The Computational Side of the Fence

You re causing ectopy...oh. He s dead. It s alright, keep trying. Those were the words of a fellow biomedical engineer this past May at the annual Heart Rhythm meeting. I was using a cath lab simulator, trying to insert an ICD lead into a simulated heart. The technician was doing his best to help me. He recommended various wires and other devices, told me where to put things, when to twist, when to feed it in further. To me, the image on the fluoroscope looked like a wire wiggling rhythmically in empty space. Eventually, I had to leave and gave up, though not without some taunting from the technician. The reason for my difficulty is that I don t normally do that type of work. I m a Ph.D. student in Tulane University s Computational Cardiac Electrophysiology Lab. As such, when I was approached about writing this article, I was concerned that I would be addressing the wrong crowd. However, I welcome this opportunity to show you what we do in a computational lab. First, I ll tell you who we are and why we re here. Next, I ll outline our major projects, and how we accomplish them. Finally, I ll try to tie our experiments in with your daily work. Dr. Natalia Trayanova, who has had over 80 publications since 1982, directs our lab. Our research analyst, Robert Blake, maintains and continually improves our software. Viatcheslav Gurev, our newest postdoctoral researcher, has been working on simulating electro-mechanical feedback. The bulk of the research work in the lab is done by a complement of eight students. Six are Ph.D. students, including David Bourn, Molly Maleckar, Samuel Kuo, Weihui Li, Xiao Jie, and myself. The other two, Jason Constantino and Hermenegild Arevalo, are master s students. Why use a computational model to study the heart? There are fundamental problems with the use of live models for cardiac electrophysiology research. Studying human subjects is typically encumbered by a lot of red tape, and the studies that can be done are limited. The use of large grids of plunge electrodes or optical mapping in humans is next to impossible. These techniques are usable in animal models. However, even the best experiments using them are limited in critical ways. Grids of plunge electrodes have coarse resolutions. The plunge electrodes can also damage the tissue and thereby cause recording artifacts. Even when the resolution is sufficient, large shocks can saturate recording equipment and prevent data collection during the critical time following said shocks. The most popular alternative, optical mapping, is limited to the surface of the tissue sample. Even with the use of mathematical tricks, it cannot reveal the pattern of conduction more than three or four cells deep. Despite these shortcomings, there is no question at this time that human and animal studies are both useful and necessary. A real heart is the system that we wish to understand. However, computational modeling of that system can contribute greatly to our understanding of harmful arrhythmias. Our computational studies combine geometrically accurate finite-element tissue models with mathematical models of cardiomyocyte ionic currents. Rather than represent the intracellular space, extracellular space, and cellular membrane in three separate compartments, we use the Bidomain model, in which every part of the mesh contains all three. This formulation allows us to accurately model electrophysiological activity and defibrillation in the heart with the least possible computational expense. For some time before I joined the lab, our experiments focused on understanding how shocks actually initiate fibrillation. Using the Bidomain model, we thoroughly investigated what is known as virtual electrode polarization (VEP). Once we understood the basic mechanisms by which it acted on the heart, we began the more complex work of studying disease states and mechano-electric feedback. The first disease state we studied was global myocardial ischemia. While global ischemia is essentially an artificial condition, understanding it was a necessary step before moving on to regional ischemia. Our first foray into global ischemia was headed by our then-postdoc, Dr. Blanca Rodriguez. By applying the most recently developed changes to existing ionic current models, she was able to develop a physiologically accurate simulation of global ischemia. The key changes in ischemia included accumulation of extracellular potassium, depletion of ATP because of anoxia, and inhibition of sodium and calcium channels due to acidosis. Simulation of ischemia was an important step. In addition to the resolution problems posed by animal models, the time course of ischemia is rapid and dynamic, making consistent and useful ischemia investigations difficult. With simulations, however, we were able to thoroughly investigate several stages of ischemia as it progressed. We published two papers on our results, and Blanca won the Heart Rhythm Young Investigator Award in 2004 for her presentation of the second. I joined the lab while these investigations were in progress. While I did some of the work involved in both of them, I was already developing the next step simulated regional myocardial ischemia. For about half of my undergraduate career, I tried to earn a degree in computer science along with biomedical engineering. Part of what drew me to the computational lab was that I would be using my programming abilities to do research. Modeling regional ischemia required new tools to be developed, and so I was chosen by my advisor to work on the project. My first summer in the lab was therefore spent programming. I wrote software to partition our model