Improving the classification performance of amnestic MCI using multiple neuropsychological scales

In this paper, MCI diagnostic model based on neuropsychological scales was researched, and the performance of which was improved by combination of multiple neuropsychological scales. The data of 41 subjects includes 15 AD patients, 12 aMCI patients and 14 normal subjects as control group. Each subject was tested on the neuropsychology scales, which include MMSE, ADAS-cog and CCWS. The feature selection was based on Support Vector Machines-Recursive Feature Elimination (SVM-RFE), then use the PCA algorithm for feature compression. Support Vector Machines was utilized to build the classification model, and the classifier performance were evaluated based on accuracy, and using 10-fold cross validation. Finally, three neuropsychology scales were combined, and the result showed that classification accuracy for AD vs NC, AD vs aMCI and aMCI vs NC respectively was 1.0, 1.0 and 0.9, and classification of the three categories is up to 0.93. The results show that combination of multiple neuropsychological scales can improve the diagnostic performance and classifier performance of aMCI.