A Classifier of MR Brain Images through KSVM Optimized with Principal Component Analysis

The most important fields in clinical analysis and interpretation isaccurate and automated classification of Medical Resonance brain images. Based on the features obtained from the captured MR brain images, they can be classified into normal and abnormal. This can be done using many different methods. We have come up with an effective method that can be used for the classification and analysis of the brain images. Initially, we use wavelet transform for the extraction of features and attributes from the images. Then, the application of Principal Component Analysis (PCA) is done for the compression of image dimensions and the resultant is given to the Kernel Support Vector Machine (KSVM). For enabling the enhancement of the generalization of K-SVM, the utilization of K-fold stratified cross validation technique is being done. The images of brain inclusive of twenty normal and One hundred and forty abnormal samples were acquired and the proposed technique was performed. The accuracy level achieved using our proposed method was 99.38%. Also, various other methods like LIN, IPOL and HPOL shows about 95%, 98.l2% and 96.88% in their accuracy levels respectively. Hence, it is obvious that the technique of proposed strategy reaches the peak accuracy compared to the other existing methods. This method is also effective and faster when done practically. Also, they can be applied in classification of brainMR images where the doctors are assisted in the diagnosis of a disease even when the patient is abnormal state or normal state to a certain extent. Thus, our proposed technique has proven to be efficient, highly accurate and less time consuming for the classification of the brain medical resonance images.