Complete “Hands-Off” Conformance Test

Controlling the White Level

Regulating the Gamma Function

Hands-off Conf Test Sampling Points

Stabilizing the Black Level

Monitor stability over time

6

Grayscale Standard Display Function • A methodology for measuring the characteristic curve of a display system – For conformance testing – For altering display system to match GSDF

• Key elements of GSDF – Target: 10% of the monitor area – Surround: 20% of Lmax

GSDF • Neither – A performance standard – Image display standard

• Does not define – – – –

Luminance Luminance range Optical density range Presentation of picture element values

7

GSDF DICOM Part 14 • What does it provide – Standardized basis for video display calibration – Table of the GSDF as a function of the JustNoticeable Difference Index

Factors that affect visual perception • Room lighting – Ambient light – Specular reflection

• Lmax and Lmin • Icons on screen • Veiling glare

8

Monitor Gamma vs Perceptual Linearization

Monitor Perceptual Linearization and JND Determination • • • •

Background is set Line Pair target in center 10% of viewing area Increase target until it is just visible Increase background to target level – Repeat previous two steps

• Example follows:

9

10

11

12

Grayscale Standard Display Function DICOM Part 14

Perceptually Linearized Response Curve

Luminance

white

black Just Noticeable Differences (No of JNDs)

Light – lux vs cd/m2

13

Optimized Softcopy Display: Putting the Pieces Together

What are the pieces needed to give an image “Pizzazz”? • Monitor Calibration • Acquisition Device – The Processed Image • Image Data Optimized for Display • The Displayed Image

14

The First Piece: Monitor Calibration • Necessary for standardization • Corrects the natural display gamma to a perceptually linearized model • Requires frequent conformance testing

The First Piece: Monitor Calibration • Setting the display black and white levels

15

Use Brightness controls to adjust the black level.

Use Contrast controls to adjust the white level.

16

The First Piece: Monitor Calibration • Setting the display black and white levels • Determining the natural gamma of the display

17

The First Piece: Monitor Calibration • Setting the display black and white levels • Determining the natural gamma of the display • Application of a Look-Up Table (LUT)

Application of an LUT Display Natural Gamma

Gamma Correction LUT Gamma Correction LUT

Display Luminance as a Function of Digital Driving Level 70000

600

60000

500

200

C o r r e c tio n F a c to r

+

300

40000

30000

20000

100

10000

0

0 0

200

400

600

800

1000

0

1200

200

400

600

800

1000

1200

Digital Driving Level

Digital Driving Level

Perceptually Linearized Display Display Luminance as a Function of Pixel Value 1000

D is p la y L u m in a n c e (c d /s q m )

D is p la y L u m in a n c e (c d /s q m )

50000

400

100

10

1 0

500

1000

1500

2000

2500

3000

3500

4000

Pixel Value

18

Conformance Validation • Manual Conformance Testing • DICOM Part 14 Grayscale Standard Display Function • Monitor Luminance Uniformity

19

Analysis of Measured Data

Conformance Validation • Manual Conformance Testing • DICOM Part 14 Grayscale Standard Display Function • Monitor Luminance Uniformity • Automated Conformance Testing • On Local Workstation • Remotely

20

Automated Conformance Testing Software - Local

Automated Conformance Testing Software - Remote

21

Automated Conformance Testing Software - Remote

The Second Piece: The Processed Image Image Acquisition

Processing

Processed Image

Processing

Detector Gain and Flatfield Corrections

Image Data Enhancement

Raw Image

22

Processed Image

Breast Image 1 algorithms • Processing unknown (proprietary)

• Is not suitable for softcopy interpretation

• Can not be optimized with linear windowing

The Third Piece: Optimization for Display • Produces: •An image with diagnostic content (pizzazz) •An image that looks familiar to the radiologist

• Accomplished through non-linear windowing

23

Non-Linear Windowing Non-Linear Windowing LUT 70000

60000

50000

40000

30000

20000

10000

0 0

500

1000

1500

2000

2500

3000

3500

4000

4500

Pixel Value

Optimized Image Breast Image 2

24

How Do the Pieces Fit Together? Optimized Image

Processed Image

•The processed image data is passed through a nonlinear window

Histogram of Data 700

600

500

400

300

200

100

0 0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

Pixel Value

25

How Do the Pieces Fit Together? Optimized Image

Processed Image

• The processed image data is passed through a nonlinear window • The non-linearly windowed data is applied to a gammacorrected (perceptually linearized) display system

Display Luminance as a Function of Pixel Value

D isplay Lum inance (cd/sqm )

