An upper limb stroke rehabilitation exercise video dataset.

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Computer Vision

Background:

• Stroke is a leading cause of disability globally, with significant challenges in low and middle-income countries.
• Limited access to rehabilitation facilities and trained professionals hinders effective post-stroke recovery in these regions.
• Telerehabilitation offers a scalable solution for remote monitoring and personalized therapy, improving accessibility.

Purpose of the Study:

• To present a novel, annotated video dataset for upper limb rehabilitation exercises.
• To facilitate the development of vision-based assessment systems for telerehabilitation.
• To support the creation of low-cost deep learning solutions for post-stroke care.

Main Methods:

• Collected 491 videos of four distinct upper limb strengthening exercises.
• Videos were captured using standard RGB cameras at 30 frames per second.
• Data was acquired under varied background and lighting conditions with ten volunteers.

Main Results:

• A comprehensive, exercise-specific video dataset was created.
• The dataset is well-annotated, suitable for training deep learning models.
• The data captures realistic variations in exercise performance and environmental conditions.

Conclusions:

• The dataset can enable the development of robust, low-cost telerehabilitation systems.
• This resource has the potential to improve post-stroke care accessibility for disadvantaged populations.
• Vision-based assessment systems are key to advancing remote rehabilitation effectiveness.