Mri Brain Image Classification Using Polynomial Kernel Principal Component Analysis with Neural Network

Magnetic Resonance (MR) Imaging has come up as widely accepted and revolutionary innovation in field of medical science and brain imaging especially. A new method is proposed here for MRI brain image classification using Polynomial Kernel Principle Component Analysis (KPCA) with Neural Network. In this paper, we are having various stages namely pre-processing, feature extraction, feature reduction and classification of MRI brain images. Here for improving the MRI image quality, imadjust function is used as preprocessing stage of MRI image. In second stage, features are reduced by Polynomial Kernel Principle Component Analysis. In last stage, MRI images are classified as normal or abnormal image by Artificial Neural Network. Different feature reduction methods like PCA, LDA, SVD, Gaussian KPCA and Polynomial KPCA with P (power of kernel) = 2, 3, 4 and 5 are used. The results show that classification rate of 99.8 % is achieved for p = 4 of KPCA.