Hybrid approach of Kernelized Fuzzy C-Means and Support Vector Machine for Breast Medical Image Segmentation

M. S. Sheeba and A. Sathya

Abstract

Medical image analysis is becoming progressively more significant in decision making of MRI analysis and computerized diagnosis system. Image segmentation is an important key process for image analysis. In this paper, a novel medical image segmentation technique is proposed which combines Kernelized Fuzzy C-Means and Support Vector Machine. In the proposed system, a robust Hyper tangent induced Kernelized Fuzzy C-Means method is constructed with the inclusion of new spatial information term firstly. And then, the new Support Vector Machine is developed for improving effective segmentation result. The input vector for SVM classifier is generated by membership function of novel FCM in which the pixel data are labeled by new FCM method. In order to accelerate the effectiveness of segmentation result and to deal non linearity, new hyper tangent based similarity measure is used in both KFCM and SVM. Experimental analysis is carried out on real left and right breast MRIs to show the efficiency of proposed method. The performance of proposed method is demonstrated through comparative analysis of proposed and existed methods. Fuzzy Partition coefficient, Fuzzy entropy, iteration count and error rate are used to measure cluster validity. Finally, it is shown that our proposed method is the most promising technique for medical image segmentation.