Can velocity increment derived from a single, lumbar-mounted inertial measurement unit serve as a surrogate for propulsion force impulse in healthy and post-stroke individuals?

Background Paretic propulsion is challenging to measure outside the laboratory. Trunk anterior acceleration reflects late-stance velocity increment, potentially corresponding to the net propulsion force generated by the legs. We evaluated whether anterior velocity increment from a lumbar-mounted inertial measurement unit (IMU) can serve as a surrogate for propulsion force impulse in healthy and post-stroke individuals. Methods Seventeen healthy and thirteen post-stroke individuals completed a 2-min self-paced treadmill walking trial. Step-by-step anterior velocity increments were derived from lumbar IMU acceleration and compared with propulsion force impulses obtained from treadmill-embedded force plates. Group-level and individual-level associations (R 2 ) between lumbar velocity increment and propulsion force impulse, and root mean squared errors (RMSE) were calculated using linear mixed models. To test whether step aggregation improves validity, we compared individual-level R 2 and RMSE across 1-, 3-, and 5-step conditions using Friedman tests. Findings Group-level associations were significant but weak (marginal R 2 : healthy = 0.32; paretic leg = 0.15; non-paretic leg = 0.05), with corresponding RMSE values of 0.061, 0.073, and 0.064 N·s/kg. Median [IQR] individual-level R 2 was 0.42 [0.34–0.48] for healthy individuals, 0.19 [0.02–0.52] for the paretic leg, and 0.03 [0.02–0.25] for the non-paretic leg, and RMSE values were 0.022 [0.020–0.025], 0.030 [0.023–0.035], 0.038 [0.027–0.042], respectively. Step aggregation did not affect R 2 but significantly reduced RMSE. Interpretation Lumbar anterior velocity increment showed poor validity for estimating propulsion force impulse in both healthy and post-stroke individuals, suggesting that a single, lumbar-mounted IMU is insufficient for accurate assessment of propulsion force.