Alzheimer disorders diagnosis system design using machine learning for EEG signal

Abstract

The diagnosis of Alzheimer's disorders (AD), a prevalent neurological disorder, can created by utilising a range of therapeutic methods, including the electroencephalogram (EEG), which has been especially successful in the past. The objective for this study is to develop a computer-aided diagnosis tool which may recognize AD from EEG data. The EEG information was cleaned up with a band-pass elliptic digital filter to remove any interference or disruptions. The filtered signal was then divided into its frequency ranges using the Discrete Wavelet Transform (DWT) method, enabling the extraction of EEG signal characteristics. To enhance the diagnostic performance, various signal features were integrated into the DWT technique, including logarithmic band power (LBP), standard deviation, and root mean square (RMS). These features were incorporated to generate feature vectors that were utilized in the diagnosis process. In order to categorize the EEG features into their respective classes, various machine learning (ML) approaches were explored, including quadratic discriminant analysis (QDA), support vector machine (SVM), and k-nearest neighbour (KNN). The performance of these approaches was then evaluated by comparing and computing their sensitivity, specificity, and overall diagnosis accuracy. When comparing the accuracy values, it can be observed that QDA has a higher accuracy of 97.5% compared to KNN's accuracy of 95%.