midag@unc medial object shape representations for image analysis & object synthesis stephen m....
Post on 20-Dec-2015
225 views
TRANSCRIPT
MIDAG@UNCMIDAG@UNC
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis
Stephen M. PizerStephen M. Pizer
Kenan ProfessorKenan Professor
Medical Image Display & Analysis GroupMedical Image Display & Analysis Group
University of North Carolina, USAUniversity of North Carolina, USA
Credits: Many on MIDAG, especiallyCredits: Many on MIDAG, especially
Daniel Fritsch, Andrew Thall, George Stetten, Daniel Fritsch, Andrew Thall, George Stetten,
Paul YushkevichPaul Yushkevich
Stephen M. PizerStephen M. Pizer
Kenan ProfessorKenan Professor
Medical Image Display & Analysis GroupMedical Image Display & Analysis Group
University of North Carolina, USAUniversity of North Carolina, USA
Credits: Many on MIDAG, especiallyCredits: Many on MIDAG, especially
Daniel Fritsch, Andrew Thall, George Stetten, Daniel Fritsch, Andrew Thall, George Stetten,
Paul YushkevichPaul Yushkevich
MIDAG@UNCMIDAG@UNC
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis
MIDAG@UNCMIDAG@UNC
WhatWhat shape shape representation is forrepresentation is for
WhatWhat shape shape representation is forrepresentation is for
Analysis from imagesAnalysis from images Extract the kidney-shaped objectExtract the kidney-shaped object Register based on the pelvic bone shapesRegister based on the pelvic bone shapes Extract shape information w/o modelExtract shape information w/o model
SynthesisSynthesis Design the objectDesign the object Deform the object, with physical realismDeform the object, with physical realism
Shape science Shape science Shape and biologyShape and biology Shape-based diagnosisShape-based diagnosis
Analysis from imagesAnalysis from images Extract the kidney-shaped objectExtract the kidney-shaped object Register based on the pelvic bone shapesRegister based on the pelvic bone shapes Extract shape information w/o modelExtract shape information w/o model
SynthesisSynthesis Design the objectDesign the object Deform the object, with physical realismDeform the object, with physical realism
Shape science Shape science Shape and biologyShape and biology Shape-based diagnosisShape-based diagnosis
MIDAG@UNCMIDAG@UNC
WhatWhat shape shape representation is forrepresentation is for
WhatWhat shape shape representation is forrepresentation is for
Analysis from imagesAnalysis from images Extract the kidney-shaped objectExtract the kidney-shaped object
Register based on the pelvic bone shapesRegister based on the pelvic bone shapes
Analysis from imagesAnalysis from images Extract the kidney-shaped objectExtract the kidney-shaped object
Register based on the pelvic bone shapesRegister based on the pelvic bone shapes
MIDAG@UNCMIDAG@UNC
WhatWhat shape shape representation is forrepresentation is for
WhatWhat shape shape representation is forrepresentation is for
SynthesisSynthesis Design the objectDesign the object Deform the object, with physical realismDeform the object, with physical realism
SynthesisSynthesis Design the objectDesign the object Deform the object, with physical realismDeform the object, with physical realism
MIDAG@UNCMIDAG@UNC
WhatWhat shape shape representation is forrepresentation is for
WhatWhat shape shape representation is forrepresentation is for
Shape science Shape science Shape and biologyShape and biology Shape-based diagnosisShape-based diagnosis
Shape science Shape science Shape and biologyShape and biology Shape-based diagnosisShape-based diagnosis
Brain structures (Gerig)
MIDAG@UNCMIDAG@UNC
Shape SciencesShape SciencesShape SciencesShape Sciences
GeometryGeometry The spatial layout: via primitivesThe spatial layout: via primitives
LandmarksLandmarks Boundary places and orientationsBoundary places and orientations Medial places, figural sizes and orientationsMedial places, figural sizes and orientations Space itselfSpace itself
StatisticsStatistics The average shapeThe average shape Modes of variation from the averageModes of variation from the average
Computer GraphicsComputer Graphics Image AnalysisImage Analysis
GeometryGeometry The spatial layout: via primitivesThe spatial layout: via primitives