into ischemic and non-ischemic zones, and then to apply appropriate ionic properties to each zone of the model. In addition to the major region selection software, I wrote many small scripts to manipulate data. The original writers of our simulation software, Drs. Jamey Eason and Felipe Aguel, added capabilities to the software to allow specification of properties on a per-element basis. Finally, after generating the appropriate model, I was able to begin the simulation experiments for regional ischemia in two dimensions. The paper based on that research is in progress, and is nearly complete. We are already running experiments for regional ischemia in 3D. While this is the area in which I focus, it is by no means the only area of study for our lab. We work on many projects in parallel. Dr. Viatcheslav Gurev and Weihui Li are working on mechano-electric feedback, an area not well investigated by simulation as of yet. Weihui is investigating the arrhythmogenic properties of the precordial thump procedure, wherein a physician can stop or start VF with a quick thump on the sternum. Dr. Gurev has built his own mechano-electric simulation software, which models both the viscoelastic contraction of the tissue, and the electrophysiological activity associated with it. David Bourn, who recently received his MSE and has nearly completed his Ph.D., did a lot of work with animal models. He visited the University of Alabama at Birmingham under the supervision of Dr. Rick Gray, where he investigated the ULV/DFT hypothesis using optical mapping of sheep ventricles. His later work with a computational model of a slice of canine ventricles helped him to understand the mechanisms involved in the ULV-DFT correlation and agreed elegantly with his work in the sheep ventricles. His paper with Molly Maleckar on the subject is currently in press. Molly, in addition to her work on the ULV/DFT paper with David, is currently running experiments with her newly developed model of ischemic scar tissue. Where my research models the first 10 minutes after complete occlusion of the LAD, hers deals with the fibrotic scar tissue formed following myocardial infarction. Given that many patients with ICDs have infarction scarring as the result of a previous episode, this research is immediately relevant to the EP field. Xiao Jie is investigating the period between ischemia phase 1A and post-infarction scarring. Her work deals with the changes in cardiac tissue after 10 minutes post-occlusion of coronary circulation. Ischemia does not proceed linearly. As with ischemia phase 1A, it s a dynamic process, and ionic concentrations and energy reserves that increase or are depleted during phase 1A may exhibit opposite behavior in phase 1B. Additionally, in phase 1B, the connexin molecules connecting cardiomyocytes begin to decouple, altering the conduction velocity and safety factor of wave propagation. Samuel Kuo s work focuses on the atria. He studies the effects of islands of acetylcholine on wave propagation and breakup. These islands occur because of the discrete nature of sympathetic innervation in the atria. Since efferent synapses are neither evenly dispersed, nor finely spaced, the atria present an electrophysiologically heterogeneous substrate for wavefront propagation and breakup. While most of our simulations rely on the Bidomain formulation, Sam s atrial simulations use cable theory, with numerical solutions programmed by Dr. Ed Vigmond of the University of Calgary. Our master s students are both working on ambitious projects as well. Hermenegild Arevalo is running simulations to investigate the mother rotor controversy. This is the question of whether ventricular fibrillation is a result of one dominant spiral wave, which drives waves of activation into the tissue, or of repeated collisions and breakup of waves in the ventricles, having no organization at all. Jason Constantino has been examining the effects of ICD electrode placement on defibrillation. He has put hours upon hours of painstaking effort into accurately modeling ICD electrodes in the ventricles. In the past, our computer models used mesh electrodes in a bath, mirroring studies with excised animal hearts, or point stimuli on the myocardium. Jason has boosted the clinical relevance of our simulated defibrillation by developing this more realistic shock protocol. Our work may affect yours in a number of ways. Solid research on mechano-electric feedback by our lab and others may someday lead to more frequent and effective use of the precordial thump procedure in place of others. An improved understanding of regional ischemia may affect the design of ICDs. New electrode configurations, especially in combination with the various regional diseases that we study, may be developed and employed by device manufacturers. A better understanding of the link between the upper limit of vulnerability and defibrillation threshold may improve defibrillation efficacy in a clinical setting. There are likely other ways in which EP lab procedures will be affected by the computational research done in our lab and others, but I couldn t begin to predict them specifically. I hope that this peek into a computational EP lab has given you a broader idea of the basic science research in the field, and perhaps even piqued your curiosity. We d love to see you at the next major EP conference, so please drop by and ask us about our posters in the Basic Science section. Finally, if you d like to follow my daily life as a Ph.D. student in a computational EP lab, please visit https://www.virtuallyshocking.com. You can also find out more about our lab by visiting https://bruteforce.bmen.tulane.edu.