Application of attention network test and demographic information to detect mild cognitive impairment via combining feature selection with support vector machine

Mild cognitive impairment (MCI) is now thought as the prodromal phase of Alzheimer's disease (AD), and the usual method for diagnosing the disease would be a battery of neuropsychological assessment. The present study proposes to integrate a feature selection scheme with support vector machine (SVM) to identify patients with MCI by using attention network test (ANT) and demographic data. Forty-two patients with MCI and forty-five normal individuals underwent ANT recording, and the reaction time and accuracy of ANT and demographics (age, gender, and educational level) were selected as original features. To select features, we first introduced some random variables as probe features in the original data, then ranked all the features according to their influence on the support vector machine decision function, and finally selected those features that had an influence higher than that of the probes. Initially 18 different features were reduced to only four features by our method. SVM classifier created by using these four features gave an 85% classification accuracy with a sensitivity of 85% and a specificity of 86%. And the area under the curve obtained by receiver operating characteristics analysis was 0.918. The experimental results demonstrate that the proposed method is a good potential use to assist identifying patients with MCI objectively and efficiently.