Ranking based Classification of AD in Hippocampus Region from structural Magnetic Resonance Imaging using Support Vector Machine

Abstract: The automatic classification of Alzheimer's disease has been widely used to investigate the Magnetic Resonance Imaging data. In this, the use of t-test based feature ranking approach is used for feature selection procedure, where the number of top features is determined using Fisher Criterion. The fisher criterion or Irving fisher is a linear discriminate for classification method that projects the high discriminate data and it perform the classification in one-dimensional space. The projection maximizes the distance between the means of two classes while minimizing the variance within each class. The proposed classification involves five systematic levels. First, the voxel based morphometry technique is used to compare the global and local areas of gray matter in patients with Alzheimer's disease differentiate with healthy controls subjects. Some of the significant differences in gray matter volume are determined as volume of interests. Second, the voxel clusters are considered as volume of interest in which each voxel is considered to be a feature. Voxel clusters are assessing and quantifying the MRI based treatments for therapeutic performance. Third, the features are ranked using t-test scores. In this, fisher criterion between AD and HCs groups are calculated for changing the number of ranked features where the size of vector maximizing the fisher criterion. It is selected as the optimal number of top discriminative features. Fourth, the classification is performed using support vector machines. Finally, a data fusion method is used to improve the classification performance among the voxel clusters. Keywords: Feature ranking, Support vector machines, Lib Support Vector Machines, Fisher criterion, Voxel based morphometry and Voxel clusters.