Statistically rigorous human movement onset detection with the maximal information redundancy criterion

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Machine Learning in Healthcare

Background:

• Stroke survivors often experience impaired ability to perform activities of daily living (ADL).
• Quantitative assessment of sensorimotor function during ADL tasks is crucial for monitoring stroke recovery.
• Identifying movement initiation in force/torque data is challenging but vital for detailed analysis.

Purpose of the Study:

• To develop an automated method for extracting movement initiation from sensorimotor force/torque measurements during ADL tasks in stroke patients.
• To evaluate the effectiveness of the Maximal Information Redundancy (MIR) criterion for movement onset detection.

Main Methods:

• Utilizing a machine learning approach based on the Maximal Information Redundancy (MIR) criterion.
• Analyzing force and torque signals from stroke patients performing standardized ADL tasks.
• Applying MIR to identify signal characteristics indicative of movement initiation, assuming increased signal redundancy.

Main Results:

• The MIR criterion successfully identified movement initiation phases in the sensorimotor data.
• The automated detection accuracy closely matched that of clinical experts.
• This method provides a quantitative measure related to the early stages of movement during ADL tasks.

Conclusions:

• Automated extraction of movement initiation using the MIR criterion is feasible and accurate.
• This technique offers a promising tool for objective and quantitative assessment of motor recovery post-stroke.
• The findings support the use of machine learning for analyzing sensorimotor data in rehabilitation.