F033 Real-world walking bout and gait event detection using wearable inertial sensors in Huntington’s disease

<h3>Background</h3> Walking detection and monitoring of temporal gait parameters under free-living conditions using wearable inertial sensors could provide valuable information on gait impairment progression in Huntington’s disease (HD). A gait event detection algorithm was previously validated in HD participants [1], but the algorithm has not been deployed under free-living conditions. <h3>Aim</h3> To compare an algorithm for walking bout and gait event detection from a single leg-worn accelerometer, with reference data under free-living conditions in HD. <h3>Methods</h3> Thirteen HD participants (51 ± 10 yrs., UHDRS-TMS=25.85 ± 19.52, UHDRS gait score=0.69 ± 0.75) wore a thigh-worn (AP: ActivPAL4, 40Hz) tri-axial accelerometer and a wrist-worn (FB: Fitbit Charge 4) device for 7 days during free-living conditions. Walking bouts and gait events were detected from thigh acceleration using the Teager-Kaiser energy operator combined with unsupervised clustering. Daily time spent walking (W_t) and step count (SC) were estimated from all detected walking bouts on each day, and compared with proprietary AP data. <h3>Results</h3> Excellent agreement was observed for Wt and SC between the proposed algorithm (W_t=121.9 ± 75.4 min, SC=4032 ± 3222) and AP software (W_t=111.1 ± 76.8 min, SC=3893 ± 2740) (ICC_2,1&gt;0.975), with the presented algorithm additionally allowing stride and stance time estimation. The wrist-worn FB device overestimated SC by 34.2% as compared with thigh data. <h3>Conclusions</h3> An accurate method for walking bout detection and estimation of gait parameters has been deployed at home in HD participants using a single leg-worn accelerometer, which provides more accurate estimates of SC compared with wrist-worn devices. <h3>Reference</h3> Lozano-García M, <i>et al. J Neurol Neurosurg Psychiatry</i> 2022.