一种融合KPCA、FastICA及SVD的腹壁源胎儿心电信号提取算法研究OACSTPCD

Objective To propose an algorithm to extract fetal ECG signals from mixed signals of maternal abdominal wall with high signal-to-noise ratio and clear waveforms by combining kernel principal component analysis(KPCA),fast independent component analysis(FastICA)and singular value decomposition(SVD).Methods Firstly,KPCA was used to downscale the maternal ECG signals,and then the improved negative entropy-based FastICA was applied to processing the downscaled data to obtain the independent components.Subsequently,sample entropy was introduced for signal channel selection,and the signal channel containing the most maternal information was selected.SVD was performed on the selected maternal channel to get an approximate estimate of the maternal ECG signals,which was then subtracted from the abdominal wall source signals to obtain a preliminary estimate of the fetal ECGs.Finally,the pure fetal ECG signals were successfully separated using a modified negentropy-based FastICA.The proposed algorithm was validated in the Abdominal and Direct Fetal Electrocardiogram Database(ADFECGDB)and the PhysioNet 2013 Challenge database.Results The proposed algorithm gained advantages in both subjective visualization and objective evaluation metrics,which had the sensitivity,positive predictive value and F1 value of fetal QRS compound wave respectively being 99.74％,98.85％and 99.30％for the ADFECGDB database,and 99.10％,97.87％and 98.48％for the PhysioNet 2013 Challenge database.Conclusion The fetal ECG signal extraction algorithm incorporating KPCA,FastICA and SVD effectively handles the additional noise while extracting fetal ECG signals,which provides strong support for the early diagnosis of fetal diseases.[Chinese Medical Equipment Journal,2024,45(7):1-7]