Machine Learning Based Analytics for the Significance of Gait Analysis in Monitoring and Managing Lower Extremity Injuries

Aug 4, 2026

This study evaluates gait analysis as a predictive tool for post-injury complications in patients with lower extremity fractures, including infection, malunion, and hardware irritation. Supervised machine learning models are assessed for predicting complications using two consecutive gait datasets collected via a chest-mounted IMU. After preprocessing 12 gait variables and standardizing data, models are trained and tested using methods such as XGBoost, Logistic Regression, SVM, LightGBM, and Random Forest, with class imbalance handled by SMOTE. Results identify XGBoost as optimal, and feature importance highlights injury-to-first-analysis duration.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Abstract

Objective

Methods

Results

Conclusion

Introduction

Clinical challenge in lower extremity rehabilitation

Weight-bearing guidelines and gaps in evidence

Evolution of gait analysis methods

Machine Learning Based Analytics for the Significance of Gait Analysis in Monitoring and Managing Lower Extremity Injuries Mostafa Rezapour, PhD1* ; Rachel B. Seymour, PhD2; Stephen H. Sims, MD2 ; Madhav A. Karunakar, MD 2 ; Nahir Habet, MS2 ; Metin Nafi Gurcan, PhD1 1Center for Artificial Intelligence, Wake Forest School of Medicine, Winston-Salem, NC, USA 2 Department of Orthopaedic Surgery, Atrium Health Musculoskeletal Institute and Wake Forest University School of Medicine, Charlotte, NC, USA * Correspondence: Mostafa Rezapour, PhD [mrezapou@wakehealth.edu](mrezapou@wakehealth.edu) Keywords: Gait variables, Lower extremity fractures, Recovery trajectory, Machine Learning, Data analysis. Abstract Objective: This study aimed to explore the potential of gait analysis as a predictive tool for assessing postinjury complications, e.g., infection, malunion, or hardware irritation, in patients with lower extremity fractures. More precisely, the research focused on determining the proficiency of supervised machine learning models in predicting complications using two consecutive gait datasets. Methods: We prospectively identified patients with lower extremity fractures at a tertiary academic center. These patients underwent gait analysis with a chest-mounted IMU device. Using customized software, the raw gait data was preprocessed, emphasizing 12 essential gait variables. The data were standardized, and several machine learning models including XGBoost, Logistic Regression, SVM, LightGBM, and Random Forest were trained, tested, and evaluated. Special attention was given to class imbalance, addressed using SMOTE. Additionally, we introduced a novel methodology to compute the Rate of Change (ROC) for gait variables, which operates independently of the time difference between the gait analyses of different patients. Results: XGBoost was identified as the optimal model both before and after the application of SMOTE. Prior to using SMOTE, the model achieved an average test AUC of 0.90 (95% CI: [0.79, 1.00]) and an average test accuracy of 86%(95% CI: [75%, 97%]) . Through feature importance analysis, a pivotal role was attributed to the duration between the occurrence of the injury and the initial gait analysis. Data patterns over time revealed early aggressive physiological compensations, followed by stabilization phases, underscoring the importance of prompt gait analysis. Conclusion: This study emphasizes the transformative potential of using machine learning, particularly XGBoost, in gait analysis for orthopedic care. By effectively predicting post-injury complications, early gait assessment becomes pivotal, revealing timely intervention points. The findings underscore a paradigm shift in orthopedics towards a more data-informed, proactive approach, paving the way for enhanced patient outcomes. 1 Introduction Orthopedic injuries to the lower extremity are commonplace in both military and civilian contexts. The ultimate aim of surgical and rehabilitative teams is the patient's return to their prior functional level, which can be a challenging balance given multiple short-term goals and associated risks [1] . Rehabilitation, especially for the lower extremities, is an intricate process. It focuses on early movement, improving range of motion, and careful weight-bearing, but must be undertaken with caution to prevent harm to fracture sites and soft tissues. In the rehabilitation journey, patient compliance, pain management, and addressing mobility challenges are all pivotal elements. Especially in older demographics, reduced mobility can precipitate a host of health concerns ranging from physical deterioration to thromboembolic events and respiratory complications [1] . Historically, clinicians were reserved in recommending early weight-bearing for patients with lower extremity fractures, wary of complications such as implant failure [2] . However, contemporary trends suggest a paradigm shift, leaning tow

FAQs

What complications does the study aim to predict using gait analysis?

The study targets post-injury complications such as infection, malunion, and hardware irritation in lower extremity fracture patients.

How was the gait data collected and prepared for modeling?

Patients underwent gait analysis using a chest-mounted IMU; raw gait data were preprocessed to emphasize 12 essential gait variables and then standardized before training models.

Which machine learning model performed best, and how was class imbalance handled?

XGBoost was the optimal model both before and after applying SMOTE. SMOTE was used to address class imbalance during training and evaluation.

Aug 4, 2026

Upload by Lucas Martin