A Spatial Model of White Matter Fiber Tracts

Mahnaz Maddah, William M. Wells, S.K. Warfield, C.-F. Westin, W. Eric L. Grimson
Proceedings of the ISMRM Annual Meeting (ISMRM'07)
Pages 1
2007

Download full paper

Abstract

We present a method to build a spatial model of the white matter fiber tracts. Knowledge of intra-bundle point correspondence between trajectories is essential for building such a model. Using the expectation maximization (EM) algorithm, fiber trajectories are clustered into bundles in a gamma mixture model context. Point correspondence is efficiently obtained though computing a distance map and a label map for each cluster at each iteration of the EM algorithm. The output of the algorithm is a probabilistic assignment of the fiber trajectories to each cluster as well as a model of each bundle defined with the average trajectory and its spatial variation. The constructed model provides an efficient way to study inter-subject and temporal variations of the fiber tracts as well as a fast 3-D visualization method for neurosurgical applications.

Trajectories of (CC) and (MCP) tracts clustered into bundles (a) and their corresponding spatial models (b). Each model is represented by the average trajectory and the isosurfaces defined by the point-wise standard deviations. Similar results are shown for roughly 3000 trajectories from major white matter fiber tracts in (c) and (d).

Reference

Maddah M, Wells WM, Warfield S, Westin CF, Grimson WEL. A spatial model of white matter fiber tracts. In Proceedings of the ISMRM Annual Meeting (ISMRM'07). 2007;1.

Bibtex entry

@InProceedings{maddahISMRM07,
  author         = {Mahnaz Maddah and William M. Wells and S.K. Warfield and   
                   Carl-Fredrik Westin and W. Eric L. Grimson},                
  title          = {A Spatial Model of White Matter Fiber Tracts},             
  booktitle      = {Proceedings of the {ISMRM} Annual Meeting (ISMRM'07)},     
  pages          = {1},                                                        
  year           = {2007}
}

Grants

NIH U54-EB005149 (NAMIC), NIH R01-MH074794