CCNCS Seminar Details
A Tissue Segmentation Method for Quantitative Analysis of Pathological Changes in the Brain
|Speaker:||Dr Ali Hojjat|
|Date/Time:||Wednesday 28 January 2009, 4.15pm|
|Location:||Computing Laboratory room S110B|
Segmentation of MR images is an important first step for the quantitative analysis of pathological changes in the brain. A wide range of image segmentation techniques have been proposed in the literature. The most commonly used methods are based on template (atlas) matching which initialize an algorithm based on a common model of brain to initialize the segmentation algorithm. These techniques are biased towards the template used for segmentation. I will describe a low level region based algorithm for the segmentation of brain tissues in MRI. The algorithm uses a gray level similarity criterion to expand the region and uses the peak of a discontinuity measure to segment the region. The segmentation method is used to segment different tissue types in MRI, including WM, GM, CSF and white matter lesions. It is also used to segment large lesions due to hemorrhage. The performance of tile algorithm is evaluated and validated on a series of pathologic three-dimensional MR images of the head.