An Algorithm for Activity Recognition and Assessment of Adults Poststroke

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Occupational Therapy

Background:

• Stroke is a leading cause of long-term disability, yet objective tools for measuring home-based activity in survivors are lacking.
• Accurate assessment of upper extremity function is crucial for recovery and independence after stroke.

Purpose of the Study:

• To integrate validated upper extremity assessments into an activity recognition system for poststroke individuals.
• To evaluate the accuracy and utility of this system in a home environment.

Main Methods:

• An exploratory descriptive study analyzed previously collected 3-month home-based data from 16 community-dwelling participants (≥6 months poststroke).
• An activity detection algorithm was tested for accuracy against manual labels.
• The algorithm assessed upper extremity movement quality, including reach, speed, and smoothness.

Main Results:

• The algorithm achieved 87% to 95% accuracy in identifying in-kitchen activities.
• It successfully detected the quality of upper extremity movements, providing detailed metrics.
• Participant demographics included 9 women and 7 men, with an average age of 63.38 years.

Conclusions:

• The developed algorithm accurately identifies poststroke activities and quantifies upper extremity movement quality in the home setting.
• This technology offers a potential supplement to clinical evaluations, enhancing treatment planning and outcomes reporting in occupational therapy.
• The findings highlight the potential of technology to provide objective, clinically relevant data for stroke rehabilitation.