Evolutionary self-adaptive multimodel prediction algorithms of the fetal magnetocardiogram

Abstract

A novel technique for the analysis, nonlinear model identification and prediction of the fetal magnetocardiogram (f-MCG) is presented. f-MCGs can be recorded with the use of specific totally non-invasive superconductive quantum interference devices (SQUID). For the analysis and classification of the f-MCG signals we introduce an intelligent method that combines the following well known advanced signal processing techniques: the genetic algorithms (GA), the multimodel partitioning (MMP) theory and the extended Kalman filters (EKF). Simulations illustrate that the proposed method is selecting the correct model structure and identifies the model parameters in a sufficiently small number of iterations and tracks successfully changes in the signal, in real time. The information provided by the proposed analysis is easily interpreted and assessed by gynecologists and consist of the clinical status of the fetus. The proposed algorithm can be parallel implemented and also a VLSI implementation is feasible.