Use of Independent Component Analysis to Estimate Multiple Fibers within a Voxel
Singh M, Jeong J
University of Southern California
Detection of crossing fibers within a voxel is a challenge in streamline tractography, particularly in clinical protocols where a relatively small number of gradient directions (6-25) are used. After verifying that the probability density function of the DTI data within voxels likely to contain a single fiber was nongaussian, an Independent Component Analysis (ICA) was performed to seperate two components (representing two fibers) in voxels identified as likely to contain crossing fibers. Results suggest that ICA can detect new tracts through fiber-crossing regions that are undetectable by conventional single-tensor streamline tractography.