Scholarlink

<p>Fetal electrocardiogram (fECG) helps identifying, analyzing, and monitoring various congenital heart diseases (CHD) in developing fetal heart during the gestation period along with uterine contraction. It is recorded using invasive and non-invasive methods, among which the latter one is preferred due to low risk and convenience. Extraction of clean fECG signal by non-invasive method is crucial due to its interference with maternal electrocardiogram (mECG), power line disturbances, baseline wander, motion artifact, uterine contraction, and high frequency noises etc. FECG is prone to signal loss during recording that leads to false diagnosis. The objective of this research study is to present a resilient methodology, consisting of empirical mode decomposition (EMD) and wavelet decomposition (WD) for clear and accurate fECG extraction from abdominal ECG (aECG) consisting of fECG, mECG, and noises. EMD helps decomposing non-stationary, non-linear, and quasi-periodic signals like ECG into intrinsic mode functions (IMFs) using the signal itself as the basis. This identifies the various frequency components present in aECG. The IMFs are selected on the basis of frequency range of mECG and fECG signal. Selective IMFs are combined with residual to generate the data matrix, which is subjected to N-level WD (with a suitable mother wavelet) for identifying various frequency components in different sub-bands. Selective wavelet sub-bands are used to obtain mECG, followed by fECG identification. The extracted mECG and fECG are subjected to filtering by Savitzky-Golay- followed by 3rd order band-pass-, derivative-, and P-point moving average-filter for clear demarcation of the fiducial points for identification of R-peaks by Pan-Tompkins algorithm followed by computation of heart rate variability (HRV). This work has been validated on publicly available DaISy database and reports an accuracy of 100% (mECG extraction) &amp; 88.9% (fECG extraction) and F1 score of 100% (mECG peaks) &amp; 88.9% (fECG peaks). HRV for mECG and fECG is found to be 81 bpm and 132 bpm, respectively for DaISy database. The results of this algorithm are compared with other state-of art method available in literature, such as, template subtraction, adaptive filtering, blind source separation, and AI techniques.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Approach for Detecting Fetal QRS and Estimating Fetal Heart Rate from Abdominal ECG using EMD and Wavelet Decomposition

  • 期刊论文

  • 发表日期：

    2025年7月14日

查看全文
  • Sanghamitra Subhadarsini Dash,
  • Malaya Kumar Nath

摘要

Fetal electrocardiogram (fECG) helps identifying, analyzing, and monitoring various congenital heart diseases (CHD) in developing fetal heart during the gestation period along with uterine contraction. It is recorded using invasive and non-invasive methods, among which the latter one is preferred due to low risk and convenience. Extraction of clean fECG signal by non-invasive method is crucial due to its interference with maternal electrocardiogram (mECG), power line disturbances, baseline wander, motion artifact, uterine contraction, and high frequency noises etc. FECG is prone to signal loss during recording that leads to false diagnosis. The objective of this research study is to present a resilient methodology, consisting of empirical mode decomposition (EMD) and wavelet decomposition (WD) for clear and accurate fECG extraction from abdominal ECG (aECG) consisting of fECG, mECG, and noises. EMD helps decomposing non-stationary, non-linear, and quasi-periodic signals like ECG into intrinsic mode functions (IMFs) using the signal itself as the basis. This identifies the various frequency components present in aECG. The IMFs are selected on the basis of frequency range of mECG and fECG signal. Selective IMFs are combined with residual to generate the data matrix, which is subjected to N-level WD (with a suitable mother wavelet) for identifying various frequency components in different sub-bands. Selective wavelet sub-bands are used to obtain mECG, followed by fECG identification. The extracted mECG and fECG are subjected to filtering by Savitzky-Golay- followed by 3rd order band-pass-, derivative-, and P-point moving average-filter for clear demarcation of the fiducial points for identification of R-peaks by Pan-Tompkins algorithm followed by computation of heart rate variability (HRV). This work has been validated on publicly available DaISy database and reports an accuracy of 100% (mECG extraction) & 88.9% (fECG extraction) and F1 score of 100% (mECG peaks) & 88.9% (fECG peaks). HRV for mECG and fECG is found to be 81 bpm and 132 bpm, respectively for DaISy database. The results of this algorithm are compared with other state-of art method available in literature, such as, template subtraction, adaptive filtering, blind source separation, and AI techniques.