UVM Theses and Dissertations
Format:
Print
Author:
Sharma, Shruti
Dept./Program:
Electrical Engineering
Year:
2010
Degree:
MS
Abstract:
Atrial Fibrillation is one of the most common diseases of the heart, predominantly affecting the elderly population. To understand the disease and in aiding radio frequency ablation procedures, "mapping" of the heart is an important function. Traditional cardiac mapping utilize hand-held bipolar probes placed sequentially at different sights. However, complexity of the Atrial Fibrillation mechanism and distortion of the electrocardiograms due to chaotic activity within the heart, allow only a crude estimate of the electrical activation sequence. Current technology using basket catheters permits recording electrograms at many different locations simultaneously. The task is the automatic determination of activation times and the corresponding frequency between successive activations. While this has been done manually, automatic procedures for such processing of the data are still in their infancy and the problem remains unresolved.
In a normal heart, events in an electrogram are clear deflections from the baseline and the time period between consecutive deflections is discernible. This period remains relatively constant, in a normal heart. In patients with Atrial Fibrillation, electrograms vary to a high degree, resulting inwhat is referred to chaotic electrical activity. In addition, electrograms become fractionated and difficult to decipher. Electrophysiologists resort to frequency analysis to ascertain the so-called "dominant frequency", as an estimate of the inverse of the activation period. For this frequency to be useful in mapping fibrillation of the heart, it should capture the rapid changes in the electrical activity. Currently, simple synthetic signals are created and analyzed in segments of 1-4 seconds using Dominant Frequency (DF) analysis.
The objective in this thesis is to formulate a methodology that can ultimately aid in the automated analysis of electrograms. These results could then be used to generate a threedimensional mapping of the electrical activity of the atrium. We propose as a solution, the Variable Window Analysis (VWA) method with the Short-Time Fourier Transform. Data adaptive overlapping windows of variable lengths are used to capture the time-varying nature of a signal. VWA improves upon the time resolution used in current methods by varying the segment length between 250 and 500 ms. The method is first tested on random samples of simulated cardiogram signals, created with varying ratios of amplitude, activation periods and morphologies to emulate real signals and then tested on real signals obtained from patients with Atrial Fibrillation. An error analysis is provided to determine the accuracy of the estimate.
VWA has not yet been tested on real fragmented signals due to the inability to detect a ground truth to calculate error. Nevertheless, there is considerable challenge in the analysis of simulated electrogram signals. It improves the correlation coefficient between the ground truth and the frequencies calculated using the current state of the art and the VWA from 0.069 to 0.91 proving to be a more reliable and accurate in capturing the changing frequency of the signal than methods currently used.
In a normal heart, events in an electrogram are clear deflections from the baseline and the time period between consecutive deflections is discernible. This period remains relatively constant, in a normal heart. In patients with Atrial Fibrillation, electrograms vary to a high degree, resulting inwhat is referred to chaotic electrical activity. In addition, electrograms become fractionated and difficult to decipher. Electrophysiologists resort to frequency analysis to ascertain the so-called "dominant frequency", as an estimate of the inverse of the activation period. For this frequency to be useful in mapping fibrillation of the heart, it should capture the rapid changes in the electrical activity. Currently, simple synthetic signals are created and analyzed in segments of 1-4 seconds using Dominant Frequency (DF) analysis.
The objective in this thesis is to formulate a methodology that can ultimately aid in the automated analysis of electrograms. These results could then be used to generate a threedimensional mapping of the electrical activity of the atrium. We propose as a solution, the Variable Window Analysis (VWA) method with the Short-Time Fourier Transform. Data adaptive overlapping windows of variable lengths are used to capture the time-varying nature of a signal. VWA improves upon the time resolution used in current methods by varying the segment length between 250 and 500 ms. The method is first tested on random samples of simulated cardiogram signals, created with varying ratios of amplitude, activation periods and morphologies to emulate real signals and then tested on real signals obtained from patients with Atrial Fibrillation. An error analysis is provided to determine the accuracy of the estimate.
VWA has not yet been tested on real fragmented signals due to the inability to detect a ground truth to calculate error. Nevertheless, there is considerable challenge in the analysis of simulated electrogram signals. It improves the correlation coefficient between the ground truth and the frequencies calculated using the current state of the art and the VWA from 0.069 to 0.91 proving to be a more reliable and accurate in capturing the changing frequency of the signal than methods currently used.