Zkoumání vlivu vzorkovací frekvence na kvalitu monitorování tepové frekvence plodu

Abstract

This thesis focuses on non-invasive fetal monitoring using fetal electrocardiography (fECG), with the main emphasis placed on the impact of changes in sampling frequency on the quality and accuracy of R-peak detection. As part of the practical work, a software application was developed in MATLAB App Designer to enable the extraction of fECG, detection of R-peaks, and evaluation of results based on comparison with reference data. The experimental study employed the template subtraction method (TS) and its modifications (TSSVD, TSLP, TSSF, TSSA) combined with two types of detectors based on continuous wavelet transform (CWT) and deep neural networks (DNN). The research was conducted using two publicly available datasets (labour, pregnancy) published by Wrobel et al. [1], with reference annotations verified by clinical experts. The results of the experiments demonstrated that decreasing the sampling frequency leads to a decline in detection accuracy, with limit frequencies identified at 320 Hz for the CWT detector and 370–380 Hz for the DNN detector. The thesis highlights that the appropriate selection of the extraction method together with a sufficient sampling frequency is crucial for reliable fetal ECG monitoring.