Transcript 9Oct_1510_Tarbox-CPI.ppt

DICOM INTERNATIONAL
CONFERENCE & SEMINAR
Oct 9-11, 2010
Rio de Janeiro, Brazil
Managing Display
Quality &
Consistency
Lawrence Tarbox, Ph.D.
Washington University in St. Louis
Overview
• Review of Grayscale Standard Display
Function
• Review of Grayscale Presentation State
• Color Presentation States
– Color Consistency
– Presentation States applied to Color Images
– Color Blending - CT-PET fusion
• Hanging Protocols
Problems of Inconsistency
•VOI chosen on one
display device
•Rendered on another
with different display
•Mass expected to be
seen is no longer seen
mass visible
mass invisible
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Problems of Inconsistency
0.5
1.0
•Not all display levels
are perceivable on all
devices
1.5
3.0
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Problems of Inconsistency
Digital Modality
Printed images does not look
like displayed images
Laser Printer
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Grayscale Standard Display
Function
• Defines a standardize output ‘unit’ for
monochromatic DICOM images called
‘P-Values’ (Presentation Values)
• Based on a model of human vision
described by Barton, et al
• A mathematical derivation that takes
into account lighting conditions
• Follows a curve that is roughly
perceptually linear
Grayscale Standard
Display Function
Grayscale Standard Display Function
1000
Despite different change
in absolute luminance
100
10
1
0
200
400
600
800
1000
.1
.01
JND Index
Same number of Just Noticeable Difference == Same perceived contrast
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Calibrate the Monitor!
• First set up monitor/graphics subsystem as
optimally as possible
• Measure the ambient light
• Measure the current response of the monitor
• Calculate a calibration LUT that converts Pvalues into digital driving levels for the
monitor
• Most high end medical imaging monitors can
auto-calibrate
Display Calibration Tools
(Photometer)
Slide Provided by Jerry Gaskill, Image Smiths Inc.
Monitor Characteristic Curve
Monitor Characteristic Curve
Luminance
100
10
0.1
Ambient Light
0.01
0
50
100
150
Digital Driving Level
200
250
300
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Perceptual linear device - LUT
Mapping P-Values to Input of Characteristic Curve (DDL’s)
300
250
DDL
200
150
100
50
0
0
50
100
150
200
250
300
P-Values
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Grayscale Standard
Display Function
Grayscale Standard Display Function
1000
Despite different change
in absolute luminance
100
10
1
0
200
400
600
800
1000
.1
.01
JND Index
Same number of Just Noticeable Difference == Same perceived contrast
Slide Provided by David Clunie, Quintiles Intelligent Imaging
Distributed Image
Consistency
Laser Printer
Digital Modality
Identical perceived contrast
Workstation
Workstation
DICOM Grayscale
Presentation State
• Separate object from the image objects
• Specifies how the referenced image is
to be displayed
– Grayscale transformation
– Graphics and text overlays
– Spatial transformations
DICOM Grayscale Image
Transformation Model
Rescale Slope/Intercept
or Modality LUT
Original
Image
Modality
LUT
Transformation
Window/Level
or VOI LUT
Mask
(Subtraction)
VOI LUT
Transformation
Presentation
LUT
Presentation
LUT
Transformation
Grayscale Transformations
Shutter
P-Values
Transformation
Image
Annotation
Spatial
Transformation
Disp. Area
Display
Annotation
Shutter, Annotation and Spatial Transformations
Spatial Transformations
Entire Image Selected
Flip Horizontal
Original Image
Scale To Fit
Transformed Image
Spatial Transformations
Part of Image Selected
Flip Horizontal
Original Image
Scale To Fit
Transformed Image
Transformation & Annotation
Part of Image Selected
Flip Horizontal
Scale To Fit
Mass behind heart
Mass behind heart
Original Image
Transformed Image
In this example,
- text annotation is specified by image relative visible anchor point
- the circle is a separate image relative graphic annotation
Limitations of Grayscale
Presentation States
• Apply to grayscale images
– no means to specify spatial transformations or
graphic annotations for color images
• Only grayscale consistency
– standard display function defined only for
luminance
• No pseudo-color capability
• No blending or fusion capability
Distributed Color Image
Consistency
Printer
Digital Modality
Different perceived color
Workstation
Workstation
Distributed Image
