Alzheimer's Disease Detection using Empirical Mode Decomposition and Hjorth parameters of EEG signal

Alzheimer's disease (AD) is a progressive neurodegenerative disorder observed in the elderly. AD diagnosis is performed through interviews or questionnaires by an experienced psychiatrist. This process is time-consuming, biased, and subject-specific. Hence, its urgent need to develop an. The paper presents an automatic AD detection system using Electroencephalogram (EEG) signal to alleviate these problems and support neurologists. Nine IMFs (Intrinsic mode functions) are generated for each EEG signal using empirical mode analysis. Ten different features are extracted from these IMFs. Three Hjorth parameters (activity, mobility, complexity) are selected using the Kruskal-Wallis test. The selected features from EEG recordings of 23 subjects (AD-12 and NC-11) are evaluated using the least-square support vector machine (LS-SVM) model with 10-fold cross-validation for three kernels. A maximum of 92.90% classification accuracy is obtained using the features of IMF-4. The results showed that the proposed method detected AD patients efficiently. Further, the proposed method can be used to detect other neurological disorders.

Automated summary

Alzheimer's disease is a common neurodegenerative disorder in the elderly, and the current diagnosis methods have limitations. In this paper, an automatic AD detection system using EEG signals is proposed, which can effectively detect AD patients and has potential for detecting other neurological disorders.