Utilizing machine learning to analyze trunk movement patterns in women with postpartum low back pain

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Data Science

Background:

• Postnatal low back pain (LBP) affects many women, impacting their quality of life.
• Understanding the biomechanical factors contributing to LBP is crucial for effective intervention.
• Existing methods for LBP assessment may not fully capture dynamic trunk movement patterns.

Purpose of the Study:

• To analyze trunk movement in women with postnatal LBP using machine learning.
• To identify key biomechanical features predictive of postnatal LBP.
• To develop accurate machine learning models for LBP prediction and classification.

Main Methods:

• Applied regression and classification algorithms to trunk movement data from 100 postpartum women (50 with LBP, 50 without).
• Utilized techniques such as Optimized optuna Regressor, Basic CNN, and Random Forest Classifier.
• Performed feature selection to identify influential biomechanical factors.

Main Results:

• Optimized optuna Regressor achieved high accuracy in regression (MSE < 0.0003, R2 > 0.99).
• Basic CNN and Random Forest Classifier demonstrated near-perfect classification accuracy (accuracy, AUC, precision, recall, F1-score = 1.0).
• Significant predictive features included pain, range of motion (flexion, extension), and average movement.

Conclusions:

• Machine learning provides a powerful tool for analyzing trunk biomechanics in postnatal LBP.
• Identified key features offer quantitative insights for risk assessment and treatment personalization.
• This approach holds potential for improving LBP diagnosis and patient outcomes.