Convolutional neural network in upper limb functional motion analysis after stroke.

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Artificial Intelligence in Healthcare

Background:

• Stroke frequently impairs upper limb function, necessitating objective movement analysis.
• Understanding functional movement patterns is crucial for effective rehabilitation strategies.

Purpose of the Study:

• To compare upper limb movement patterns in post-stroke individuals with healthy controls during activities of daily living.
• To develop and apply a Convolutional Neural Network (CNN) for classifying upper limb motion data.
• To identify specific upper body segments most sensitive to movement changes post-stroke.

Main Methods:

• Utilized an optical, marker-based motion capture system for data acquisition.
• Compared motion features of dominant/non-dominant limbs in healthy participants versus paresis/non-paresis limbs in stroke survivors.
• Developed a novel CNN application for classifying motion data into two class label configurations.

Main Results:

• The CNN successfully classified upper limb movement data.
• Significant differences in upper limb motion patterns were observed between stroke survivors and healthy individuals.
• The arm segment was identified as the most sensitive indicator of changes in lifting movement trajectories.

Conclusions:

• CNNs are a viable tool for analyzing functional upper limb movement patterns in stroke rehabilitation.
• Objective quantification of movement deficits can inform personalized therapy.
• The arm's sensitivity highlights its importance in assessing post-stroke motor recovery.