Fetal phonocardiogram extraction using single channel blind source separation

Frequent and long-term monitoring of fetal health status is still a challenging task in high-risk pregnancies. This paper presents a fully noninvasive method to extract fetal heart sound (FHS) from acoustic signals recorded from the maternal abdominal surface. The proposed algorithm is based on single channel blind source separation (SCBSS), which utilizes empirical mode decomposition (EMD) and nonnegative matrix factorization (NMF) to extracts different sources from audio signal mixtures. To evaluate the algorithm performance, a dataset of 110 pregnant women were collected at Hafez Hospital of Shiraz University of Medical Sciences. The algorithm performance has been tested on 50 randomly selected sample data of recording signals. The results show very good separation of the fetal heart sound. Quantitative results show that the fetal heart rate (FHR) calculated from the processed channel corresponding to the fetal phonocardiogram (fPCG) has an accuracy of 96% as compared to the simultaneously recorded FHS by means of ultrasound.

[1]  Hyunsoo Kim,et al.  Sparse Non-negative Matrix Factorizations via Alternating Non-negativity-constrained Least Squares , 2006 .

[2]  Ferenc Kovács,et al.  A rule-based phonocardiographic method for long-term fetal heart rate monitoring , 2000, IEEE Transactions on Biomedical Engineering.

[3]  F. Stanley,et al.  Effects of frequent ultrasound during pregnancy: a randomised controlled trial , 1993, The Lancet.

[4]  Chris H. Q. Ding,et al.  Convex and Semi-Nonnegative Matrix Factorizations , 2010, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  Keikichi Hirose,et al.  Single-Mixture Audio Source Separation by Subspace Decomposition of Hilbert Spectrum , 2007, IEEE Transactions on Audio, Speech, and Language Processing.

[6]  Gabriel Rilling,et al.  On empirical mode decomposition and its algorithms , 2003 .

[7]  Alioune Ngom,et al.  Fast sparse representation approaches for the classification of high-dimensional biological data , 2012, BIBM.

[8]  G. Clifford,et al.  A Review of Fetal ECG Signal Processing; Issues and Promising Directions. , 2010, The open pacing, electrophysiology & therapy journal.

[9]  Wei Liu,et al.  Blind Second-Order Source Extraction of Instantaneous Noisy Mixtures , 2006, IEEE Transactions on Circuits and Systems II: Express Briefs.

[10]  Gian Luca Romani,et al.  Fetal cardiac time intervals: validation of an automatic tool for beat-to-beat detection on fetal magnetocardiograms , 2005 .

[11]  O Axelsson,et al.  Ultrasound and adverse effects , 2002, Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology.

[12]  Zoltán Benyó,et al.  An advanced method in fetal phonocardiography , 2003, Comput. Methods Programs Biomed..

[13]  Maria Romano,et al.  An algorithm for FHR estimation from foetal phonocardiographic signals , 2010, Biomed. Signal Process. Control..

[14]  H. Sebastian Seung,et al.  Learning the parts of objects by non-negative matrix factorization , 1999, Nature.

[15]  Anupam Shukla,et al.  System simulation and comparative analysis of foetal heart sound de-noising techniques for advanced phonocardiography , 2007 .

[16]  Hyunsoo Kim,et al.  Nonnegative Matrix Factorization Based on Alternating Nonnegativity Constrained Least Squares and Active Set Method , 2008, SIAM J. Matrix Anal. Appl..

[17]  Alioune Ngom,et al.  A new Kernel non-negative matrix factorization and its application in microarray data analysis , 2012, 2012 IEEE Symposium on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB).

[18]  Bin Gao,et al.  Single channel blind source separation , 2011 .

[19]  Emmanuel Vincent,et al.  Enforcing Harmonicity and Smoothness in Bayesian Non-Negative Matrix Factorization Applied to Polyphonic Music Transcription , 2010, IEEE Transactions on Audio, Speech, and Language Processing.

[20]  S. Emery,et al.  Dynamic analysis of beat-to-beat fetal heart rate variability recorded by SQUID magnetometer: quantification of sympatho-vagal balance. , 2002, Early human development.