Scoring upper-extremity motor function from EEG with artificial neural networks: a preliminary study.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Medicine

Background:

• Current motor function assessments for chronic stroke survivors are often subjective and require in-person evaluations by trained professionals.
• There is a need for more objective, accessible, and scalable methods for motor function assessment.

Purpose of the Study:

• To investigate the feasibility of automatically scoring upper-extremity motor function using electroencephalography (EEG) signals and artificial neural networks (ANNs).
• To develop a novel, objective method for assessing motor function in chronic stroke survivors.

Main Methods:

• Collected EEG data from healthy participants and chronic stroke survivors during a simple button-clicking task.
• Utilized convolutional neural network (CNN) models trained on participants' Fugl-Meyer motor assessment scores to predict motor function.
• Evaluated model performance using within-participant and cross-participant testing.

Main Results:

• The proposed method achieved high prediction accuracy, demonstrating strong correlation coefficients (r=0.9921 within-participant, r=0.9867 cross-participant) with established motor function scores.
• Statistical significance was confirmed with very low p-values (p<10^-11) for both testing scenarios.
• The results indicate robust performance of the ANN model in predicting motor function from EEG data.

Conclusions:

• The developed method shows significant potential for use as a stable and objective measurement tool for motor function assessment.
• This approach could overcome the limitations of traditional, subjective clinical assessments.
• Automated EEG-based motor function scoring offers a promising avenue for remote and objective patient monitoring in stroke rehabilitation.