A Multicenter Standardized Gait Database from the ORITEL Network for Cerebral Palsy Gait Analysis.

Area of Science:

• Biomedical Engineering
• Clinical Biomechanics
• Computational Neuroscience

Background:

• Gait analysis is crucial for cerebral palsy (CP) management.
• Multicenter data variability hinders clinical decision-making and research translation.
• Standardized data and analytical tools are needed for reliable CP gait assessment.

Purpose of the Study:

• To create a standardized, multicenter gait database for CP.
• To develop and validate a machine learning (ML) classifier for distinguishing CP gait patterns (hemiplegia, diplegia) from typically developing individuals.
• To establish a baseline for data-driven rehabilitation planning in CP.

Main Methods:

• Unified data processing (nomenclature, normalization, resampling) across eight ORITEL labs using Vicon/BTS systems.
• Feature engineering combining kinematic, spatiotemporal, and anthropometric variables.
• Machine learning classification using a multilayer perceptron (MLP) with techniques to address class imbalance and dimensionality reduction (PCA).

Main Results:

• A standardized multicenter gait database was successfully created.
• The MLP classifier achieved 82% test accuracy, with high F1-scores for diplegia (0.81) and controls (0.98), and 0.63 for hemiplegia.
• Key discriminants identified include double support, stride length, and knee flexion, offering clinical interpretability.

Conclusions:

• The developed resource and ML pipeline advance multicenter data standardization for CP gait analysis.
• This standardized approach provides a clinically grounded baseline for supporting rehabilitation planning.
• The findings suggest distinct gait patterns in hemiplegia and diplegia, aligning with treatment recommendations.