Fetal ECG extraction based on adaptive neuro-fuzzy interference system

Abstract

The aim of this paper is evaluation of the best setting options for adaptive neuro-fuzzy interference system (ANFIS) in case of fetal electrocardiogram (fECG) elicitation from two ECG signals. Thoracic ECG (tECG) signal represents maternal ECG (mECG). Abdominal ECG (aECG) signal is a mixture of mECG, fECG and additive noises (e.g. power line interference, motion artifact, ambient noise...). While additive noises can be easily eliminated by ordinary linear filters, relationship between tECG and maternal component of aECG is fully nonlinear. ANFIS is able to handle this nonlinear relationship. Quality of mECG suppression by ANFIS is affected by changing ANFIS parameters, namely number of membership functions (mf), type of mf and number of epochs. The influence of each ANFIS parameter on suppression result is described in the paper. Results of fECG filtering are assessed by signal to noise ratio (SNR) and root mean square error (RMSE). Experimental results indicate that ANFIS have the potential to improve the diagnostic and monitoring quality of fECG signals while preserving their clinically important features.