Validity of computational methods for estimating ground reaction forces and center of pressure during gait in individuals with subacute stroke

Ground reaction forces (GRF) and center of pressure (CoP) are essential for biomechanical gait analysis in individuals with stroke but are traditionally measured using force plates, limiting clinical accessibility. Recently developed computational methods can estimate GRF and CoP from kinematic gait data, although their validity in individuals with stroke remains unclear. This study evaluated the validity of two computational methods for estimating GRF and CoP during gait in individuals with subacute stroke. Marker trajectories and force plate data from 20 individuals with subacute stroke and 20 healthy controls were obtained from an open-access dataset. GaitDynamics and OpenGRF were used to estimate GRF and CoP from gait kinematics. Estimated GRF and CoP were evaluated using waveform similarity, magnitude error, and the accuracy of clinically relevant gait metrics. Linear mixed-effects models assessed the effects of computational method, limb group, and walking speed. For the vertical GRF, both methods met the predefined validity thresholds for waveform similarity ( r = 0.96–0.98) and magnitude error (RMSE = 6.63–9.87 %BW) across all groups. Anteroposterior estimation approached these thresholds ( r = 0.87–0.97; RMSE = 2.82–4.21 %BW), whereas mediolateral estimation remained insufficiently accurate ( r = 0.75–0.92; RMSE = 1.25–1.72 %BW). For most clinically relevant gait metrics, mean absolute errors remained below the corresponding minimal detectable change threshold. Estimation accuracy did not differ significantly between groups and was not associated with walking speed. Both methods show promise for extending kinetic gait analysis beyond traditional force plate-based laboratory settings, although further improvements in mediolateral estimation are required.