A Combined sEMG and Accelerometer System for Monitoring Functional Activity in Stroke.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Wearable Sensor Systems

Background:

• Traditional assessments of functional independence in stroke patients rely on subjective, paper-based questionnaires.
• Remote monitoring using body-worn sensors offers a more objective alternative for physical activity assessment.

Purpose of the Study:

• To investigate the classification accuracy of a combined surface electromyography (sEMG) and accelerometer (ACC) sensor system for monitoring activities of daily living (ADL) in stroke patients.
• To identify the minimal sensor configuration for accurate ADL classification with minimal misclassification errors.

Main Methods:

• Collected sEMG and ACC data from 10 hemiparetic stroke patients performing 11 identification tasks and 10 non-identification tasks.
• Analyzed sensor data using a multilayered neural network and an adaptive neuro-fuzzy inference system.
• Determined optimal sensor placement for high classification accuracy and low misclassification rates.

Main Results:

• A configuration of 4 ACC sensors and specific sEMG sensors (upper arms, forearm, thigh) achieved high accuracy.
• Mean sensitivity of 95.0% and specificity of 99.7% for identification tasks.
• Mean misclassification error of less than 10% for non-identification tasks.

Conclusions:

• A hybrid sEMG and ACC wearable sensor system is feasible for automatic recognition of motor tasks.
• This technology can objectively assess functional independence in stroke patients.
• The findings support the use of wearable sensors for remote rehabilitation monitoring.