Recent advances in human gait analysis using artificial intelligence and machine learning: a systematic review

This systematic review synthesizes advancements in marker-based optical motion capture (MoCap) gait analysis using artificial intelligence (AI), machine learning (ML), and deep learning (DL) from 2018 to 2025. Traditional MoCap faces challenges like marker occlusion, missing data, noise, and labor-intensive processing. AI/ML/DL methods offer transformative solutions to these limitations. The review focused on applications in athletes, healthy populations, and sports rehabilitation. Following the PRISMA 2020 guidelines, major databases were searched, yielding 27 studies that met the inclusion criteria. Data extraction focused on AI methodologies, technical implementations, performance improvements, and clinical applications. Included studies demonstrated diverse AI approaches, with neural networks (22.2%) and long short term memory (LSTM)/recurrent neural network (RNN) architectures (18.5%) being the most common. Vicon systems dominated the MoCap technology market, accounting for 66.7%. Performance improvements included a reduction of up to 18% to 54% in tracking errors and over 90% classification accuracy for gait abnormalities. AI/ML/DL has significantly advanced marker-based gait analysis by providing robust solutions for handling missing data, reducing noise, and enabling automated pattern recognition. Deep learning, particularly LSTM and attention-based models, has demonstrated superior performance in handling the temporal dynamics of gait.