TY - GEN
T1 - Fetal ECG Extraction Using Independent Components and Characteristics Matching
AU - Alkhodari, Mohanad
AU - Rashed, Abdelrahman
AU - Alex, Meera
AU - Yeh, Nai Shyong
N1 - Funding Information:
V. ACKNOWLEDGEMENT This research is supported by the Biosciences and Bioengineering Research Institute (BBRI) at American University of Sharjah. The authors would like to thank the office of Graduate Affairs and Research at the College of Engineering for funding this work.
Publisher Copyright:
© 2018 IEEE.
PY - 2019/2/14
Y1 - 2019/2/14
N2 - In this paper, further investigations into a simpler automated use of Independent Component Analysis (ICA) in the process of Fetal ECG (FECG) extraction are performed. Extracting FECG signals through abdominal electrodes helps clinicians in diagnosing the overall health of the fetus non-invasively. In the ICA technique, FECG signals are separated from Abdominal ECG (AECG) mixtures containing maternal and noise signals. 300,000 Data samples of three AECG recordings are obtained from PhysioNet database at 1 kHz sampling frequency. Data are pre-processed through MATLAB software by centering, whitening, and filtering techniques. Then, a simpler Fast ICA algorithm is developed and used to smoothly distinguish between AECG components through automatic signal characteristics matching. Moreover, further analysis of the extracted FECG signal is performed to determine the fetus heart rate. Results successfully show efficient automatic separation between the FECG, Maternal ECG (MECG), and noise from the AECG recordings. In addition, the developed characteristics matching algorithm automatically identified the fetus signal and smoothed it to be ready for further fetal health observations. The integration of AECG signal characteristics as a prior information into the ICA algorithm promises to assist clinicians in decision making when diagnosing fetal health conditions non-invasively.
AB - In this paper, further investigations into a simpler automated use of Independent Component Analysis (ICA) in the process of Fetal ECG (FECG) extraction are performed. Extracting FECG signals through abdominal electrodes helps clinicians in diagnosing the overall health of the fetus non-invasively. In the ICA technique, FECG signals are separated from Abdominal ECG (AECG) mixtures containing maternal and noise signals. 300,000 Data samples of three AECG recordings are obtained from PhysioNet database at 1 kHz sampling frequency. Data are pre-processed through MATLAB software by centering, whitening, and filtering techniques. Then, a simpler Fast ICA algorithm is developed and used to smoothly distinguish between AECG components through automatic signal characteristics matching. Moreover, further analysis of the extracted FECG signal is performed to determine the fetus heart rate. Results successfully show efficient automatic separation between the FECG, Maternal ECG (MECG), and noise from the AECG recordings. In addition, the developed characteristics matching algorithm automatically identified the fetus signal and smoothed it to be ready for further fetal health observations. The integration of AECG signal characteristics as a prior information into the ICA algorithm promises to assist clinicians in decision making when diagnosing fetal health conditions non-invasively.
KW - Abdominal ECG (AECG)
KW - Blind Source Separation (BSS)
KW - Fetal ECG (FECG)
KW - Fetus Heart Rate (FHR)
KW - Independent Component Analysis (ICA)
KW - Maternal ECG (MECG)
UR - http://www.scopus.com/inward/record.url?scp=85063480892&partnerID=8YFLogxK
U2 - 10.1109/CSPIS.2018.8642725
DO - 10.1109/CSPIS.2018.8642725
M3 - Conference contribution
AN - SCOPUS:85063480892
T3 - 2018 International Conference on Signal Processing and Information Security, ICSPIS 2018
BT - 2018 International Conference on Signal Processing and Information Security, ICSPIS 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 International Conference on Signal Processing and Information Security, ICSPIS 2018
Y2 - 7 November 2018 through 8 November 2018
ER -