Nonlinear open-loop gain of the baroreflex using artificial feedforward neural networks

Abstract

The baroreflex control system is inherently nonlinear. Clinical measurements which rely on linear models provide adequate representation of the system as long as the input perturbations to the reflex loop are small and fall within the linear region of the response curve. The authors propose a novel technique that combines approximation power of a class of artificial neural networks (ANNs), Volterra nonlinear block representation and eigen-analysis to provide estimates of the open-loop gain of the baroreflex. A range of eigen-parameters are extracted from the converged weight matrices of the ANN to provide a range of possible values of the gain factor of simulated baroreflex response curve.