Towards context-aware free-living digital biomarkers for gait assessment

Inertial Measurement Units (IMUs) have emerged as popular tools for gait related fall risk analysis due to their low cost, portability, and ability to capture spatial and temporal gait characteristics (e.g., step length and step time, respectively). Free-living analysis using IMUs is an unobtrusive means for wearers to perform daily activities without being confined to a lab setting, enabling habitual data capture. However, the lack of contextual factors in IMU data poses challenges in interpreting gait characteristics accurately. Factors such as terrain, environmental obstacles or even weather conditions can significantly influence gait characteristics and arising fall risk. To address this limitation, wearable cameras have been integrated with IMUs to provide absolute context. Wearable cameras can capture a participant’s near complete environment context, enabling researchers to examine video data with corresponding IMU data to determine cause of abnormal gait characteristics such as high step time variability. Although wearable cameras offer enhanced contextual information, concerns regarding the time-consuming process of manual review and privacy remain. Machine learning and computer vision techniques show promise in automating the review process and upholding privacy. Incorporating wearable video cameras and advanced algorithms could provide a comprehensive understanding of gait patterns and potential fall risks in free-living analysis.