LandmarksLandmarks Boundary places and orientationsBoundary places and orientations Medial places, figural sizes and orientationsMedial places, figural sizes and orientations Space itselfSpace itself
StatisticsStatistics The average shapeThe average shape Modes of variation from the averageModes of variation from the average
Computer GraphicsComputer Graphics Image AnalysisImage Analysis
MIDAG@UNCMIDAG@UNC
Options for Primitives Options for Primitives Options for Primitives Options for Primitives
Space: Space: xxii for grid elements for grid elements
Landmarks: Landmarks: xxii described by local geometry described by local geometry
Boundary: (Boundary: (xxii ,normal ,normalii) spaced along boundary) spaced along boundary
Figural: nets of diatoms sampling figuresFigural: nets of diatoms sampling figures
Space: Space: xxii for grid elements for grid elements
Landmarks: Landmarks: xxii described by local geometry described by local geometry
Boundary: (Boundary: (xxii ,normal ,normalii) spaced along boundary) spaced along boundary
Figural: nets of diatoms sampling figuresFigural: nets of diatoms sampling figures
MIDAG@UNCMIDAG@UNC
Primitives for shape Primitives for shape representation: Landmarksrepresentation: Landmarks
Primitives for shape Primitives for shape representation: Landmarksrepresentation: Landmarks
Sets of points of special Sets of points of special geometrygeometry
Sets of points of special Sets of points of special geometrygeometry
MIDAG@UNCMIDAG@UNC
Primitives for shape Primitives for shape representation: Boundariesrepresentation: Boundaries
Primitives for shape Primitives for shape representation: Boundariesrepresentation: Boundaries
Boundary points with normalsBoundary points with normals Boundary points with normalsBoundary points with normals
MIDAG@UNCMIDAG@UNC
Object Representation by M-RepsObject Representation by M-Reps
MIDAG@UNCMIDAG@UNC
Each M-figure Represented by Each M-figure Represented by Net of Medial PrimitivesNet of Medial Primitives
o
o
ooo
o
o
o
oo
MIDAG@UNCMIDAG@UNC
Each M-figure Represented by Each M-figure Represented by Net of Medial PrimitivesNet of Medial Primitives
MIDAG@UNCMIDAG@UNC
Figural ModelsFigural ModelsFigural ModelsFigural Models
Figures: successive medial involutionFigures: successive medial involution Main figureMain figure ProtrusionsProtrusions IndentationsIndentations Separate figuresSeparate figures
Hierarchy of figuresHierarchy of figures Relative positionRelative position Relative widthRelative width Relative orientationRelative orientation
Figures: successive medial involutionFigures: successive medial involution Main figureMain figure ProtrusionsProtrusions IndentationsIndentations Separate figuresSeparate figures
Hierarchy of figuresHierarchy of figures Relative positionRelative position Relative widthRelative width Relative orientationRelative orientation
o
o
ooo
o
o
o
oo
MIDAG@UNCMIDAG@UNC
Primitives’ Desired Primitives’ Desired PropertiesProperties
Primitives’ Desired Primitives’ Desired PropertiesProperties
GeometryGeometry Intuitive: simple, global + localIntuitive: simple, global + local Efficiently deformableEfficiently deformable Easily extracted or createdEasily extracted or created Spatial tolerance inherentSpatial tolerance inherent
StatisticsStatistics Unimodality: normally distributedUnimodality: normally distributed Via geometrical, tolerance-sensitive metricVia geometrical, tolerance-sensitive metric
GeometryGeometry Intuitive: simple, global + localIntuitive: simple, global + local Efficiently deformableEfficiently deformable Easily extracted or createdEasily extracted or created Spatial tolerance inherentSpatial tolerance inherent
StatisticsStatistics Unimodality: normally distributedUnimodality: normally distributed Via geometrical, tolerance-sensitive metricVia geometrical, tolerance-sensitive metric
MIDAG@UNCMIDAG@UNC
Figural Models Figural Models with Boundary Deviationswith Boundary Deviations
Figural Models Figural Models with Boundary Deviationswith Boundary Deviations
HypothesisHypothesisAt a global level, a figural At a global level, a figural
model is the most intuitivemodel is the most intuitive
At a local level, boundary At a local level, boundary deviations are most intuitivedeviations are most intuitive
