Mild Cognitive Impairment Detection from EEG Signals Using Combination of EMD Decomposition and Machine Learning

Mild cognitive impairment (MCI) is the earliest stage of dementia, and its detection is crucial for disease management. Electroencephalography (EEG) has gained popularity as a tool for identifying brain disorders. This article presents methods for diagnosing MCI from EEG signals using empirical mode decomposition (EMD). The proposed methods are tested using EEG data from 32 healthy control (HC) cases and 29 MCI cases. First, major artifacts are removed through preprocessing. EMD is then used to decompose signals into intrinsic mode functions (IMFs), from which features are extracted using one of the following measures: energy, log energy entropy, sure entropy, threshold entropy, Shannon entropy, or band power. Multiple machine learning approaches are then used to classify whether the resulting features are for MCI or HC. Results reveal that the proposed methods, especially the one that combines EMD and entropy, produce competitive results. For instance, the achieved results for a KNN classifier using sure entropy in terms of classification accuracy, specificity, sensitivity, and F-score are 97.60
^%
, 98.59%, 96.49%, and 97.44
^%
, respectively. The findings demonstrate accurate detection of MCI using the proposed methods, indicating potential for clinical diagnosis applications.