Background: Distal radius fractures (DRF) commonly occur in early postmenopausal females as the first fragility fracture. Although the incidence of DRF in this set of patients may be related to a lower ability to control their balance and gait, the detailed gait characteristics of DRF patients have not been examined.
Research question: Is it possible to identify the physical and gait features of DRF patients using in-shoe inertial measurement unit (IMU) sensors at various gait speeds and to develop a machine learning (ML) algorithm to estimate patients with DRF using gait?
Methods: In this cross-sectional case control study, we recruited 28 postmenopausal females with DRF as their first fragility fracture and 32 age-matched females without a history of fragility fractures. The participants underwent several physical and gait tests. In the gait performance test, the participants walked 16 m with the in-shoe IMU sensor at slower, preferred, and faster speeds. The gait parameters were calculated by the IMU, and we applied the ML technique using the extreme gradient boosting (XGBoost) algorithm to predict the presence of DRF.
Results: The fracture group showed lower hand grip strength and lower ability to change gait speed. The difference in gait parameters was mainly observed at faster speeds. The amplitude of the change in the parameters was small in the fracture group. The XGBoost model demonstrated reasonable accuracy in predicting DRFs (area under the curve: 0.740), and the most relevant variable was the stance time at a faster speed.
Significance: Gait analysis using in-shoe IMU sensors at different speeds is useful for evaluating the characteristics of DRFs. The obtained gait parameters allow the prediction of fractures using the XGBoost algorithm.
Keywords: Accidental fall; Gait analysis; Gait speed; Machine learning; Radius fracture.
Copyright © 2023. Published by Elsevier B.V.