HypothesisHypothesisAt a global level, a figural At a global level, a figural
model is the most intuitivemodel is the most intuitive
At a local level, boundary At a local level, boundary deviations are most intuitivedeviations are most intuitive
o
o
ooo
o
o
o
oo
MIDAG@UNCMIDAG@UNC
Union and Difference of M-figuresUnion and Difference of M-figures
MIDAG@UNCMIDAG@UNC
Medial PrimitivesMedial PrimitivesMedial PrimitivesMedial Primitives
xx, (, (bb,,nn) frame, r, ) frame, r, (object angle) (object angle) Imply boundary segmentsImply boundary segments
with tolerancewith tolerance
Similarity transform equivariantSimilarity transform equivariant Zoom invariance implies width-proportionality ofZoom invariance implies width-proportionality of
tolerance of implied boundarytolerance of implied boundary boundary curvature distributionboundary curvature distribution spacing along netspacing along net interrogation aperture for imageinterrogation aperture for image
xx, (, (bb,,nn) frame, r, ) frame, r, (object angle) (object angle) Imply boundary segmentsImply boundary segments
with tolerancewith tolerance
Similarity transform equivariantSimilarity transform equivariant Zoom invariance implies width-proportionality ofZoom invariance implies width-proportionality of
tolerance of implied boundarytolerance of implied boundary boundary curvature distributionboundary curvature distribution spacing along netspacing along net interrogation aperture for imageinterrogation aperture for image
n
MIDAG@UNCMIDAG@UNC
3D kidney model extracted from CT3D kidney model extracted from CT3D kidney model extracted from CT3D kidney model extracted from CT
Four figure model of the kidneysFour figure model of the kidneys
Red represents indentation figuresRed represents indentation figures
Four figure model of the kidneysFour figure model of the kidneys
Red represents indentation figuresRed represents indentation figures
MIDAG@UNCMIDAG@UNC
Need for Special End PrimitivesNeed for Special End Primitives
Represent Represent non-blobby objectsnon-blobby objects
angulated edges, corners, creasesangulated edges, corners, creases still allow rounded edges , corners, creasesstill allow rounded edges , corners, creases allow bent edgesallow bent edges
ButBut Avoid infinitely fine medial samplingAvoid infinitely fine medial sampling Maintain tangency, symmetry principlesMaintain tangency, symmetry principles
MIDAG@UNCMIDAG@UNC
End PrimitivesEnd Primitives
Rounded end primitive Rounded end primitive in cross-sectionin cross-section
Corner primitive Corner primitive in cross-sectionin cross-section
MIDAG@UNCMIDAG@UNC
Displacements from Displacements from Figurally Implied BoundaryFigurally Implied Boundary
Displacements from Displacements from Figurally Implied BoundaryFigurally Implied Boundary
Boundary implied by figural model Boundary after displacements
MIDAG@UNCMIDAG@UNC
Coarse-to-fine representationCoarse-to-fine representationCoarse-to-fine representationCoarse-to-fine representation
For each of three levelsFigural hierarchyFor each figure,
net chain, successively smaller tolerance
For each net tile, boundary displacement
chain
For each of three levelsFigural hierarchyFor each figure,
net chain, successively smaller tolerance
For each net tile, boundary displacement
chain
MIDAG@UNCMIDAG@UNC
Multiscale Medial ModelMultiscale Medial ModelMultiscale Medial ModelMultiscale Medial Model
From larger scale medial net Coarsely sampled Smooother figurally implied boundary Larger tolerance
Interpolate smaller scale medial net Finer sampled More detail in figurally implied boundary Smaller tolerance
Represent medial displacements
From larger scale medial net Coarsely sampled Smooother figurally implied boundary Larger tolerance
Interpolate smaller scale medial net Finer sampled More detail in figurally implied boundary Smaller tolerance
Represent medial displacements
MIDAG@UNCMIDAG@UNC
Multiscale Medial ModelMultiscale Medial ModelMultiscale Medial ModelMultiscale Medial Model
From larger scale medial net, interpolate smaller scale medial net
and represent medial displacements
From larger scale medial net, interpolate smaller scale medial net
and represent medial displacementsb.
