Neuromuscular Control During Tai Chi in Individuals With Knee Osteoarthritis: Broadly Preserved Muscle Synergy Architecture With Selective Alterationsin Functional Recruitment

Abstract Background: Although Tai Chi is widely recommended for knee osteoarthritis (KOA), its muscle-level control, group differences and the movements that best reveal them remain unclear. Using muscle synergies extracted by non-negative matrix factorization (NMF), we identified Tai Chi's key lower-limb muscles, tested whether synergy architecture or recruitment differed between groups and ranked movements for assessment research.Methods: Bilateral surface EMG from 16 lower-limb muscles was recorded in 69 participants (52 KOA and 17 healthy controls) during eight-form Tai Chi (17 movement-by-side conditions). Synergies were extracted by shared-basis consensus NMF and evaluated using five measures: synergy number, muscle weights, whole-vector similarity, functional contribution (each muscle's share of reconstructed EMG), and activation profiles (cluster-based permutation testing). Contribution compositions were centred-log-ratio transformed and compared using studentised permutation tests with global Benjamini--Hochberg correction. Other statistical tests used predefined within-family adjustments for multiple comparisons. The measures formed a descriptive composite index for ranking movements.Results: Reconstructed EMG contribution was dominated by the quadriceps (≈ 49%), assisted by hip and posterior-chain muscles, forming six bilaterally symmetric synergy modules. Synergy architecture showed limited KOA--HC differences: synergy number, overall similarity and muscle weights differed minimally. Differences concentrated in recruitment rather than structure: 11/272 contribution contrasts survived global correction, with KOA showing higher gluteus medius contribution in two conditions and controls higher contribution from the tibialis anterior, vastus medialis, biceps femoris, medial gastrocnemius and rectus femoris in selected conditions. Activation profiles differed in 26/99 synergies (34 cluster-corrected windows, 24 KOA-higher), most numerous in Repulse Monkey but longest in Cloud Hands. The integrated index ranked Heel Kick and Repulse Monkey as most KOA-divergent, which the activation-profile measure alone would have missed.Conclusions: Tai Chi recruits a compact set of key lower-limb muscles organised into six bilaterally paired synergy modules whose architecture was broadly similar between groups. KOA was associated with selective, movement-specific differences in relative contribution and activation profiles within these largely shared synergies. No single measure captured every affected movement, so a descriptive integrated index ranked candidates. These muscles and movements are candidates for task-specific motor-control training and compact KOA assessment batteries, warranting prospective evaluation.