Srovnání algoritmů založených na analýze nezávislých komponent při zpracování plodového elektrokardiogramu

Abstract

The aim of this bachelor thesis is the processing of non-invasive fetal electrocardiogram (fECG) using algorithms based on independent component analysis (ICA) and subsequent evaluation of the quality of fECG extraction using these algorithms. These algorithms are: fast independent component analysis (Fast ICA), joint approximation diagonalization of Eigen-matrices (JADE ICA), flexible independent components analysis (Flex ICA) robust accurate and direct independent components analysis (Radical), robust independent components analysis (Robust ICA) , second order blind identification (SOBI), information maximization of independent components analysis (Infomax), equivalent robust independent components analysis (ERICA), curtosis maximization of independent component analysis (kICA), simultaneous blind signal extraction (SIMBEC) and algorithm for extraction of multiple unknown sources (Amuse). The study was performed on real data from the extended abdominal and direct fetal electrocardiogram database (ADFECGDB). Evaluation of extraction quality is evaluated by determining the overall probability of correct detection (ACC), sensitivity (SE), positive predictive value (PPV) and harmonic mean between SE and PPV (F1). The best results were obtained in the experiment by the Flex ICA algorithm with ACC > 80 % in 8 of 12 records.