Date of Award
Summer 1995
Document Type
Thesis - Restricted
Degree Name
Master of Science (MS)
Department
Biomedical Engineering
First Advisor
Ropella, Kristina M.
Second Advisor
Myklebust, Joel
Third Advisor
Prieto, Thomas E.
Abstract
Computerized arrhythmia interpretation systems suffer in their ability to detect atrial fibrillation from the surface electrocardiogram (ECG). Atrial fibrillation comprises one of the largest classes of arrhythmia (greater than 1%) seen in the 12 lead electrocardiogram. Poor detection of atrial fibrillation results in inappropriate alarms and a ceasing of other ECG interpretation functions. Furthermore, inaccurate detection of atrial fibrillation necessitates a cardiologist verifying each diagnosis made by such systems Proper detection of atrial fibrillation is essential to the functioning of computerized arrhythmia monitors and ECG interpretation systems in that it eliminates inappropriate alarms as well as reduces the time spent by cardiologists reading electrocardiograms. In this thesis research, we propose the use of magnitude-squared coherence as a method of detecting atrial fibrillation from the surface ECG. This study is the first to address the use of magnitude-squared coherence (MSC) for analysis of cardiac arrhythmias recorded from the surface electrocardiogram. Previously, MSC has been applied to intracardiac data and was shown to be a reliable discriminator of fibrillatory and nonfibrillatory cardiac rhythms. The purpose of this thesis is to determine whether MSC can discriminate between atrial fibrillation and nonfibrillatory rhythms (sinus rhythm and atrial flutter) recorded from the surface ECG. Magnitude-squared coherence is a frequency domain measure of the linear phase relation between two signals. During atrial fibrillation, the activity recorded at one site is unlikely related (in time and morphology) to that recorded from a different site thus, MSC between those two sites is expected to be lower due to the continually changing phase relation. In contrast during nonfibrillatory rhythms where there is an orderly spread of excitation through the cardiac tissue, MSC is expected to be greater. In this research, MSC is performed on the remainder ECGs between two orthogonal surface leads (II and V 1) in an attempt to improve detection of atrial fibrillation. For 68 remainder ECG recordings (23 sinus rhythm, 22 atrial flutter, and 23 atrial fibrillation), MSC was computed between leads II and VI and the mean MSC in several frequency bands was examined. The performance of MSC was compared to ventricular irregularity, percent power and regularity index in their ability to properly detect atrial fibrillation. Results show that mean MSC in the 2-9Hz band was low for atrial fibrillation in comparison to moderate to high for sinus rhythm and atrial flutter. Mean MSC in the 2-9Hz band, R-R variability, and percent power in the 5-9 Hz band for lead V I were able to differentiate atrial fibrillation from sinus rhythm and atrial flutter to varying degrees. Of particular note, mean MSC performed better than the other methods in discriminating atrial fibrillation from atrial flutter. Results suggest that a combination of these algorithms (MSC, R-R variability and percent power) will likely provide better detection of atrial fibrillation in terms of sensitivity and specificity. In this research, only the regularity index failed to discriminate atrial fibrillation from sinus rhythm and atrial flutter. The breakdown in phase during atrial fibrillation can be quantified from the surface ECG using MSC. However, a larger database and additional rhythms need to be evaluated for further validation of MSC results. Since MSC is relatively independent of signal amplitude and morphology, does not require explicit detection of atrial activity and bases it detection of atrial fibrillation on characteristics unique to atrial fibrillation (breakdown in phase), we believe that MSC is a more desirable and more reliable algorithm for atrial fibrillation detection by ECG interpretation systems. Finally, the possibility of obtaining more localized measurements relative to current lead systems, which might improve the performance of MSC from the surface ECG, is being investigated.
Recommended Citation
Sadek, Lara E., "Detection of Atrial Fibrillation from the Surface Electrocardiogram Using Bivariate Analysis" (1995). Master's Theses (1922-2009) Access restricted to Marquette Campus. 4085.
https://epublications.marquette.edu/theses/4085