Lumbar Acceleration Gait Estimation: "Step-by-Step" Algorithm Updates and Improvements.

BACKGROUND: Digital health technologies, such as accelerometry, offer low participant burden and provide quantitative metrics with ease of deployment, making them increasingly popular for gait monitoring. Remote gait monitoring delivers quantifiable, continuous health measures over extended periods, surpassing the limited insights from single clinic or laboratory visits and offering a more comprehensive health perspective. Numerous gait algorithm implementations, inspired by prior research, aim to standardize these metrics across devices. The SciKit Digital Health (SKDH) package exemplifies this as a device-agnostic framework. OBJECTIVE: This study introduces a series of literature-informed enhancements to the SKDH gait algorithm, improving its performance against reference standards and reducing the need for manual parameter adjustments across diverse populations. METHODS: A block-wise refinement process was undertaken, examining each algorithmic component for potential enhancements and evaluating their cumulative impact on the complete gait algorithm and the metrics generated. RESULTS: Using data from healthy adult and pediatric participants, the novel gait event estimation method reduced the mean absolute error by more than 50% compared with its predecessor. Following the updates, the intraclass correlation values for final gait metric concordance with the in-laboratory reference improved markedly, from 0.50-0.74 to 0.81-0.90. Additionally, the systematic bias observed in the previous version's gait speed estimation was rectified, narrowing the difference from the reference from 0.065-0.230 to 0.00-0.03 m/s. CONCLUSIONS: The findings from this study provide robust evidence supporting the validity of the enhancements made to the gait algorithm. They demonstrate that a single lumbar accelerometer can capture gait characteristics with high accuracy and reliability across various speeds and age groups.