Identification of Alzheimer’s Disease Using Novel Dual Decomposition Technique and Machine Learning Algorithms from EEG Signals

Abstract

Alzheimer’s disease (AD) is one of the neurodegenerative disorders. The rate of AD prevalence is rapidly increasing worldwide. The existing clinical invasive methods and neuro-imaging techniques to detect AD are time-consuming, subjective, and expensive. To overcome these issues, we proposed a new automatic framework for detecting AD at an early stage based on the dual decomposition method. Initially, EEG signals of mild cognitive impairment (MCI), AD, and normal control (NC) patients are divided into five subbands by employing discrete wavelet transform (DWT). Subsequently, a Variational mode decomposition (VMD) is applied to these five EEG subbands for further decomposition into various intrinsic mode functions (IMFs). Afterward, three different multiscale permutation entropy (PE) features, namely Shannon PE (SPE), Tsalli’s PE (TPE), and Renyi PE (RPE), have been measured from each IMF. Later, these features have been used to train and test ensemble bagged tree (EBT), k-nearest neighbor, support vector machine (SVM), decision tree (DT), and neural networks with a 10-fold cross-validation scheme. The proposed method has been verified using EEG signals of 59-AD, 7-MCI, and 102-NC subjects. The results obtained from the proposed DWT-VMD method provide 95.20% accuracy for three-class and 97.70% for two-class classification using an EBT classifier with 10-fold cross-validation. It shows a significant ability to distinguish AD from MCI. The proposed dual decomposition method can employ for other neurodegenerative disorders such as Parkinson’s disease, epilepsy, various sleep disorders, and major depressive disorders.