Regression-Based Machine Learning for Predicting Lifting Movement Pattern Change in People with Low Back Pain

Area of Science:

• Biomechanics
• Rehabilitation Science
• Machine Learning in Healthcare

Background:

• Machine learning (ML) is vital in healthcare, yet regression algorithms' efficacy in predicting lifting movement changes needs assessment.
• Low back pain (LBP) affects numerous individuals, and understanding movement alterations post-rehabilitation is crucial.

Purpose of the Study:

• To pilot regression-based machine learning models for predicting trunk, hip, and knee movement changes after a 12-week strength training program in LBP patients.
• To identify the most accurate regression algorithms for forecasting specific joint movement alterations.

Main Methods:

• Feature extraction calculated sagittal plane range of motion for trunk, hip, and knee.
• Twelve distinct regression machine learning algorithms were employed.
• Model performance was evaluated based on predictive accuracy for movement changes.

Main Results:

• Ensemble Tree with LSBoost achieved the highest accuracy for predicting trunk movement alterations.
• Ensemble Tree (LSBoost) also demonstrated superior predictive precision for hip movement.
• Gaussian regression with an exponential kernel yielded the highest prediction accuracy for knee movement.

Conclusions:

• Specific regression models, including Ensemble Tree (LSBoost) and Gaussian regression, can accurately predict changes in trunk, hip, and knee movements.
• These predictive models offer potential for enhanced visualization of treatment outcomes in LBP rehabilitation.
• This pilot study highlights the utility of ML regression in quantifying and predicting rehabilitation-induced biomechanical changes.