Text

theses

The T-wave of the electrocardiogram electrically represents ventricular repolarization – the relaxation phase of the cardiac cycle. Discrete states of abnormal T-wave morphology are known to be associated with both pathologic and non-pathologic causes. The links between causative factors and morphologic effects, however, are described as sensitive, but not specific. This thesis aimed to develop a device that could accurately quantify the T-wave’s characteristic morphologic features on a beat-to-beat basis, in real-time, to improve specificity. T-wave feature data were extracted with minimal noise using a novel analog electronic device design that allowed corrections for baseline drift and motion artifacts. T-wave morphologies were then approximated by geometric composite figures constructed from each T-wave’s constituent data, namely, its height and leading and trailing edge slopes. It was hypothesized that the T-wave approximation figures would convey clinically relevant information to an observer, notwithstanding their composition from highly compressed data. Simulated T-wave monitor designs were tested on 2,604 T-waves from thirty-two real and synthetic ECG data sources. Results from the study concluded that over 88% of the geometric composite figures were reasonable approximations of T-wave morphology. Noise on the T-wave signal was the primary cause for less-than-reasonable approximations. Feature accuracies were found to have less than 3% error when tested against a smaller subset of 260 T-wave controls. Clinical meaningfulness of the composite figures was demonstrated by observation of T-wave alternans and the effects of oxygen saturation levels on T-wave morphology. Average baseline drift was held to within 0.010 mV across a wide variety of input conditions. Complete transient response recovery from ±300 mV input pulses sometimes occurred in less than one heartbeat. The present novel methodology employed in the successfully tested T-wave monitor design can be extended to other ECG components, and has the potential to improve the accuracy of arrhythmia detection and classification in future applications.

ETD_4993

M.S.

Includes bibliographical references

by David Winfield Smith

doi:10.7282/T34747WT

ETD graduate

The author owns the copyright to this work.