Combined Volumetric and Surface (CVS) Registration FS workshop Gheorghe Postelnicu, Lilla Zöllei, Bruce Fischl A.A. Martinos Center for Biomedical Imaging, MGH
Surface-based (2D) registration does an excellent job of aligning cortical folds, but doesn’t apply to non-cortical structures (e.g. basal ganglia). Volumetric (3D) registration applies to the entire brain but doesn’t in general align folding patterns. Solution: integrate them Introduction and motivation
Why Aligning Folds in the Volume is Hard! Affine transform of surfaces from one subject mapped to another.
Template Affine Nonlinear pial surfaceWM surface
Outline Proposed Framework: –Automated Surface-based registration –Initialization via an Elastic Transform –Intensity-based Volumetric Registration Experiments Validation
accurate, topologically correct reconstructions of the cortical surfaces (L/R white/pial) registration in the spherical space 1-to-1 mapping Surface-based registration Metric distortion Topology preservation Data term
Reconstruction of cortical surfaces
Surface Morph spherical registration for each of the hemispheres independently Sulcal pattern projected on the sphere TemplateRegistered Source Source
Surface Morph
Elastic morph to diffuse vector field from the cortical surfaces to the rest of the volume regularity constraint: –elastic deformation, i.e. a smooth, orientation- preserving deformation satisfying our choice: Navier operator from the linear elasticity theory + finite element method (FEM)
Elastic Morph Iterate for i=1:n create mesh set up BC’s from surface setup linear elastic system solve system solve topology problems Tetrahedral mesh used for one iteration of the elastic solver. Notice how the mesh is denser near the input surfaces
Example of Morph Dynamics
Results – Elastic Morph Template Affine HAMMER Elastic
Intensity-based registration Nonlinear optical flow intensity based registration [Fischl, NeuroImage04] G I : intensity term G T : topology constraining term G S : smoothness term
Resulting Morph Template Elastic CVS Template
Step 2: Elastic deformation
Step 3: Intensity deformation
Experimental setup Compare: FLIRT, HAMMER, CVS Template: single, randomly selected scan; rest in the data set is registered to it Accuracy: Generalized Jaccard overlap metric
Experiments 1.Adult brain inter-subject [R. Buckner]: –40 subjects (66 cortical+21subcortical manually segmented labels) 2.IBSR dataset, CMA, MGH ( –10 subjects –15 GM/WM structures; mostly sub-cortical –common with Joshi
The last column of the cortical measures represents the overlap measure computed in the surface space. The vertical lines represent the standard error of the mean of the measurement. 2
TEMPLATE
FLIRT
HAMMER
ELASTIC
CVS
TEMPLATE
FLIRT
HAMMER
ELASTIC
CVS
TEMPLATE
3 Overlap measure: JaccardIn figures: mean and standard error
Template 3
FLIRT 3
HAMMER 3
CVS 3
FLIRT HAMMER CVS 3 Template
CVS registration script mri_cvs_register --template $template --mov $subjid USAGE: mri_cvs_register Required Arguments: --mov subjid: subjid for subject to be moved / registered --template subjid: subjid for subject to be kept fixed (template)** Optional Arguments --outdir directory: output directory where all the results are written (default is SUBJECTS_DIR/mov/cvs) --noaseg: do not use aseg volumes in the registration pipeline (default is 0) --nocleanup: do not delete temporary files (default is 0) --nolog: do not produce a log file (default is 0) --version: print version and exit --help: print help and exit
Applications to tractography Goal: fiber bundle alignment Study: compare CVS to methods directly aligning DWI-derived scalar volumes Conclusion: high accuracy cross-subject registration based on structural MRI images can provide improved alignment Zöllei, Stevens, Huber, Kakunoori, Fischl: “Improved Tractography Alignment Using Combined Volumetric and Surface Registration”, NeuroImage 51 (2010),
Mean Hausdorff distance measures for three fiber bundles CSTILFUNCINATE
Average tracts after registration mapped to the template displayed with iso-surfaces FLIRTFA-FNIRTCVS
Average tracts after registration mapped to the template displayed with morphed tracks FLIRTFA-FNIRTCVS
Average tracts after registration mapped to the template displayed with morphed tracks FLIRTFA-FNIRTCVS
Ongoing development Improve CVS capability to register ex-vivo to in-vivo acquisitions Implemented MI-based volumetric registration (for CVS step 3) to accommodate intensity profile differences Qualitative preliminary results on 4 subjects L. Zöllei, Allison Stevens, Bruce Fischl: Non-linear Registration of Intra-subject Ex-vivo and In-vivo Brain Acquisitions, Human Brain Mapping, June 2010
Target (in-vivo) Masked target 2-step CVS CVS with MI Subject 1
Subject 2 Target (in-vivo) Masked target 2-step CVS CVS with MI
Subject 3 Target (in-vivo) Masked target 2-step CVS CVS with MI
Subject 4 Target (in-vivo) Masked target 2-step CVS CVS with MI