A model of muscle spasticity in opensim

Abstract

Introduction: Computer simulations of human movement are commonly used to study normal and abnormal gait, for instance of subjects with spasticity. However, they generally do not include explicit models of spasticity. Our goal, therefore, was to develop a computer model of spasticity and to test this model by dynamic simulation of instrumented spasticity tests. Patients/materials and methods: Spasticity was modelled in OpenSim [1,2] using a controller plug-in. This controller excites any muscle towhich it is assignedwhen themuscle fibres are stretched faster than a specified “threshold” velocity. Following Lance [3], the spastic excitation was dependent only on fibre stretch velocity, which was multiplied by a constant “gain”. A “delay” was used to represent the stretch reflex latency. To test the spasticity model, forward dynamic simulations were created of instrumented spasticity test data of the hamstrings [4], collected on a female subject with cerebral palsy (15 years, 1.60m, 52kg). The knee was manually extended at three different velocities with the subject supine and the hip fixed. The force exerted on the shank, the knee angle, and EMG was measured. The scaled musculoskeletal model was put into the position measured at the beginning of the test and the measured forces were imposed on the model. The model was run forward in time, and the resulting movements, muscle length and velocity, and muscle activation were evaluated. Simulations were runwithout andwith spasticity, simulatedwith different values for gain, threshold, and delay. Results:Without spasticity, the simulation overestimated knee extension, muscle activation was zero, and the muscle fibres extendedsmoothly (see Figure).Whenadding the spastic controller to the semitendinosusmuscle, themuscle reacted to the stretch by becoming active, resulting in lower peak knee angle, slower fibre lengthening velocity, and lower peak fibre length. On close examination, an abrupt stop (‘catch’) was seen in the fibre length change (Fig. C, zoom). With increasing gain, a more spiky pattern was seen in muscle activation (Fig. B), resembling a clonus effect. Increasing threshold and delaymoderated the effects of the spastic controller. The experimental knee angle could be reproduced well by optimizing the model’s parameters, but only if muscle contracture as present in the subject was included. Discussion & conclusions: With a spasticity model dependent only on muscle fiber velocity, spastic behaviour can be predicted, including a sudden increase in muscle activity, a break (‘catch’) in the stretch of the muscle, and clonus. Prolonged muscle activity beyond the duration of the stretch, as often observed in patients with spasticity including our test subject (Fig. B, black line) was not predicted by our model. Future steps include testing the model on subjects with different levels of spasticity and applying the model to evaluate the effects of spasticity on the performance of dynamic tasks such as gait.