Machine Learning-Based Peripheral Artery Disease Identification Using Laboratory-Based Gait Data.

Area of Science:

• Biomechanics
• Machine Learning
• Medical Diagnostics

Background:

• Peripheral artery disease (PAD) stems from atherosclerosis, impairing leg blood flow and altering muscle function and gait.
• Underdiagnosis of PAD delays treatment, leading to poorer clinical outcomes.

Purpose of the Study:

• To develop machine learning (ML) models for early identification of individuals with PAD.
• To establish ML as a tool for distinguishing PAD patients from healthy individuals.

Main Methods:

• Utilized overground walking biomechanics data from PAD patients and healthy controls.
• Gait signatures analyzed included joint angles, torques, powers, and ground reaction forces (GRF).
• Developed classification models using Neural Networks and Random Forest algorithms.

Main Results:

• ML models achieved 89% accuracy in classifying PAD using all gait variables.
• Models using only GRF variables reached 87% accuracy.
• GRF-based ML models provided the most informative data for PAD classification.

Conclusions:

• ML models can effectively classify individuals with and without PAD based on gait signatures.
• Gait analysis combined with ML shows promise for early PAD detection.
• GRF features are particularly valuable for ML-based PAD classification.