Evaluation of Non-Invasive Foetal ECG Extraction Algorithms using Non-Stationary Mixtures. Andreotti, F., Behar, J., Zaunseder, S, Clifford, G D, & Oster, J In Congresso Brasileiro de Engenharia Biomedica, pages 1585—-1588, Uberlandia, Brazil, 2014.
abstract   bibtex   
Abstract: The non-invasive foetal electrocardiogram obtained from abdominal electrodes is an alternative for assessing the foetal health state. This study evaluates the behavior of two well-known methods (Independent Component Analysis and Template Subtraction). These techniques are often referred to in literature for extracting the FECG from the abdominal mixture. Our analysis quantifies the extraction results in terms of accuracy of foetal peak detection using the F1 measure. 1750 five-minute-long simulations considering the absence and presence of non-stationarity mixtures served as dataset for this study. In terms of the F1 measure, on average ICA (97.3 ± 0.1 %) was found to outperform TS (88.1 ± 0.2 %). Moreover, we have shown that the presence of non-stationarities can degrade the performance of both methods considerably, indicating a beat-by-beat adaptive approach may be preferable.
@inproceedings{andreotti2014evaluation,
abstract = {Abstract: The non-invasive foetal electrocardiogram obtained from abdominal electrodes is an alternative for assessing the foetal health state. This study evaluates the behavior of two well-known methods (Independent Component Analysis and Template Subtraction). These techniques are often referred to in literature for extracting the FECG from the abdominal mixture. Our analysis quantifies the extraction results in terms of accuracy of foetal peak detection using the F1 measure. 1750 five-minute-long simulations considering the absence and presence of non-stationarity mixtures served as dataset for this study. In terms of the F1 measure, on average ICA (97.3 ± 0.1 {\%}) was found to outperform TS (88.1 ± 0.2 {\%}). Moreover, we have shown that the presence of non-stationarities can degrade the performance of both methods considerably, indicating a beat-by-beat adaptive approach may be preferable.},
address = {Uberlandia, Brazil},
author = {Andreotti, Fernando and Behar, Joachim and Zaunseder, S and Clifford, G D and Oster, J},
booktitle = {Congresso Brasileiro de Engenharia Biomedica},
pages = {1585----1588},
title = {{Evaluation of Non-Invasive Foetal ECG Extraction Algorithms using Non-Stationary Mixtures}},
year = {2014}
}

Downloads: 0