Predicting gait kinematics in youth with cerebral palsy using clinically informed machine learning algorithms

Area of Science:

• Pediatric Orthopedics
• Rehabilitation Medicine
• Biomechanical Engineering

Background:

• Gait deviations are prevalent in youth with Cerebral Palsy (CP), affecting their development.
• Individual and treatment factors influence changes in gait patterns during growth.
• Predicting gait kinematics changes is crucial for effective management.

Purpose of the Study:

• To employ Machine Learning (ML) to predict gait kinematics changes in youth with CP.
• To evaluate the relationship between patient and treatment factors and gait alterations.
• To develop a predictive framework for clinical decision support.

Main Methods:

• Collected kinematic gait data from ambulatory youth with spastic CP (GMFCS I-III) using instrumented gait analysis (IGA).
• Trained Gradient Boosting Regressor (GBR) models to predict changes in Gait Profile Score (GPS) and Gait Variable Scores (GVS).
• Utilized patient factors (age, motor function) and treatment factors (orthopedic surgery) as predictors.

Main Results:

• Analyzed 702 evaluation pairs from youth with GMFCS levels I, II, and III.
• Younger participants (<11 years) and those undergoing orthopedic surgery showed greater GPS improvement.
• Baseline gait pattern, motor function, and trunk/pelvis positioning influenced gait changes.

Conclusions:

• Gait kinematics improvement in youth with CP is significantly associated with age, orthopedic surgery, and baseline factors.
• The developed ML framework can enhance clinical decision support for predicting gait changes.
• Findings aid providers in optimizing treatment strategies for ambulatory youth with CP.