Graduate Student Talk: Soft Segmentation of CT Brain Data

December 4, 2006
Halligan Ext Conf. Room
Speaker: Alexandra Lauric, Tufts University
Host: Sarah Frisken


Unlike research on brain segmentation of MRI data, research on CT brain segmentation is less prevalent. Because MRI is better at differentiating soft tissue, it is generally preferred over CT for brain segmentation. We have begun to explore methods for soft tissue segmentation of the CT brain data with a goal of enhancing the utility of CT for brain imaging. In this study, we consider the effectiveness of existing algorithms, previously used on MRI data, for segmenting brain tissue in CT images. Three methods (Bayesian Classification, Fuzzy c-Means and Expectation Maximization) were used to segment brain matter, bone and cerebrospinal fluid. While these methods outperformed the most commonly used threshold-based segmentation, our results show the need for developing new imaging protocols for optimizing CT imaging to differentiate soft tissue detail and for designing segmentation methods tailored for CT.