A machine learning approach to assessing gait patterns for Complex Regional Pain Syndrome

Area of Science:

• Biomedical Engineering
• Neurology
• Rehabilitation Science

Background:

• Complex Regional Pain Syndrome (CRPS) causes chronic pain and functional decline.
• Objective assessment of physical function is crucial for evaluating CRPS treatment efficacy.
• Current assessment methods may not fully capture subtle functional changes.

Purpose of the Study:

• To determine the feasibility of using accelerometer-based gait analysis for assessing physical performance in CRPS patients.
• To investigate the effectiveness of machine learning in classifying gait patterns associated with CRPS.
• To explore short-distance walking tests for objective CRPS evaluation.

Main Methods:

• Ten CRPS patients and ten healthy controls participated.
• Gait data was collected using accelerometers during short walking tests (2.4m and 20m).
• Multilayer Perceptron (MLP) neural networks were employed for gait pattern classification using extracted features.

Main Results:

• A classification accuracy of 99.38% was achieved using 3 selected features on the 2.4m test.
• An independent validation on a 20m test dataset yielded a prediction accuracy of 85.7%.
• The study identified key gait features indicative of CRPS-related functional impairment.

Conclusions:

• Accelerometer-based gait analysis is a feasible and accurate method for assessing physical function in CRPS.
• Machine learning, particularly MLP, can effectively differentiate between CRPS and healthy gait patterns.
• This approach offers a promising tool for objective CRPS therapy outcome monitoring.