MIDAG@UNCMIDAG@UNC
Multiscale Medial/Boundary ModelMultiscale Medial/Boundary ModelMultiscale Medial/Boundary ModelMultiscale Medial/Boundary Model
From medial net Coarsely sampled, smoother implied
boundary Larger tolerance
Represent boundary displacements along implied normals Finer sampled, more detail in boundary Smaller tolerance
From medial net Coarsely sampled, smoother implied
boundary Larger tolerance
Represent boundary displacements along implied normals Finer sampled, more detail in boundary Smaller tolerance
MIDAG@UNCMIDAG@UNC
Shape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image Analysis
SegmentationSegmentation Extract an object from imageExtract an object from image
RegistrationRegistrationFind geometric transformation that brings Find geometric transformation that brings
two images into alignmenttwo images into alignment 3D/3D3D/3D 3D/2D3D/2D
Shape MeasurementShape MeasurementFind how probable a shape isFind how probable a shape is
SegmentationSegmentation Extract an object from imageExtract an object from image
RegistrationRegistrationFind geometric transformation that brings Find geometric transformation that brings
two images into alignmenttwo images into alignment 3D/3D3D/3D 3D/2D3D/2D
Shape MeasurementShape MeasurementFind how probable a shape isFind how probable a shape is
MIDAG@UNCMIDAG@UNC
Shape Repres’n in Image AnalysisShape Repres’n in Image AnalysisShape Repres’n in Image AnalysisShape Repres’n in Image Analysis
SegmentationSegmentation
Find the most probable deformed mean Find the most probable deformed mean model, given the imagemodel, given the image
Probability involvesProbability involvesProbability of the deformed model (prior)Probability of the deformed model (prior)Probability of the image, given the Probability of the image, given the
deformed model (likelihood)deformed model (likelihood)
SegmentationSegmentation
Find the most probable deformed mean Find the most probable deformed mean model, given the imagemodel, given the image
Probability involvesProbability involvesProbability of the deformed model (prior)Probability of the deformed model (prior)Probability of the image, given the Probability of the image, given the
deformed model (likelihood)deformed model (likelihood)
MIDAG@UNCMIDAG@UNC
Probability of a deformed modelProbability of a deformed modelProbability of a deformed modelProbability of a deformed model
From training setFrom training set via principal components analysis, coarse-to-finevia principal components analysis, coarse-to-fine
-C * Geometric difference from typical shape-C * Geometric difference from typical shape
From training setFrom training set via principal components analysis, coarse-to-finevia principal components analysis, coarse-to-fine
-C * Geometric difference from typical shape-C * Geometric difference from typical shape
MIDAG@UNCMIDAG@UNC
Medialness: medial strength of Medialness: medial strength of a medial primitive in an imagea medial primitive in an imageMedialness: medial strength of Medialness: medial strength of a medial primitive in an imagea medial primitive in an image
Probability of image | deformed modelProbability of image | deformed model Sum of boundariness values Sum of boundariness values
at implied boundary positionsat implied boundary positions in implied normal directionsin implied normal directions with apertures proportional to with apertures proportional to
tolerancetolerance
Boundariness value Boundariness value Intensity profile distance from mean (at scale)Intensity profile distance from mean (at scale)
statistical, based on training setstatistical, based on training set Intensity differencesIntensity differences
via Gaussian derivativesvia Gaussian derivatives
Probability of image | deformed modelProbability of image | deformed model Sum of boundariness values Sum of boundariness values
at implied boundary positionsat implied boundary positions in implied normal directionsin implied normal directions with apertures proportional to with apertures proportional to
tolerancetolerance
Boundariness value Boundariness value Intensity profile distance from mean (at scale)Intensity profile distance from mean (at scale)
statistical, based on training setstatistical, based on training set Intensity differencesIntensity differences
