Detection of Cardiac Alternans via Statistical Analysis of Eigenvalues of a Dynamical System

Adam Petrie,Xiaopeng Zhao

2010-01-01

Abstract:Predicting cardiac alternans is a crucial step toward detection and prevention of ventricular fibrillation, a heart rhythm disorder that kills hundreds of thousands of people in the US each year. According to the theory of dynamical systems, cardiac alternans is mediated by a period-doubling bifurcation, which is associated with variations in a characteristic eigenvalue. Thus, knowing the eigenvalues would allow one to predict the onset of alternans. The existing criteria of alternans either adopt unrealistically simple assumptions and thus produce erroneous predictions or rely on complicated intrinsic functions which are not possible to measure experimentally. Unfortunately, the true state space of whatever model is appropriate for the heart is unknown. In this work, we present a maximum likelihood method to estimate the eigenvalues of a pseudostate space using a trust region method and genetic algorithms. It allows one to compute the dominant eigenvalues of a system using the time history of a single measurable variable, e.g. the transmembrane voltage or the intracellular calcium concentration in cardiac experiments. Numerical examples based on a cardiac model verify the accuracy of the method.