## All Theses and Dissertations (ETDs)

#### Author's School

School of Engineering & Applied Science

#### Author's Department/Program

Electrical and Systems Engineering

English (en)

January 2009

Dissertation

#### Degree Name

Doctor of Philosophy (PhD)

Martin Arthur

#### Abstract

Background. Diabetes and obesity are two major risk factors for cardiovascular disease. Both can cause changes due to cardiac sources in body-surface potentials: BSPs), that is, in electrocardiograms: ECGs). By identifying the major effects of diabetes and obesity on BSPs, we hope to reveal the electrical phenotype of diabetes in body-surface ECGs in the presence of obesity. Methods. A Bidomain Platform was constructed to link heart-surface transmembrane potentials: TMPs) and BSPs. The Forward-Problem Module of the platform calculates BSPs from a bidomain-model of myocyte TMPs and torso anatomy. The platform also contains a Cardio-myocyte TMP Estimation Module in which an innovative method, named regularized waveform identification: RWI), was developed. It is a new approach used to reconstruct the TMPs from BSPs, that is, solving the electrocardiographic inverse bidomain problem. Using normal TMPs, BSPs were simulated on obese torsos and compared to BSPs on a normal torso to determine ECG changes that might accompany certain obese habitus. BSPs on a normal torso were also simulated with both normal TMPs and TMPs whose duration was increased in a manner expected to occur in the diabetic. In addition, BSPs were measured, heart and torso models were found on two adult male subjects: one normal and one obese diabetic. BSPs and estimated TMPs in these subjects, found by using the RWI method, were compared to identify ECG changes that might be found in the obese diabetic in a clinical setting. Results. Forward-problem solutions found for obese heart-torso models with normal TMPs compared to normal had relative errors: RE) of 12, 30, and 68\% for 20\% left-ventricular hypertrophy, 16\% abdomen extension, and displaced heart, respectively. These results suggest that standard 12-lead ECG measurement could be significantly affected by the anatomical changes associated with obesity. Simulation results also showed diabetic electrical remodeling may have a strong impact on BSPs. An RE of 125\% was observed between normal and diabetic BSPs due to prolongation of recovery that might accompany diabetes. Energy reduction of BSPs was found in both simulated and measured BSPs with obesity. Although QT interval prolongation found in simulated BSPs was not seen in the ECGs recorded from the obese diabetic subject, QT dispersion(QTd) was found increased in diabetic in both simulated and measure ECGs. Obviously, no statistical conclusions can be reached with our limited data set, but the suggestive results call for further clinical observations. TMPs were estimated in realistic, normal heart-torso model simulations using the RWI method. REs of about 15\% were found for up to 10\% noise added to BSPs; and for errors in heart size of 10\% and heart location of $\pm 1$ cm, which were significant improvements over conventional regularization methods alone. Conclusions. In this study, we characterized electrical changes with diabetes and anatomical changes with obesity; then independently evaluated their influences on body surface potentials: BSPs). These results suggest that standard 12-lead ECG measurements could be strongly influenced by the anatomical changes associated with obesity. Body-surface maps and inverse solutions to the heart-surface that minimize volume-conductor effects are likely to be more useful in investigating the influence of diabetic electrical remodeling among obese diabetic patients. Simulation results showed that the RWI inverse solution performed much better than traditional regularization methods alone and is robust in the presence of noise and geometric error. By incorporating temporal information, in the form of the basic TMP wave shape, estimation accuracy was enhanced while maintaining computational simplicity.