Estimating Muscle Forces in Patients With Cerebral Palsy During Walking Using Static Optimization and Computed Muscle Control

Cerebral palsy (CP) is a group of neurological disorders that presents significant challenges for clinical rehabilitation. While muscle forces could aid clinical decision-making, direct in vivo measurement is infeasible and ethically questionable. Consequently, model-based methods such as static optimization (SO) and computed muscle control (CMC) have gained attention. Although SO and CMC have been compared for healthy individuals, it remains uncertain whether one approach yields more accurate predictions across varying severities of crouch gait in CP. We evaluated SO and CMC using opensim to estimate muscle forces and activations from an openly available dataset with delineations based on crouch severity. Predicted muscle activations were validated against experimental electromyography data using Spearman's rank correlation coefficients (ρ) and root-mean-squared error (RMSE), while joint moment tracking was assessed using reserve moments. A sensitivity analysis was conducted to examine the influence of tendon slack length on force predictions. Results showed that while CMC predicted generally higher muscle forces than SO, both methods yielded variable ρ values (-0.7 to 0.9) and RMSEs (0.14 to 0.7) across muscle groups and crouch severities. ρSO was significantly higher than ρCMC for the medial hamstrings, and crouch severity significantly influenced the ρ difference between methods for the lateral hamstrings and rectus femoris. However, RMSEs did not consistently reflect these trends. CMC was more sensitive to tendon slack length variations. Overall, neither method currently provides sufficiently validated muscle force estimates for clinical application in CP, emphasizing the need for further methodological refinement.