1000

100

10

1 0

500

1000

1500

2000

2500

3000

3500

4000

Pixel Value

26

How Do the Pieces Fit Together? Optimized Image

Processed Image

• The processed image data is passed through a nonlinear window • The windowed data is applied to a gamma-corrected (perceptually linearized) display system • The result is a composite function that: • is optimized for viewing • can be “log luminance linear” (i.e. look like film)

Composite Transformation (Display Gamma/Window LUT/GCT)

D isplay Lum inance (cd/sqm )

1000

100

10

1 0

500

1000

1500

2000

2500

3000

3500

4000

Pixel Value

27

Processed Image

Optimized Image

Breast Image Comparison

Processed Image

Optimization for Display

Processed Image

Image Acquisition

Processed Data

Non-Linear Window Non-Linear Windowing LUT

700 70000

600 60000

500

+

300

200

100

S ca lin g F a ct o r

400

50000

40000

30000

20000

10000

0 0

0 0

500

1000

1500

2000

2500

3000

3500

4000

4500

500

1000

1500

2000

2500

3000

3500

4000

4500

Pixel Value

5000

Pixel Value

Processing

+

Processing D is p la y L u m in a n c e ( c d /s q m )

1000

100

10

1 0

500

1000

1500

2000

2500

3000

3500

4000

Digital Driving Level

Perceptually Linearized Display

Raw Image

Monitor Calibration Image Presentation Gamma Correction LUT

Display Gamma 600

Composite Transformation (Display Gamma/Window LUT/GCT)

70000

1000 60000

+

300

200

100

0

D is p la y L u m in a n c e (c d /s q m )

50000

400 C or r e c tion Fa c tor

D is p la y L u m in a n c e (c d /s q m )

500

40000

30000

20000

100

10

10000

0

0

500

1000

1500

2000 Digital Driving Level

2500

3000

3500

4000

0

200

400

600

800

1000

1200

Digital Driving Level

1 0

500

1000

1500

2000

2500

3000

3500

4000

Pixel Value

28

Flat Panel Calibration • “Out of the box” Vendor Supplied LUT • DICOM Part 14 calibration using 256 sampling points – Get less than 256 display values – Have Zero JND Shifts between some DLs

• Optimized DICOM Part 14 calibration using “pseudo shades of gray” – Get TRUE 256 display values – No Zero JND shifts

Linear LUT – “out of the box”

29

Linear LUT – showing zero JND shifts Red < 1 jnd Blue > 3 jnd

Video LUT with 256 Calibration values

30

256 Cal values - Showing JND Shifts

Optigrayscale – TRUE 256 JNDs

31

Optigrayscale 256 JNDsAnnotated

Differing #’s of Calibration Points Linear

“256” cal values True 256 cal values

32

Evaluation of SMPTE Pattern SMPTE – Society of Motion Picture and Television Engineers Pattern developed to evaluate Television systems and Cinema cameras

SMPTE Pattern – Evaluation Frequency Physician - Daily prior to each reading session Technologist - Daily on acquisition workstation - Weekly on Dx monitors in work area

Medical Physicist - Monthly – Department QC review - Annually – detailed eval/comparison with prior years data

33

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%)

34

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners

35

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center Corners

36

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center Corners Black bar on White

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center Corners Black bar on White White bar on Black

37

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center Corners Black bar on White White bar on Black

SMPTE Test Pattern Evaluation 0 / 5% Black 95/100% White Density Step Ramp (0 to 100%) High Contrast Bars Center Corners Low Contrast Bars Center Corners Black bar on White White bar on Black Letters

38

SMPTE found on some Workstations Center Targets For measuring Lmax and Lmin Veiling Glare will Increase black level measurement Pattern Should NOT be used for this measurement

AAPM TG18 • To publish: – Assessment of Display Performance for Medical Imaging Systems • Pre-print draft (ver 9.0) issued 09 Oct 02 • Available at: http://deckard.mc.duke.edu/~samei/tg18

• Document contains descriptions of: – Assessment Instruments – Test Patterns – Software

39

Assessment of Display Performance • • • •

Geometric Distortions Reflection Luminance Response Luminance Uniformity • Resolution • Noise

• • • • •

Veiling Glare Color Uniformity Artifacts Moiré patterns Physical defects

Excellent Presentation on Monday -- Display Evaluation Workshop

AAPM TG18 QC Pattern • Multiple inserts embeded in a midvalue background – – – – – – – –

Gridlines Luminance Patches Line Pair Patterns Cx patches 100, 75, 50, 25% of Lmax Contrast Detail “Quality Control” Vertical Bars White and Black Bars Horizontal Area for Flat Panel evaluation – Border around outside

40

Thank you

Questions? Image Target Patterns GSDF Measure Software Available at: Image-Smiths.com Or E-mail me: [email protected]

41