Detection of Alzheimer's Disease From EEG Signals Using Improved MCh-EVDHM-Based Rhythm Separation

In this letter, we propose a new framework for Alzheimer's disease (AD) detection using electroencephalogram (EEG) signals. The EEG signals are decomposed into a set of elementary components using improved multichannel eigenvalue decomposition of Hankel matrix (MCh-EVDHM) technique. A rhythm separation method is proposed based on improved MCh-EVDHM technique. Then, the total energy and statistical features are extracted from the EEG rhythms. The features are classified into AD and healthy classes using machine learning classifiers. The proposed framework achieved an accuracy of 98.9% and 95.6% in eyes closed and eyes open states, respectively. The proposed framework is compared with the state-of-the-art methods from the literature and found to be more robust, and provides comparable performance measures. Furthermore, the performance of the proposed framework is validated from a combination of EEG signals recorded during eyes open and closed states and achieved an accuracy of 97.3%. The model size of the classifier utilized in the proposed framework is also presented.