Accurate Gait Assessment and Reduced Patient Burden from a Chest-Mounted Accelerometer

Gait analysis is increasingly being used in clinical trials and routine clinical care to characterize mobility and impairment in those with neurological disorders and other mobility impairments. Gait assessment methods have been approved as primary endpoint in a clinical trial, elucidating the need to minimize participant burden while ensuring accurate results. Traditional methods often rely on cumbersome equipment, limiting their use in everyday environments. In this study, we explore the potential of chest acceleration data for estimating gait parameters, leveraging wavelet analysis and machine learning techniques. The chest location offers several advantages: it is minimally obtrusive, can capture whole-body dynamics, and allows for continuous monitoring during daily activities. Our stride event segmentation approach achieves low error rates for stride time and stance time compared to motion capture (RMSE of 0.043 seconds and 0.067 seconds, respectively). Additionally, we found a deep learning model achieved an RMSE of 0.117 meters in estimating stride length and a less computationally expensive generalized inverted pendulum model achieved an RMSE = 0.132 meters. This chest-based method successfully estimated other gait parameters such as duty factor and double support duration, although performance was more varied. These findings suggest that chest acceleration data can provide an efficient and non-invasive method for gait analysis, which could enhance the utility of wearable devices in clinical trials and routine monitoring.Clinical Relevance-Validation of these novel algorithms support their use for quantifying gait and mobility impairment while reducing patient burden and allowing simultaneous measurement of important cardiac signals.