A 10-item Fugl-Meyer Motor Scale Based on Machine Learning

Area of Science:

• Neurology
• Rehabilitation Medicine
• Biomedical Engineering

Background:

• The Fugl-Meyer motor scale (FM) is a standard for assessing motor function post-stroke.
• Its extensive item count (50 items) limits practical clinical application.

Purpose of the Study:

• To develop a machine learning-derived short form of the FM (FM-ML) for stroke patients.
• To evaluate the efficiency and psychometric properties of the FM-ML against existing FM versions.

Main Methods:

• A random lasso machine learning method identified 10 key items for the FM-ML.
• Artificial neural networks were used to calculate FM-ML scores.
• Concurrent validity, predictive validity, responsiveness, and test-retest reliability were assessed for all FM versions.

Main Results:

• The 10-item FM-ML demonstrated comparable psychometric properties to longer FM versions.
• FM-ML achieved high concurrent validity (r=0.95-0.99), responsiveness (r=0.78-0.91), and test-retest reliability (ICC=0.88-0.92).
• FM-ML used significantly fewer items, offering a 80% reduction compared to the original FM.

Conclusions:

• The 10-item FM-ML shows promising efficiency and robust psychometric properties.
• This tool has the potential to enhance the speed and effectiveness of motor function evaluations in stroke rehabilitation.