Hand Grasp Motion Intention Recognition Based on High-Density Electromyography in Chronic Stroke Patients

Area of Science:

• Biomedical Engineering
• Neurorehabilitation
• Robotics

Background:

• Stroke significantly impairs motor function and quality of life.
• Robot-assisted rehabilitation offers superior outcomes compared to conventional methods.
• Effective robot assistance requires precise recognition of patient's volitional movement intentions.

Purpose of the Study:

• To investigate the efficacy of high-density electromyography (HD-EMG) for recognizing hand grasp motion intentions in chronic stroke patients.
• To develop and validate algorithms for intention recognition to facilitate robot-assisted therapy.

Main Methods:

• HD-EMG signals were recorded from muscles involved in hand grasp in three chronic stroke patients.
• An adaptive onset detection algorithm was employed to pinpoint the initiation of grasp motions.
• A convolutional neural network (CNN) was trained to classify four distinct hand grasp motions from HD-EMG data.

Main Results:

• The grasp onset detection achieved an average true positive rate of 91.6% and a false positive rate of 9.8%.
• The CNN model accurately classified grasping motions with an average accuracy of 76.3%.
• These findings demonstrate reliable intention recognition capabilities.

Conclusions:

• HD-EMG provides a viable method for accurate hand grasp intention recognition in chronic stroke patients.
• This technology holds significant potential for enhancing the effectiveness of robot-based rehabilitation strategies.
• Improved intention recognition can lead to more intuitive and responsive robotic assistance for stroke recovery.