Energy-Dependent RLS Architecture for the Separation of Fetal ECG Using Thoracic and Abdominal Lead ECG of Mother

• Published in: IEEE transactions on instrumentation and measurement Vol. 72; pp. 1 - 12
• Main Authors: Dhas, D. Edwin, Suchetha, M.
• Format: Journal Article
• Language: English
• Published: New York IEEE 2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Abdomen; Abdominal channels; adaptive filter; Adaptive filters; Correlation coefficients; Datasets; Electrocardiography; Energy; fetal electrocardiogram (FECG); Field programmable gate arrays; field-programmable gate array (FPGA); Filtering algorithms; Filtration; Finite impulse response filters; Hardware; recursive least-square (RLS) filter; Separation; Signal to noise ratio; Synthetic data; Thorax
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2023.3328093

This article proposes a fetal electrocardiogram (FECG) separation approach based on an energy-dependent recursive least-square (RLS) filtering approach that uses the mother's R-peaks collected from both the abdomen and the thorax. This approach initially identifies the mother's R-peaks from the thorax electrocardiogram (ECG), which is used to represent the mother's R-peaks in both the abdominal and thorax channels. Instead of using the recent abdominal and thorax ECG (TECG) samples, the proposed filter also considers the energy of <inline-formula> <tex-math notation="LaTeX">L_{1} </tex-math></inline-formula> number of mother's past R-peak abdominal and thorax samples along with the energy of <inline-formula> <tex-math notation="LaTeX">L_{2} </tex-math></inline-formula> number of non-R-peak abdominal samples for estimating the R-peak energy factor. The energy factor is estimated for each sample for the updation of weights in the RLS filter. An architecture for the filter is also proposed, which can be used in hardware implementation. The evaluation of the proposed filtering approach was performed using datasets such as Synthetic and Daisy with the evaluation metrics, namely, correlation coefficient, fetal R-peak detection accuracy (PDA), fetal-to-maternal signal-to-noise ratio (SNR), and percent root-mean-square difference. With filter length <inline-formula> <tex-math notation="LaTeX">P=24 </tex-math></inline-formula>, the proposed filter results in correlation, SNR, and percent root-mean-square difference of 0.9901, 9.03 dB, and 80.84%, respectively. For the Daisy and Synthetic datasets, the PDA was estimated as 96.4% and 98.12% respectively. The architecture of the proposed filter was implemented in Virtex VC707 hardware, which utilizes a power of 1.378 W, resulting in a maximum clock frequency and throughput (TP) of 128.43 MHz and <inline-formula> <tex-math notation="LaTeX">31.5~\text {Mb/s} </tex-math></inline-formula>, respectively, with a wordlength of <inline-formula> <tex-math notation="LaTeX">L=24 </tex-math></inline-formula> bits.