Correlation Analysis Approach Between Features and Motor Movement Stimulus for Stroke Severity Classification of EEG Signal Based on Time Domain, Frequency Domain, and Signal Decomposition Domain

Abstract

The healing process of a stroke necessitates tools for measuring relevant parameters to facilitate monitoring, evaluation, and medical rehabilitation. Accurate parameter measures can be observed in stroke patients' severity to ascertain suitable interventions by identifying components pertinent to monitoring, evaluation, and medical rehabilitation. The components are derived from the observation collection process utilizing an EEG device, accompanied by a motor stimulus, to ensure the acquisition of EEG signals for monitoring, evaluation, and medical rehabilitation while preventing any loss of information during data collection. The acquired information encounters challenges due to the signal's unstable, nonlinear, and non-stationary characteristics, necessitating efforts to stabilize, render stationary, and linearize it through suitable signal processing and feature extraction techniques to achieve a pertinent feature composition. The subsequent difficulty is achieving the objectives of medical monitoring, evaluation, and rehabilitation, necessitating the correlation between EEG signal characteristics and motor movement stimuli, ensuring that the process adheres to appropriate parameter identification and scheduling per the established plan. In response to this difficulty, a correlation analysis methodology is established, incorporating normalcy tests, significance tests, and correlation analysis to ensure that the relevant factors for identifying stroke severity categorization patterns are precisely identified beforehand. The correlation analysis strategy employs raw data situations, preprocessing, feature extraction, feature selection, and correlation analysis for classification purposes. Our experimental findings indicate that the correlation analysis approach for assessing stroke severity classification patterns is evident in the Hajorth Complexity feature, utilizing the Shoulder motor movement stimulus and the SVM classification type, achieving an accuracy significant value of 98%. These findings confirm the efficacy of correlation analysis between EEG signal features and motor movement stimuli in identifying the optimal parameters within a reduced dimensional space to assess stroke severity effectively.