Effect of different musculoskeletal model scaling methods on muscle force prediction for patients with cerebral palsy and equinus gait

Patient-specific musculoskeletal models are always acquired by scaling the generic ones. Different scaling methods can influence joint kinematics and affect musculotendon kinematics. The latter is an input of EMG-driven modelling and static optimization for muscle force estimation. For children with cerebral palsy (CP) and equinus gait, the ankle kinematics is a key indicator for gait classification that can be affected by scaling methods. Effects of scaling methods on muscle force estimation for such a paediatric group is not investigated yet. This study aimed at evaluating the modelling performance with two scaling methods (scaling only by static marker positions and by both static marker positions and joint angles pre-calculated from a static pose). In this study, three children with CP and equinus gait underwent standard gait analysis. Inverse kinematics, inverse dynamics, muscle analysis, static optimization and EMG-assist modelling were conducted to obtain the tibialis anterior (TA), lateral gastrocnemius (LG), medial gastrocnemius (MG) and soleus (SL) muscle forces. The coefficient of multiple correlation (CMC) and root mean squared error (RMSE) values were calculated to compare the difference between the two scaling methods. Triceps surae forces calculated by static optimization showed very good to the excellent similarity between two scaling methods. Conversely, TA force estimation seemed to be more sensitive to the scaling method chosen. For the EMG-assist modelling, LG and MG muscle forces showed a good agreement between two scaling models in contrast to SL and TA. In conclusion, TA muscle force estimation is susceptible to the scaling method irrespective of the muscle modelling approach. A possible reason may be due to different definitions of ankle joint axes and degrees of freedom. In EMG-driven modelling, SL’s mono-articular role and its optimized muscle excitation may be the reason for its sensitivity to scaling methods. Future studies should not only involve more participants and EMG channels but also apply medical imaging and other clinical assessment methods to validate the effect of scaling methods on the performance of muscle modelling approaches.