An automated MRI segmentation framework for brains with tumors and multiple sclerosis lesions

Brain tissue is an intricate anatomical structure and hence thorough detection of numerous brain ailments much relies on precise segmentation of three major tissues, viz., cerebro-spinal fluid (CSF), gray matter (GM) and white matter (WM) in MR brain images. This problem has been addressed in literature, but many key open issues still remains to be investigated. As an initial stride in this development, an automated method for segmentation of deformities like atrophy and tumor in brain MR images is developed. The paper next concentrates on segmenting multiple sclerosis (MS) lesions in WM of central nervous system (CNS). A modified algorithm that relies on the histon based fast fuzzy C-means (HFFCM) is developed. In the former, the experimentation is carried out using brain web datasets and in the latter, the datasets used were from the MICCAI grand challenge II workshop for segmenting MS lesions. The results obtained from the proposed algorithms were compared with the existing methods using performance metrics such as specificity, sensitivity, accuracy, relative absolute volume difference (RAVD), average symmetric absolute surface distance (ASASD) etc. It is observed that the results of segmentation accuracies from the proposed methods were very high when compared with the existing methods.