Paralysed muscle, or skeletal muscle which is to be used for cardiac assistance, may be given an artificial function if it is electrically stimulated to contract and the response can be adequately controlled. To design a controller, a model of the muscle or system is usually required. The most commonly used models are analogues, originating from A.V. Hill's model. However muscles exhibit many nonlinear and time-varying phenomena which, if they are to be modelled, make the analogue complex and cumbrous to work with. The system may further be complicated by pathological changes and secondary effects of stimulation. We propose that such a system can be modelled by nonlinear networks ('neural networks'). The radial basis function network (RBF) has two advantages over the better-known multi-layer perceptron (MLP). We describe the use of an RBF network to model rabbit muscle that is supramaximally stimulated at irregular inter-pulse intervals.
No takes yet. Share an insight, caveat, or question.
Donaldson et al. (1995) studied this question.