An Automated EEG-based mild cognitive impairment diagnosis framework using spectral and functional connectivity features

Accurate and early diagnosis of mild cognitive impairment (MCI) is essential for preventing the progress of Alzheimer's and other types of dementia. However, the MCI diagnosis is very challenging due to its complicated manifestations. On this subject, many studies proposed automatic techniques for MCI diagnosis based on electroencephalogram (EEG) signals. In this study, an automatic EEG-based MCI diagnosis framework is presented, which uses spectral and functional connectivity EEG features. The sequential backward feature selection (SBFS) algorithm is utilized to select the best subset of features. K-nearest neighbor (KNN), support vector machine (SVM) with linear kernel (LSVM), and SVM with radial basis function (RBF) kernel (RBFSVM) are evaluated to select the best ones for the proposed framework. The proposed framework is validated using a public dataset consisting of 16 and 18 EEG signal of healthy control (HC) and MCI subjects. It is validated using 10-fold cross-validation by the metrics such as accuracy (AC), sensitivity (SE), and specificity (SP). The best performance of the proposed framework was achieved by LSVM classifier, which provided an average AC of 98.8%, SE of 98.5%, and SP of 98.8%. These results indicate the accurate classification performance of the proposed framework. Moreover, these results indicate the proposed framework is more accurate than previous works for EEG-based automatic MCI diagnosis. This framework could be developed to use it as a computer-aided diagnosis (CAD) tool for clinical applications.