Loading...

Fetal R Detection from Mixed Maternal and Fetal MCG Signals

Kharabian Masouleh, Shahrzad | 2009

883 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 40255 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Shamsollahi, Mohammad Bagher; Sameni, Reza
  7. Abstract:
  8. Analyzing cardiac function of the fetus during pregnancy is proved to be an important prenatal care procedure. Traditional methods like auscultation and ultrasonography could only lead to anatomical information about the fetal heart. So in the recent decades many researches on the abdominal electrical signals of the pregnant women have been done. Nowadays, it is possible to record the heart magnetic signals. With regard to the morphological similarity between the electrical and magnetical signals of the heart and the superiority of the magnetic ones, one could assume more diagnostic capacity for the fetal MCG. It should be mentioned that finding the location of the fetal R waves could help us in the diagnosis of many of the heart arrhythmias. Also it could be used as the first step in many of the fetal complete signal extraction procedures. This project aims to detect the fetal R waves from the MCG multichannel abdominal signals in highly noisy situation or with the fetus movements. In order to quantitatively evaluating the fetal R-peak detector algorithms, a simulated MCG/ECG data set is used and then fetal MCG signals from the Friedrich Schiller university (Jena-Germany) are utilized as the real data. In this thesis two different approaches for the fetal R detection problem is discussed. In the first approach, primarily the maternal MCG signals are eliminated with the use of ICA or πCA and then in order to localize the fetal R peaks on the output signals, a method which is not sensitive to the fetus movements is used. Also, to achieve better results for the real MCG data, a dimensionality reduction procedure is performed using PCA combined with the Kurtosis criterion. The results in this part are shown to be good. The next part discuses the methods which do not need complete maternal signal removal and are only based on the maternal QRS elimination and the detection the fetal R-peaks. These methods are working very well on the simulated data set as we were able to achieve the promising results of the detection error rate less than 6% and the sensitivity above 99% for the abdominal signals with SNR=-10dB.
  9. Keywords:
  10. Blind Sources Separation (BSS) ; Independent Component Analysis (ICA) ; Principal Component Analysis (PCA) ; Fetal Magnetocardiogram ; R-Peak Detection ; Hilbert Transform ; Multiscale Principal Components Analysis (MSPCA)

 Digital Object List

 Bookmark