Preliminary Analysis of Hippocampal Shape from NAMIC Darmouth Data
description
Transcript of Preliminary Analysis of Hippocampal Shape from NAMIC Darmouth Data
NA-MICNational Alliance for Medical Image Computing
Preliminary Analysis ofHippocampal Shape fromNAMIC Darmouth Data
Processing: Martin Styner
Data: Andy Saykin
National Alliance for Medical Image Computing http://na-mic.org
General Items
• UNC shape pipeline worked without any modification on the Dartmouth data
• Processing pipeline not yet ready for Core 3 since no point&click User Interface
• Hippocampal shape highly variable in data– I expect the need for a larger dataset in order
to show consistent, reliable shape differences
• Timeline– Data received: Sept 8 2005– This report with full results: Sept 14 2005
National Alliance for Medical Image Computing http://na-mic.org
Shape Analysis Workflow
Hippocampus Segmentation
Preprocessing &Parameterization
SPHARM-PDM Shape
QCShape &Corresp.
Alignment& Scale
Feature Computatione.g. Subdivision or
Mean Shape Difference
QC of Features & Statistical Results
Statistical AnalysisOf Features
National Alliance for Medical Image Computing http://na-mic.org
Pre-processing
• Sample case• Segmentation
after pre-processing (establishing spherical topology)
• 0.5mm^3 resolution
National Alliance for Medical Image Computing http://na-mic.org
SPHARM Shape
• Spherical parametrization• Mirroring of right hippocampus• Alignment to first order ellipsoid• Alignment to template
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
National Alliance for Medical Image Computing http://na-mic.org
SPHARM Shape
• Overlay of voxel segmentation (red) with SPHARM (blue)
• Average Error ~ 0.12mm
National Alliance for Medical Image Computing http://na-mic.org
SPHARM Correspondence
National Alliance for Medical Image Computing http://na-mic.org
SPHARM Mean Hippo
Overall Mean Left Mean Right Mean
National Alliance for Medical Image Computing http://na-mic.org
Asymmetry: MeshValmet
Left vs RightAsymmetry ofMean Shapes(only 6 cases)
National Alliance for Medical Image Computing http://na-mic.org
Statistical Testing
• Only 6 cases: Do not interprete this data, it’s just an example
• Global p-value for asymmetry shape difference: 0.34
Mean difference Raw p-valueMean Overlay
National Alliance for Medical Image Computing http://na-mic.org
Hippocampal Subdivision
• Skeleton based Subdivision (11 parts)
• Fusion into 3 (head, body, tail) or 5 parts (ant/post head, ant/post body, tail)
National Alliance for Medical Image Computing http://na-mic.org
Hippocampal Subdivision
• Subdivision Automatic QC images
National Alliance for Medical Image Computing http://na-mic.org
Hippocampal SubdivisionNAMIC01 l hippo 0 4477 423 1850 2204
NAMIC02 l hippo 0 3895 311 1642 1942
NAMIC03 l hippo 0 4398 336 2054 2009
NAMIC04 l hippo 0 3621 329 1566 1726
NAMIC05 l hippo 0 2610 265 1147 1198
NAMIC06 l hippo 0 3263 301 1412 1549
NAMIC01 r hippo 1 4788 368 2099 2322
NAMIC02 r hippo 1 4747 361 2103 2283
NAMIC03 r hippo 1 4168 315 1903 1949
NAMIC04 r hippo 1 3416 291 1465 1660
NAMIC05 r hippo 1 3315 294 1466 1556
NAMIC06 r hippo 1 3787 316 1640 1831
Total Tail Body Head
Mean Right 4037 324 1779 1934
Mean Left 3711 327 1612 1771
P-value Asymmetry 0.14 0.84 0.15 0.10
National Alliance for Medical Image Computing http://na-mic.org
Hippocampal subdivision
• Agreement with local shape analysis
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
Trend
P < 0.5
National Alliance for Medical Image Computing http://na-mic.org
Conclusion SPHARM
• Success! (ALL images in this ppt are from Darmouth hippocampus data)
• Hippocampus shape is variable, but correspondence and alignment worked very nicely
• Of course, no real analysis was possible with 6 datasets, but asymmetry analysis shows good example