Performance Evaluation of Kernel Based Techniques for Brain MRI Data Classification

Magnetic resonance (MR) imaging has been playing an important role in neuroscience research for studying brain images. The classifications of brain MRI data as normal and abnormal are important to prune the normal patient and to consider only those have the possibility of having abnormalities or tumor. Classification of MRI data along with skull in MR images results in reduction of efficiency to a great extent. Thus the removal of skull is done prior to classification. The statistical and gray level co-occurrence features are extracted from MR images before and after skull removed images. An advanced kernel based techniques such as support vector machine (SVM) and relevance vector machine (RVM) for the classification of volume of MRI data as normal and abnormal are deployed. Validation is done with stratified Holdout approach. The results are compared with radiologist results and performance measures such as sensitivity, specificity, and correspondence ratio for skull stripping and classification accuracy are calculated.