A Novel Deep Learning Method Based on an Overlapping Time Window Strategy for Brain-Computer Interface-Based Stroke...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Stroke is a major cause of death and disability worldwide.
• Brain-computer interfaces (BCIs) are crucial for motor rehabilitation in stroke survivors.
• Classifying motor intentions from brain activity is key for effective BCI-based rehabilitation.

Purpose of the Study:

• To present a novel method for training EEG-based BCI models using overlapping time windows.
• To enhance the classification accuracy of motor attempts (MA) for stroke patients undergoing BCI rehabilitation.

Main Methods:

• Employed three deep learning models: Convolutional Neural Network (CNN), Graph Isomorphism Network (GIN), and Long Short-Term Memory (LSTM).
• Utilized overlapping time windows for model training and compared different window lengths.
• Implemented a vote-counting strategy (VS) with the LSTM model.

Main Results:

• The deep learning approach with overlapping time windows significantly improved classification accuracy.
• The LSTM model combined with vote-counting achieved the highest average classification accuracy of 90.3% with a 70-unit window size.
• Experimental results confirmed the efficacy of the overlapping time window strategy.

Conclusions:

• Overlapping time windows are an effective strategy for enhancing the performance of EEG-based BCIs in stroke rehabilitation.
• The proposed method, particularly the LSTM-VS approach, offers a promising advancement for BCI rehabilitation efficiency.
• This technique can lead to more effective motor function recovery for stroke patients.