Department: Bioengineering
Faculty Advisor(s): Andrew McCulloch

Primary Student
Name: Matthew James Gonzales
Email: mjgonzal@ucsd.edu
Phone: 858-204-6070
Grad Year: 2013

Atrial fibrillation (AF) is a leading cause of stroke and correlates with congestive heart failure, yet a poor understanding of the basic mechanisms of AF make pharmacological therapy and catheter ablation surgeries variably successful. Motivated by important differences between human and animal AF, we aim to use patient-specific computational models in concert with in-vivo mapping studies in humans to investigate the basic mechanisms of AF and optimize existing treatments. Automatic construction of high-quality 3D patient-specific finite element models from non-invasive imaging data (such as computed tomography) remains an unsolved problem, especially for high-order hexahedral models represented as cubic Hermite splines or NURBS. We aimed to construct patient-specific cubic Hermite models of patients referred for ablation surgery, with the aim to use multiscale models of cardiac electrophysiology to guide clinical decision-making in real time. Finite element analysis of high-order solid models with topological complexity has not been completed to our knowledge, likely due to difficulties at regions of irregular mesh topology. We used the concepts of ensemble frames and transition maps between overlapping coordinate charts to support C1-continuous finite element solutions over the entire domain of simulation. We constructed smooth geometries by formulating a volumetric analog to an interpolating subdivision surface algorithm. We demonstrate a smooth electrophysiology solution on a patient-specific finite element model that makes use of graphical processing unit (GPU) acceleration, and solves in a clinical time scale (minutes to hours). Future work will include the incorporation of anisotropic electrical resistivity by diffusion tensor imaging of human hearts ex-vivo. Our results may increase the success rate of AF ablation surgeries, and the methods we develop for solid modeling could be useful for many finite element applications where high-order hexahedral geometries are preferred.

« Back to Posters or Search Results