The effects of gastrocnemius and soleus hyper-reflexia on the gait pattern in predictive simulations

Abstract

1. Introduction Children with cerebral palsy (CP) often exhibit complications in gait caused by neural, muscular and skeletal impairments. Treating the main source of impairment is challenging due to their multilevel nature. Predictive simulations synthesizing gait without tracking experimental data can be used to gain insight into how common neural impairments, such as hyper-reflexia, affect gait patterns. 2. Research Question How do gastrocnemius and soleus hyper-reflexia affect the gait pattern in predictive simulations? 3. Methods A generic planar OpenSim [1] model with 14 Hill-type musculotendon actuators and nine degrees of freedom was used. The reflex-based muscle control model described by Geyer and Herr [2] was implemented into SCONE (http://scone.software, version 0.20.0 [3]). In this controller, muscle excitations are a combination of constant motor signals and reflexes, based on the muscle length and muscle force and the active phase of gait. The initial pose and reflex gains were optimized by minimizing the cost of transport using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) [4]. Cost of transport was calculated and minimised using a metabolic energy cost model [5], and penalties were applied to prevent falling, and overstretching of joints. The kinematics of the predicted simulations were compared to normal gait data [6]. Three levels of hyper-reflexia were implemented for either the gastrocnemius or the soleus. A force-based reflex [7] was added with a constant value of 0.5, 1.0 or 1.5 during the whole gait cycle. After optimization, the resulting kinematics were compared to the normal gait simulations. 4. Results Simulations with normal reflex control agreed well with the normative data (Fig. 1). Gastrocnemius hyper-reflexia levels of 0.5 and 1.0 resulted in subtle changes in kinematics, mainly producing greater ankle plantar flexion. A level of 1.5 gastrocnemius hyper-reflexia had a major effect on the gait pattern, resulting in a toe-walking gait pattern. Soleus hyper-reflexia of 0.5 and 1.0 also resulted in subtle changes. Optimization with a level of 1.5 did not result in a stable gait pattern due to excessive knee extension. 5. Discussion Our predictive simulations captured some of the expected gait features due to calf muscle hyper-reflexia, including increased plantar flexion in stance and swing and an increased plantar flexion-knee extension coupling for gastrocnemius hyper-reflexia. However, surprisingly, soleus hyper-reflexia either barely affected the gait pattern or it resulted in an unstable gait pattern, since the system could not find a way to compensate for the higher level of soleus hyper-reflexia. Further studies focussing on validation against pathology and interventions are needed to assess whether further tuning of our model and controller is required. These would be the first steps towards allowing subject-specific predictive simulations to be used in clinical practise. Figure PresentedL20049408512020-02-19 <br />