Consistency
Printer
Digital Modality
Identical perceived color
Workstation
Workstation
True and Pseudo-Color
Goals for Color
• Color consistency
– standard function
– defined for image output space of existing
color images
• Transformation and annotation pipeline
• Pseudo-color for grayscale images
• Blending of grayscale images
– alpha blending function
– colorizing superimposed image
Standard Color Space
•
•
•
•
•
•
•
GSDF filled a void
Color consistency already standardized
ICC - International Color Consortium
Graphics and pre-press industry
CIE Colorimetry
Profiles of input and output devices
Commercial Off-The-Shelf color management
software handles conversion
• Perceptual rendering intent
Three New SOP Classes
• Color Presentation State
• Pseudo-Color Presentation State
• Blending Presentation State
• ICC Profile
– Defines output of all color presentation states
– Optionally present in all color images
• PCS-Values (analogous to grayscale PValues)
– Profile Connection Space (CIELAB or CIEXYZ)
Commonality
• All presentation states share identical
– Spatial transformation pipeline
– Graphic and text annotation pipeline
• Choice of output space
– P-Values for grayscale
– PCS-Values for color and pseudo-color and
blending
Old Grayscale Pipeline
Rescale Slope/Intercept
Grayscale
Stored Pixel
Values
Modality
LUT
Transformation
Mask
Subtraction
Window
or VOI LUT
Presentation
LUT
VOI LUT
Transformation
Presentation
LUT
Transformation
Device
Independent
Values
P-Values
Grayscale & Color Pipeline
Rescale Slope/Intercept
Grayscale
Stored Pixel
Values
Modality
LUT
Transformation
Mask
Subtraction
Window
or VOI LUT
Presentation
LUT
VOI LUT
Transformation
Presentation
LUT
Transformation
Device
Independent
Values
P-Values
Pseudo
Color
Palette
Color LUT
Transformation
Profile
Connection
Space
Transformation
True Color
Stored Pixel
Values
Indexed Color
Stored Pixel
Values
ICC Input
Profile
Palette
Color LUT
Transformation
PCS-Values
Common Spatial &
Annotation Pipeline
Device
Independent
Values
Shutter
Transformation
Im age
Relative
Annotation
Spatial
Transformation
Displayed
Area Relative
Annotation
Display
Blending Pipeline
Underlying
Im age
Grayscale
Stored Pixel
Values
Rescale Slope/Intercept
Window
or VOI LUT
Modality
LUT
Transformation
VOI LUT
Transform ation
Device
Independent
Values
Pseudo
Color
Superim posed
Im age
Grayscale
Stored Pixel
Values
Modality
LUT
Transformation
VOI LUT
Transformation
Palette
Color LUT
Transformation
Relative
Opacity
ICC Input
Profile
Blending
Operation
Profile
Connection
Space
Transformation
PCS-Values
Blending for CT-PET
select
underlying
select
superimposed
Blending for CT-PET
select
underlying
select
superimposed
[register]
Blending for CT-PET
select
underlying
select
superimposed
[register]
resample
Blending for CT-PET
select
underlying
select
superimposed
[register]
within slices
resample
Blending for CT-PET
select
underlying
select
superimposed
[register]
within slices
resample
[between slices]
Blending for CT-PET
select
underlying
select
superimposed
rescale and
window
[register]
within slices
resample
[between slices]
Blending for CT-PET
select
underlying
select
superimposed
rescale and
window
[register]
within slices
resample
[between slices]
pseudo-color
Blending for CT-PET
select
underlying
select
superimposed
rescale and
window
[register]
within slices
resample
[between slices]
blend
pseudo-color
Color - Conclusion
• Color consistency using industry
standard
• Transformation/annotation for color
images
• Exchange of pseudo-color information
• Support for specifying sets of images
to be blended, and how to blend (but
not register or resample) them
Hanging Protocols
•
•
•
•
“Default display protocols”
A set of instructions
How to layout a class of images for display
Order, orientation, windowing, processing
• Not specific to a particular patient’s images
• Hence a protocol, not a presentation state
Hanging Protocols
Hanging Protocols
L
L
Old
Lateral
F
Old Study
L
New
Frontal
New
Lateral
F
F
New Study
L
New
Townes
F
Hanging Protocol Goals
• Encode
– Applicability of protocol (type of display & images)
– Selection of images
– Display of selected images
• Store centrally, retrieve and exchange
– Persistent composite objects
– Query, retrieval and media encoding
• Vendor neutrality
– Interchange between sites, PACS and workstations
– Survive upgrades and replacements
– “Public” library of “good” hanging protocols ?
