Integrated approach of brain segmentation using neuro fuzzy k-means

Abstract

A proposed method using neuro-fuzzy k-means for the segmentation process of brain has been developed successfully, simulated and assessed. The proposed method has been assessed by using clinical brain images of magnetic resonance imaging (MRI) technology, to segment the three main tissues of the brain. The proposed system is able to segment the three important regions of the brain, which are white matter, grey matter and cerebrospinal fluid (CSF) more accurately, as compared to the benchmarked algorithms. Furthermore, the developed method’s misclassification rate (MR) has been significantly minimized by 88%, 27%, 88%; 82%, 71%, 84%; and 82%, 29%, 83%, as compared to k-means, fuzzy logic, and radial basis function (RBF) for white matter, grey matter and CSF, respectively. Also, from the visual interpretation, it is observed that the brain’s edges are well preserved and the tissues are clearly segmented. From these measures, the proposed integrated approach is shown to be accurate in segmenting the MRI brain tissue with reduced misclassified pixels.