Visualization-based Brain Mapping Jim Brinkley (PI) Structural Informatics Group Dept Biological Structure University of Washington, Seattle Funded by the Human Brain Project NIDCD NIMH
Goals • An information framework for brain mapping • Organized around structure – Spatial – Symbolic • Implemented in a distributed, webaccessible system • Optimized for language mapping
Photo
Visual Brain Mapper (VBM)
Requirements for visualizationbased brain mapping • Accurate reconstruction of brain surface as seen at neurosurgery • Different problem than surface reconstruction for unfolding, flattening, analysis
Reconstructing the surface • Isosurface following • Needs cleanly segmented cortex: the skullstripping problem
Approaches to skull-stripping • Manual • Deformable models • 3 -D region growing/mathematical morphology
A shape-based approach • Represent “knowledge” of expected shape and range of variation of cortical envelope • Use to guide the search for gray-CSF boundary • Infer the boundary where none is present • Allow user corrections
Representing shape • Geometric constraint networks • Set of local constraints • Interact together to generate a global shape description
Constraint Network Vi (Vi 1, Vi 2, …) Cij Vj (Vj 1, Vj 2, …)
Geometric Constraint Network Vi (Vi 1, Vi 2, …) Cij Vj (Vj 1, Vj 2, …)
![Radial contour model Structural component Ri [Ri, inner … Ri, outer ] Cij Rj](https://present5.com/presentation/8154cd96bdd599daa97005e3e30a323b/image-12.jpg "Radial contour model Structural component Ri [Ri, inner … Ri, outer ] Cij Rj")
Radial contour model Structural component Ri [Ri, inner … Ri, outer ] Cij Rj [Rj, inner, Rj, outer ]
![Radial contour model Shape component Ri Cij = [Ri/Rjmin … Ri/Rjmax] Rj](https://present5.com/presentation/8154cd96bdd599daa97005e3e30a323b/image-13.jpg "Radial contour model Shape component Ri Cij = [Ri/Rjmin … Ri/Rjmax] Rj")
Radial contour model Shape component Ri Cij = [Ri/Rjmin … Ri/Rjmax] Rj
Constraint Propagation
Radial surface model
Radial surface model neighbors
Shape-based skull stripping • • A module of VBM Load previously trained shape model Load MRI, MRV, MRA volumes Find cortical envelope Mask skull Find cortical surface Find veins and arteries
Mid-sagittal plane
Anterior Commissure
Posterior commissure
Cortical bounding box
Initial model
Initial model - end-on
Initial model - lateral view
Propagation from poles
Define one radial
Propagation from one radial
Error
Correct error
Generate mask
Generate mask
Error in mask
Generate mask
Generate isosurface
Extract veins
Mask veins
Vein isosurface
Extract arteries
Mask arteries
Artery isosurface
Vessels
All
Map
Workflow Select Model Align Find Envelope Define Landmarks Segment Cortex Mask Volume Propagate & Search Segment Veins Iso. Surface Select Patient Segment Arteries Map
Select Patient Align Segment Cortex Segment Veins Segment Arteries Map Workflow Architecture
End User Web Interfaces Authoring Programs Symbolic Model Builder Visual Brain Mapper Brain Map Visualizer Experiment Manager Analyzer Internet Knowledge Server Repository Manager Graphics Server Structural Information Servers Symbolic Knowledge Metadata Symbolic Information Images Spatial Knowledge 3 -D Models Spatial Information 3 -D Maps
Contributors • • Kevin Hinshaw Jeff Prothero Andrew Poliakov Richard Martin David Corina Ken Maravilla George Ojemann • • • Rex Jakobovits Kate Mulligan Hector Lettich Linda Shapiro Cornelius Rosse
Скачать презентацию Visualization-based Brain Mapping Jim Brinkley PI Structural Informatics
8154cd96bdd599daa97005e3e30a323b.ppt