Decision support for stroke rehabilitation therapy via describable attribute-based decision trees

Area of Science:

• Computational biomechanics
• Rehabilitation engineering
• Clinical assessment tools

Background:

• Stroke survivors often require extensive upper extremity rehabilitation.
• Objective and reliable assessment of movement quality is crucial for effective therapy.
• Telerehabilitation offers a scalable solution but requires robust assessment tools.

Purpose of the Study:

• To develop and validate a computational framework for movement quality assessment.
• To assist physical therapists in telerehabilitation settings.
• To evaluate upper extremity function in stroke survivors using kinematic data.

Main Methods:

• A decision tree model was developed using key kinematic attributes.
• Data was collected from eight stroke survivors performing a reach-to-grasp cone task.
• Movement quality scores were correlated with expert therapist ratings using a custom rubric.

Main Results:

• The computational framework reliably assessed movement quality.
• The framework's scores showed high correlation with therapist ratings.
• The decision tree model demonstrated potential for large-scale, unsupervised rehabilitation data analysis.

Conclusions:

• The proposed framework can serve as an assistive tool for physical therapists.
• It offers a reliable method for objective movement quality assessment in telerehabilitation.
• This approach may reduce rehabilitation time and costs, especially for home-based training.