Integrating EEG Artifact Removal with Deep Learning for Accurate Motor Imagery Classification in Acute Stroke

Nowadays, the number of stroke incidents has increased, particularly in aging populations and areas with limited access to healthcare. This rise has highlighted the need for innovative assistance solutions to improve the quality of life of stroke survivors. This study aimed to address the challenges of noisy and artifact-laden Electroencephalogram (EEG) data by employing preprocessing with autonomous and semi-autonomous artifact removal techniques, including the use of EEGLab. This study focuses on the classification of EEG data for stroke patients using Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs). Combining the strengths of CNNs and LSTMs, a hybrid model leverages both spatial and temporal features of EEG data to improve classification accuracy. The hybrid CNN-LSTM model outperforms individual CNN and LSTM models in classifying motor imagery tasks, demonstrating its potential for robust brain-computer interface applications in stroke rehabilitation. This integrative approach not only improves classification performance but also sets the stage for more effective therapeutic interventions, ultimately aiming to enhance patient outcomes.