Multistage principal component analysis based method for abdominal ECG decomposition
| Autor: | Robertas Petrolis, Vladas Gintautas, Algimantas Kriščiukaitis |
|---|---|
| Rok vydání: | 2015 |
| Předmět: |
Principal Component Analysis
Physiology Computer science Orientation (computer vision) Speech recognition Biomedical Engineering Biophysics Subtraction Heart Rate Fetal Maximal amplitude Signal Independent component analysis Electrocardiography Pregnancy Physiology (medical) Component (UML) Abdomen Principal component analysis Humans Female Fetal Monitoring Electrodes Algorithms Energy (signal processing) Ultrasonography |
| Zdroj: | Physiological Measurement. 36:329-340 |
| ISSN: | 1361-6579 0967-3334 |
| DOI: | 10.1088/0967-3334/36/2/329 |
| Popis: | Reflection of fetal heart electrical activity is present in registered abdominal ECG signals. However this signal component has noticeably less energy than concurrent signals, especially maternal ECG. Therefore traditionally recommended independent component analysis, fails to separate these two ECG signals. Multistage principal component analysis (PCA) is proposed for step-by-step extraction of abdominal ECG signal components. Truncated representation and subsequent subtraction of cardio cycles of maternal ECG are the first steps. The energy of fetal ECG component then becomes comparable or even exceeds energy of other components in the remaining signal. Second stage PCA concentrates energy of the sought signal in one principal component assuring its maximal amplitude regardless to the orientation of the fetus in multilead recordings. Third stage PCA is performed on signal excerpts representing detected fetal heart beats in aim to perform their truncated representation reconstructing their shape for further analysis. The algorithm was tested with PhysioNet Challenge 2013 signals and signals recorded in the Department of Obstetrics and Gynecology, Lithuanian University of Health Sciences. Results of our method in PhysioNet Challenge 2013 on open data set were: average score: 341.503 bpm(2) and 32.81 ms. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|