Methodology for Classification and Segmentation of Multiple Sclerosis Affected Brain MRI Images

A methodology is developed for classification of brain MRI images as normal and affected by multiple sclerosis (MS). The classification of brain MR images into normal and MS using Back propagation neural network (BPNN) has given maximum accuracy of 96% for the combination of all the texture features. The segmentation of MS lesions is carried out using k-means segmentation with an accuracy of 86%. The k-means segmentation method has given more sensitivity, specificity and accuracy compared to Fuzzy-C-Means (FCM). The work carried out has given good classification & segmentation accuracy and has scope for further improvement.