Quantitative Assessment of Lower-Limb Multi-joint Synergy in Knee Osteoarthritis by Integrating Spatiotemporal Dual-Topology Features Under the HSMR Framework.

Abstract

To address the subjective and limited quantification in knee osteoarthritis (KOA) gait assessment, this study proposes an HSMR-based spatiotemporal dual-topology algorithm for objectively quantifying lower-limb multi-joint synergy. This algorithm utilizes the Human Skeleton and Mesh Recovery (HSMR) framework to reconstruct a biomechanically constrained Skeletal Kinematics Enveloped by a Learned body model (SKEL) parametric model and combines Qualisys for accuracy validation, enabling lower-limb joint angle extraction from monocular videos. Using data from 68 participants (32 controls, 36 KOA), temporal topology analysis quantifies single-joint complexity and stability, while a 3D spatiotemporal fusion algorithm characterizes multi-joint spatial synergy through 3D hip-knee-ankle cyclograms and projections. Based on this, compensatory changes in temporal motion sequences and variations in spatial hip-knee-ankle coordination mechanisms are objectively analyzed for KOA patients classified by Kellgren-Lawrence (KL) grades. At the temporal level, knee joint topological complexity in the KOA group decreased progressively with increasing KL grade, with KL2 and KL3-4 decreasing by approximately 7.8% and 12.8%, respectively, compared with controls; ankle joint cyclic stability in the KOA group showed an overall decline of about 35-40%. At the spatial coordination level, the swing-phase similarity score for KL3-4 decreased by 32.7% compared with controls, and the hip-knee two-dimensional projection similarity score decreased by 34.8%. The spatiotemporal dual-topology features of this algorithm can significantly distinguish KOA pathological grades, making it a high-precision and interpretable tool for digital gait analysis.