Optimization of adaptive filter control parameters for non-invasive fetal electrocardiogram extraction

• Published in: PloS one Vol. 17; no. 4; p. e0266807
• Main Authors: Kahankova, Radana, Mikolasova, Martina, Martinek, Radek
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 11.04.2022; Public Library of Science (PLoS)
• Subjects: Abdomen; Adaptive algorithms; Adaptive control; Adaptive filters; Algorithms; Analysis; Automation; Biomedical engineering; Clinical medicine; Computer and Information Sciences; Computer engineering; Computer science; Cybernetics; EKG; Electrical engineering; Electrocardiogram; Electrocardiography; Electrocardiography - methods; Electrodes; Engineering and Technology; Evaluation; Experiments; Female; Fetal monitoring; Fetal Monitoring - methods; Fetus - physiology; Fetuses; Heart rate; Humans; Hybrid systems; Hypoxia; Information processing; Least mean squares; Least mean squares algorithm; Least-Squares Analysis; Methods; Morphology; Noise; Optimization; Parameters; Physical Sciences; Pregnancy; Research and Analysis Methods; Signal processing; Signal Processing, Computer-Assisted; Signal quality; Transversal filters; Czech Republic
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0266807

This paper is focused on the design, implementation and verification of a novel method for the optimization of the control parameters of different hybrid systems used for non-invasive fetal electrocardiogram (fECG) extraction. The tested hybrid systems consist of two different blocks, first for maternal component estimation and second, so-called adaptive block, for maternal component suppression by means of an adaptive algorithm (AA). Herein, we tested and optimized four different AAs: Adaptive Linear Neuron (ADALINE), Standard Least Mean Squares (LMS), Sign-Error LMS, Standard Recursive Least Squares (RLS), and Fast Transversal Filter (FTF). The main criterion for optimal parameter selection was the F1 parameter. We conducted experiments using real signals from publicly available databases and those acquired by our own measurements. Our optimization method enabled us to find the corresponding optimal settings for individual adaptive block of all tested hybrid systems which improves achieved results. These improvements in turn could lead to a more accurate fetal heart rate monitoring and detection of fetal hypoxia. Consequently, our approach could offer the potential to be used in clinical practice to find optimal adaptive filter settings for extracting high quality fetal ECG signals for further processing and analysis, opening new diagnostic possibilities of non-invasive fetal electrocardiography.