Абстрактный

A Novel Approach to Improve Kernel SVM Classification Algorithm Using Hybrid Approach Based on Texture and Statistical Features of Alzheimer��?s Stage Using Neuro MR Image

Shaik Basheera*, M Satya Sai Ram

Alzheimer’s (AD) stage classification is carried on 54 number of T2 weighted Magnetic resonance Images of the subject at different stages. Generating the statistical and textual features are collected by performing segmentation of white matter, gray matter, cerebral spinal fluid, area calculated taken as attributes for the data frame. Classification of the data is carried by Supported vector machines and multilayer perceptron algorithm and compared the result of the classification using Area under curve (AUC), Classification Accuracy (CA), F1, Precision, recall and it is observed that multilayer perceptron with a two Hidden layers having given better 100% classification accuracy kernel SVM gives 96.29% of classification accuracy, by hybrid approach it is possible to increase the kernel SVM performance to 100% classification accuracy.