via Gaussian derivativesvia Gaussian derivatives
b
rR(- )bx
rR()b
MIDAG@UNCMIDAG@UNC
Figurally implied boundaries Figurally implied boundaries and rendering, via 4-figure modeland rendering, via 4-figure model
Figurally implied boundaries Figurally implied boundaries and rendering, via 4-figure modeland rendering, via 4-figure model
MIDAG@UNCMIDAG@UNC
3D DSL Model Deformation3D DSL Model Deformation3D DSL Model Deformation3D DSL Model Deformation
Initial Position of Model in Target ImageInitial Position of Model in Target Image
MIDAG@UNCMIDAG@UNC
3D DSL Model Deformation3D DSL Model Deformation3D DSL Model Deformation3D DSL Model Deformation
Figural DeformationFigural DeformationIteration 3Iteration 3
MIDAG@UNCMIDAG@UNC
3D DSL Model Deformation3D DSL Model Deformationwith interfigural penaltieswith interfigural penalties
3D DSL Model Deformation3D DSL Model Deformationwith interfigural penaltieswith interfigural penalties
Initial positionInitial position After optimizationAfter optimization
MIDAG@UNCMIDAG@UNC
Shape Repres’n in Image AnalysisShape Repres’n in Image AnalysisShape Repres’n in Image AnalysisShape Repres’n in Image Analysis
RegistrationRegistration
Find the most probable deformation, Find the most probable deformation, given the imagegiven the image
RegistrationRegistration
Find the most probable deformation, Find the most probable deformation, given the imagegiven the image
MIDAG@UNCMIDAG@UNC
Shape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image Analysis
Prior-free medial shape analysisPrior-free medial shape analysis
Cores: height ridges of medialness (Pizer, Cores: height ridges of medialness (Pizer, Fritsch, Morse, Furst)Fritsch, Morse, Furst)
Statistical analysis of medial diatoms Statistical analysis of medial diatoms (Stetten)(Stetten)
Prior-free medial shape analysisPrior-free medial shape analysis
Cores: height ridges of medialness (Pizer, Cores: height ridges of medialness (Pizer, Fritsch, Morse, Furst)Fritsch, Morse, Furst)
Statistical analysis of medial diatoms Statistical analysis of medial diatoms (Stetten)(Stetten)
MIDAG@UNCMIDAG@UNC
Shape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image Analysis
Cores: height ridges of medialnessCores: height ridges of medialness Cores: height ridges of medialnessCores: height ridges of medialness
M I P @ U N CM I P @ U N C
MIDAG@UNCMIDAG@UNC
Shape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image AnalysisShape Rep’n in Image Analysis
Statistical analysis of medial diatomsStatistical analysis of medial diatoms Statistical analysis of medial diatomsStatistical analysis of medial diatoms
a
sphere cylinder slab
MIDAG@UNCMIDAG@UNC
sphere
slab cylinder
MIDAG@UNCMIDAG@UNC
sphere
slab cylinder
MIDAG@UNCMIDAG@UNC
sphere
slab cylinder
MIDAG@UNCMIDAG@UNC
a
myocardium
left ventricle
left atrium
mitral valve
epicardium
slab
cylinder
cap
MIDAG@UNCMIDAG@UNC
sphere
slab cylinder
MIDAG@UNCMIDAG@UNC
sphere
slab cylinder
MIDAG@UNCMIDAG@UNC
MIDAG@UNCMIDAG@UNC
Shape Rep’n in CAD/CAMShape Rep’n in CAD/CAM Shape Rep’n in CAD/CAMShape Rep’n in CAD/CAM
Stock figural modelsStock figural models Deformation tools: large scaleDeformation tools: large scale Coarse-to-fine specificationCoarse-to-fine specification Figural connection toolsFigural connection tools Direct rendering, according to display Direct rendering, according to display
needsneeds
Stock figural modelsStock figural models Deformation tools: large scaleDeformation tools: large scale Coarse-to-fine specificationCoarse-to-fine specification Figural connection toolsFigural connection tools Direct rendering, according to display Direct rendering, according to display
needsneeds
MIDAG@UNCMIDAG@UNC
Deformation in CAD/CAMDeformation in CAD/CAM Deformation in CAD/CAMDeformation in CAD/CAM
MIDAG@UNCMIDAG@UNC
Shape Rep’n in CAD/CAMShape Rep’n in CAD/CAM Shape Rep’n in CAD/CAMShape Rep’n in CAD/CAM
Design models for image analysisDesign models for image analysis Design models for image analysisDesign models for image analysis
MIDAG@UNCMIDAG@UNC
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis
Medial Object Shape Medial Object Shape RepresentationsRepresentations
for Image Analysis & Object Synthesisfor Image Analysis & Object Synthesis Figural models, at successive levels of Figural models, at successive levels of
