Context Independent Expectation Maximization Algorithm for Segmentation of Brain MR Images

For analyzing neurological disorders, realistic analysis of brain MRIs serves as a prerequisite step. This realistic analysis can be best described by segmenting the image into its constituent parts. Unfortunately, segmentation carried out by human visual system (HVS) is always influenced by certain factors. For example, inter-observer, intra-observer variability and large medical datasets. These factors make routine clinical applicability of HVS, a non practical way of examining MRIs. Therefore, to address this problem a fully automatic method is need of the hour. This paper discusses a highly efficient method i.e. the Expectation Maximization (EM) that precisely separates various parts of brain from a brain MRI. It works on the phenomenon of pixel labeling. The results obtained through this method are quite encouraging and are likely to contribute significantly in analyzing brain MRIs.