Difference between revisions of "Projects:TensorBasedStatistics"
From NAMIC Wiki
| (One intermediate revision by the same user not shown) | |||
| Line 1: | Line 1: | ||
| − | |||
| − | |||
Back to [[NA-MIC_Collaborations|NA-MIC_Collaborations]], [[Algorithm:Utah|Utah Algorithms]] | Back to [[NA-MIC_Collaborations|NA-MIC_Collaborations]], [[Algorithm:Utah|Utah Algorithms]] | ||
| − | + | = Tensor Based Statistics = | |
Developing new methods for statistical group tests of diffusion tensor data. Our goal is to be able to detect subtle white matter differences between schizophrenics and normal controls beyond what can be detected today using current methods. | Developing new methods for statistical group tests of diffusion tensor data. Our goal is to be able to detect subtle white matter differences between schizophrenics and normal controls beyond what can be detected today using current methods. | ||
| − | + | = Description = | |
* We are applying our methods to the cingulum bundle data from Harvard. First, we have developed a new measure of anisotropy, called geodesic anisotropy (GA), that we are comparing with the standard fractional anisotropy (FA) to see if it gives more statistical power in group tests. Our results on the cingulum bundle data have shown similar results using FA or GA, that is, they show no advantage to using GA over FA. Another method that we are developing compares distributions of data within regions of interest rather than just a single summary measurement. Our preliminary results have not shown differences that are detectable using the mean FA, and we are working to improve upon the methodology. Finally, we are building DTI processing tools to help Core 3 researchers, including methods for interpolating, filtering, and segmenting DTI. | * We are applying our methods to the cingulum bundle data from Harvard. First, we have developed a new measure of anisotropy, called geodesic anisotropy (GA), that we are comparing with the standard fractional anisotropy (FA) to see if it gives more statistical power in group tests. Our results on the cingulum bundle data have shown similar results using FA or GA, that is, they show no advantage to using GA over FA. Another method that we are developing compares distributions of data within regions of interest rather than just a single summary measurement. Our preliminary results have not shown differences that are detectable using the mean FA, and we are working to improve upon the methodology. Finally, we are building DTI processing tools to help Core 3 researchers, including methods for interpolating, filtering, and segmenting DTI. | ||
| Line 14: | Line 12: | ||
** Possible collaboration on comparing new and old anisotropy measures in the context of Schizophrenia. | ** Possible collaboration on comparing new and old anisotropy measures in the context of Schizophrenia. | ||
| − | + | = Key Investigators = | |
| − | * Utah: Tom Fletcher, Ross Whitaker | + | * Utah: Tom Fletcher, Ross Whitaker |
| − | * BWH: Sylvain Bouix, Marek Kubicki, Martha Shenton | + | * BWH: Sylvain Bouix, Marek Kubicki, Martha Shenton |
| − | + | = Links = | |
* http://pnl.bwh.harvard.edu/index.html | * http://pnl.bwh.harvard.edu/index.html | ||
| − | |||
| − | |||
Latest revision as of 19:23, 27 November 2007
Home < Projects:TensorBasedStatisticsBack to NA-MIC_Collaborations, Utah Algorithms
Tensor Based Statistics
Developing new methods for statistical group tests of diffusion tensor data. Our goal is to be able to detect subtle white matter differences between schizophrenics and normal controls beyond what can be detected today using current methods.
Description
- We are applying our methods to the cingulum bundle data from Harvard. First, we have developed a new measure of anisotropy, called geodesic anisotropy (GA), that we are comparing with the standard fractional anisotropy (FA) to see if it gives more statistical power in group tests. Our results on the cingulum bundle data have shown similar results using FA or GA, that is, they show no advantage to using GA over FA. Another method that we are developing compares distributions of data within regions of interest rather than just a single summary measurement. Our preliminary results have not shown differences that are detectable using the mean FA, and we are working to improve upon the methodology. Finally, we are building DTI processing tools to help Core 3 researchers, including methods for interpolating, filtering, and segmenting DTI.
- March 25, 2005. University of Utah visit of Harvard VA for collaborative work on DTMRI.
- Presentation of non linear statistics for tensors by Tom Fletcher.
- Possible collaboration on comparing new and old anisotropy measures in the context of Schizophrenia.
Key Investigators
- Utah: Tom Fletcher, Ross Whitaker
- BWH: Sylvain Bouix, Marek Kubicki, Martha Shenton
