An Efficient Brain Image Segmentation based on Gradient Based Watershed transform in Level set method and classification using shape features for a medical diagnosis system

ABSTRACT We propose a simple, fast, robust and efficient technique to extract the skeleton based shape signatures for the brain image classification for a medical diagnosis system. The Improved Brain image classification system uses five shape features- two features have derived from combination of skeleton, region, and boundary information and the other three have been derived from distance mapped functional(level contours). All these shape features exhibit invariance to rotation and scaling. Brain image classification is one of the utmost imperative parts of clinical investigative tools. Brain images typically comprise noise, inhomogeneity and sometimes deviation. Therefore, precise segmentation of brain images is a very challenging task. Nevertheless, the process of perfect segmentation of these images is very important and crucial for a spot-on diagnosis by clinical tools. This research presents a more accurate segmentation using Gradient Based watershed transform in level set method for a medical diagnosis system.