Machine Learning-Based Prediction of Transition to Functional Upper Limb Recovery After Intensive Inpatient...

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Artificial Intelligence

Background:

• Upper limb functional recovery post-stroke is unpredictable, posing a challenge for rehabilitation.
• Developing accurate prognostic models is crucial for effective stroke patient care.

Purpose of the Study:

• To develop and validate machine learning (ML) prognostic models for predicting functional upper limb recovery in early subacute stroke patients.
• To identify key predictors of motor, dexterity, and strength recovery.

Main Methods:

• Retrospective analysis of 960 stroke patients undergoing intensive inpatient rehabilitation.
• Defined recovery outcomes: Fugl-Meyer score (motor), Box and Block Test (dexterity), and pinch strength.
• Utilized Random Forest and other ML algorithms with cross-validation and temporal validation.

Main Results:

• Random Forest models achieved high accuracy (AUROC 0.800-0.958) in temporal validation.
• Baseline motor severity and corticospinal tract integrity were key biological predictors.
• Earlier rehabilitation and robot-assisted therapy showed associations with improved outcomes.

Conclusions:

• ML-based prediction models integrating neural and rehabilitation factors can enhance precision rehabilitation planning.
• These models improve functional outcome stratification for early subacute stroke survivors.
• Supports individualized treatment strategies to optimize upper limb recovery post-stroke.