Abstract: Image segmentation is an indispensible part of the visualization of human tissues during the analysis of Magnetic Resonance Imaging (MRI). MRI is an advanced medical imaging technique which provides rich information for detecting Cerebrospinal Fluid (CSF) in brain images. The changes in the CSF protein level forms abnormal brain deposits strongly linked to variety of neurological diseases. The traditional clustering methods yield more false positives. The proposed system encompasses the following steps, Pre-Processing the MRI brain images using Contrast Limited Adaptive Histogram Equalization, Clustering by Fuzzy C Means (FCM), Total Variation FCM (TVFCM), Anisotropic Diffused TVFCM (ADTVFCM), Optimizing the clustering techniques using Particle Swarm Optimization (PSO) (FCM-PSO, TVFCM-PSO and ADTVFCM-PSO). The clustering methods provide only local optimal solution. In order to achieve global optimal solution, the clustering methods are further optimized using PSO. The performance of the optimized clustering methods is measured using defined set of Simulated MS Lesion Brain database. The optimized clustering methods finds the CSF present in MRI brain images with 98% of accuracy, 92% of sensitivity and 97% of specificity.
P. Tamije Selvy, V. Palanisamy and M. Sri Radhai, 2013. Segmentation of CSF in MRI Brain Images Using Optimized Clustering Methods. Asian Journal of Information Technology, 12: 109-116.