tolerancetolerance Boundary displacementsBoundary displacements
Work in progressWork in progress Segmentation and registration toolsSegmentation and registration tools Statistical analysis of object populationsStatistical analysis of object populations CAD tools, incl. direct renderingCAD tools, incl. direct rendering Connection relative critical manifoldsConnection relative critical manifolds ……
Figural models, at successive levels of Figural models, at successive levels of tolerancetolerance
Boundary displacementsBoundary displacements
Work in progressWork in progress Segmentation and registration toolsSegmentation and registration tools Statistical analysis of object populationsStatistical analysis of object populations CAD tools, incl. direct renderingCAD tools, incl. direct rendering Connection relative critical manifoldsConnection relative critical manifolds ……
MIDAG@UNCMIDAG@UNC
Application: Image guided Application: Image guided planning & delivery of radiotherapyplanning & delivery of radiotherapy
Application: Image guided Application: Image guided planning & delivery of radiotherapyplanning & delivery of radiotherapy
Planning in 3DPlanning in 3D Extracting normal anatomyExtracting normal anatomy Extracting tumorExtracting tumor Planning beam posesPlanning beam poses
Patient placementPatient placement Verification of plan via Verification of plan via
portal imageportal image
Planning in 3DPlanning in 3D Extracting normal anatomyExtracting normal anatomy Extracting tumorExtracting tumor Planning beam posesPlanning beam poses
Patient placementPatient placement Verification of plan via Verification of plan via
portal imageportal image
MIDAG@UNCMIDAG@UNC
Finding Treatment Pose from Finding Treatment Pose from Portal Radiograph and Planning DRRPortal Radiograph and Planning DRR
Finding Treatment Pose from Finding Treatment Pose from Portal Radiograph and Planning DRRPortal Radiograph and Planning DRR
Planning CT Scan
Patient Setup
Planning DRR
Candidate DRRs Portal Image
Planning pose Candidate poses Treatment pose
MIDAG@UNCMIDAG@UNC
Medial Net Shape ModelsMedial Net Shape ModelsMedial Net Shape ModelsMedial Net Shape Models
Medial nets, positions onlyMedial net
MIDAG@UNCMIDAG@UNC
Integrated Medialness vs. Pose OffsetIntegrated Medialness vs. Pose OffsetIntegrated Medialness vs. Pose OffsetIntegrated Medialness vs. Pose Offset
Image Match Behavior Near Patient Pose Along Three Rotation Axis
-70
-65
-60
-55
-50
-45
-40
-35
-30
-5.00 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00
Distance to truth (deg)
Ima
ge
ma
tch
Gantry
Table
Collimator
Image Match Behavior Near Patient Pose Along Three Translation Axis
-80
-60
-40
-20
0
20
40
60
-5.00 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00
Distance to truth (cm)Im
ag
e m
atc
h
X
Y(zoom)
Z
MIDAG@UNCMIDAG@UNC
MIDAG@UNCMIDAG@UNC
Representing Boundary DisplacementsRepresenting Boundary DisplacementsRepresenting Boundary DisplacementsRepresenting Boundary Displacements
Along figurally implied boundary normalsAlong figurally implied boundary normals Coarse-to-fineCoarse-to-fine Captures along-boundary covarianceCaptures along-boundary covariance
Useful for renderingUseful for rendering
Along figurally implied boundary normalsAlong figurally implied boundary normals Coarse-to-fineCoarse-to-fine Captures along-boundary covarianceCaptures along-boundary covariance
Useful for renderingUseful for rendering
MIDAG@UNCMIDAG@UNC
Summing Medialness on Medial NetSumming Medialness on Medial Netvia Medial Weighting Functionvia Medial Weighting Function
Summing Medialness on Medial NetSumming Medialness on Medial Netvia Medial Weighting Functionvia Medial Weighting Function
MIDAG@UNCMIDAG@UNC
CT Slice of Kidneys in AbdomenCT Slice of Kidneys in AbdomenCT Slice of Kidneys in AbdomenCT Slice of Kidneys in Abdomen
MIDAG@UNCMIDAG@UNC
Object Shape Object Shape
Brain structures (Gerig)Brain structures (Gerig)Brain structures (Gerig)Brain structures (Gerig)
MIDAG@UNCMIDAG@UNC
Geometric aspects : Geometric aspects : TransformationsTransformations
Euclidean: Euclidean: translation and rotationtranslation and rotation
Similarity:Similarity: translation, rotation, zoomtranslation, rotation, zoom
AffineAffine