2008 Summer Project Week:LobeParcellation

Return to Project Week Main Page

Key Investigators

• BWH: Sylvain Bouix, Yogesh Rathi, Padmapriya Srinivasan
• SPL: Kilian Pohl
• Kitware: Brad Davis

Objective

To build a lobe parcellation algorithm for our 3T data set. The spatial priors are based on 100 existing manually-lobe parcellated 1.5T data.

Approach, Plan

Our plan is to use an algorithm described in Reference 1 to construct probability maps of the lobes using an unbiased registration approach. The algorithm uses a label space representation that allows for direct registration. We will then tune the EMsegmenter of Slicer3 to perform the segmentation.

Progress

We have been using a MATLAB(R) implementation of the atlas building algorithm to construct the atlas and fine tuning it for 10 subjects.

References

1. James Malcolm, Yogesh Rathi,and Allen Tannenbaum, "Label Space: A multi-object Shape Representation", IWCIA 2008,LNCS 4958, pp. 185-196.
2. Motoaki Nakamura, Dean F. Salisbury, Yoshio Hirayasu, Sylvain Bouix, Kilian M. Pohl, Takeshi Yoshida, Min-Seong Koo, Martha E. Shenton, and Robert W. McCarley, "Neocortical Gray Matter Volume in First-Episode Schizophrenia and First-Episode Affective Psychosis: A Cross-Sectional and Longitudinal MRI Study", Biol Psychiatry 2007;62:773–783