Using a Hanging Protocol
• Given a current exam (e.g. reading
worklist)
• Find potentially applicable protocols
• Retrieve them from archive
• Select one from those available
• Select image +/- other studies to which
it applies
• Display selected images as instructed
Finding a Protocol
• Definition Module
– Name, description, level, creator, creation
datetime
– Modality, anatomy, laterality
– Procedure, reason for procedure
– Number of priors
• Environment Module
– Number of screens
– Size(s) of screens
– Color or grayscale bit depth
Selecting Images
• Definition of “image sets”
• By attribute values
– Specific attributes, e.g. Modality, Anatomy
– Specific values, e.g, CT, Chest
– Supports all VRs, coded sequences, private elements
and multi-frame functional groups
• By time
– Relative time (today, yesterday, within last week)
– Abstract priors (last, oldest, pre-operative, etc.)
Successful Selection
• All hanging protocols depend on consistent
and reliable (and standard) information being
present in the images
• DICOM Hanging Protocols don’t solve this
integration problem
• Ideally - modality inserts correct anatomy
and procedure and reason and orientation
codes, and uses standard technique
descriptions
• Worst case (typically?) - modality protocol
(or operator) inserts recognizable Series
Description
Information for Hanging
Anterior
L
Foot
Right
Modality: Mammography
Anatomic Region: Breast
Image Laterality: L
View Code: Medio-Lateral Oblique
Patient Orientation: A\FR
Priors
• Concept of the “current” study required
• Protocol chooses priors based on
– Relative time
– Abstract temporal ranges (previous, last, etc.)
– Abstract coded descriptions (“pre-operative”)
• Does NOT specify how to find them or get
them
• May have been pushed, may need a query
• May be hard to find by abstract descriptions
• Creative use of queries or out-of-band
information
Mapping to Image Boxes
• Image Sets are mapped to Image Boxes
• Image Box types
–
–
–
–
–
Tiled (e.g. 3x4)
Stack (single image paged manually)
Cine (time-based play back)
Processed (e.g. MPR, 3D)
Single (e.g. a place for a report or waveform)
• Specify
– Scrolling mode
– Playback rate
Mapping to Image Boxes
• Filtering
– By attribute, or abstract, e.g. “category” of “image plane” “axial”
• Sorting
– By attribute, or abstract, e.g. “along axis” “increasing”
• Orientation
– E.g. rotate/flip until row left column posterior (L\P)
• Annotation
– Patient demographics, technique and graphics on or off
Processing & Presentation
•
•
•
•
•
•
•
•
•
•
Reformatting, e.g., MPR, 3D, slab
Thickness, interval
View direction, e.g., axial, sagittal, coronal
Type, e.g., MIP, surface, volume
VOI Type (windowing), e.g., brain, bone
Pseudo-color type, e.g., hot iron
Invert grayscale
True size
Synchronized scrolling (by Display Set number)
Navigation and localization
Display of Image Boxes
• Entire display environment from 0,0 to
1,1
• Individual screens are not
distinguished
(1.0, 1.0)
(0.33, 1.0)
2560
pixels
(0.0, 0.4)
1024
pixels
2048
(0.33, 0.0)
(0.0, 0.0)
1024
pixels
pixels
(1.0, 0.0)
Display of Image Boxes
• Image Sets displayed in Image Boxes
• Image Boxes rendered at relative
location
(1.0, 1.0)
DS 10:
AXIAL
2560
1024
1024
DS 7:
MPR
SAG
DS 8:
AXIA
L
DS 6:
MPR
COR
DS 9:
3D
VOL
(0.0, 0.0)
2048
Hanging Protocols Conclusion
•
•
•
•
•
•
•
Interchangeable
Vendor neutral
Multi-modality
Support selection of priors
Full richness of current display modes
Flexible
Extensible
• Non-trivial to implement and retrofit
• Dependent on reliable image attributes