A NEW APPROACH FOR MRI IMAGE CLASSIFICATION

Abstract

Magnetic Resonance (MR) Imaging has come up as widely accepted and revolutionary innovation in field of brain imaging. MRI is an imaging technique used in medical science to view the internal structures of the body in detail & having incomparable role of importance in brain imaging. MRI is a powerful tool to provide detail and reliable information about human brain, so it proved very helpful in the study of human brain. Here a work is done by simulating a method in MATLAB using artificial neural network to automatically classify brain MRI images. The proposed method for the classification of MRI Brain images consists of various stages namely Pre-processing, Feature extraction, Dimensionality reduction and Classification. Here for improving the MRI image quality, pre-processing is employed in MATLAB using imadjust function. The features are extracted using discrete wavelet transform. In Wavelet Transform, different frequencies are examined with different resolutions. Due to this reason, DWT is widely used for feature extraction. Then the features are reduced using kernel principal component analysis (KPCA). Two commonly used kernels are the polynomial kernel and Gaussian kernel. In this project, we are using polynomial kernel in which order of polynomial are varying (p=2, 3, 4) for achieving good classification rate. In last stage artificial neural network (ANN) is used as a multi class classification technique to classify between normal or abnormal (brain tumor infected) MRI images. The simulation results show that we obtain good classification rate from proposed method.