The Virtual Heart
The 70-year-old patient in the Auckland Hospital in New Zealand had suspiciously low blood pressure. The doctors were stumped. But they had an unusual experimental tool at their disposal: a unique computer program that analyzes a magnetic-resonance imaging (MRI) scan, measuring the motion of a patient’s heart and comparing it to that of a “healthy” virtual heart constructed not of blood and tissue but from mathematical equations. The analysis handed the clinic’s experts the smoking gun: part of the heart was twisting in a pattern often associated with a partially blocked valve, which, untreated, would probably kill the patient within three years.
To diagnose this disorder, surgeons would normally have to crack open the patient’s chest. But the software had accurately identified the problem in about 15 minutes. “It helps point out where the heart wall may be failing,” says Peter Hunter, the University of Auckland bioengineer whose team developed the software in collaboration with the German company Siemens.
The MRI analysis program is just one of a rapidly growing number of medical applications emerging from an ambitious global effort known as the cardiome project. The goal of this multilab endeavor is to build a virtual heart: a computer model that accurately depicts everything from a single cardiac cell up to the whole organ, from the interwoven electrochemical activities of millions of cells to the delicately synchronized pumping of blood. The model should even be able to “suffer” from the blocked arteries, weakened muscles, and erratic electrical rhythms that characterize heart diseases.
Medical researchers have been working on computer models of the heart for decades. But thanks to exponential leaps in available computer power, rapid progress in describing the precise and complex details of how the heart actually works, and the fashioning of mathematical representations of those details, increasingly lifelike models of the heart are beginning to yield real health dividends. Insights gleaned from the virtual-heart project are leading to new approaches to diagnosis, surgery, and drug discovery, with the potential to improve or even save the lives of the more than 13 million people in the United States alone who suffer ailments ranging from heart attacks caused by clogged coronary arteries to potentially fatal abnormal heartbeats triggered by rare genetic mutations. “We can do a good job now of modeling on a computer what happens to cardiac cells in heart failure, and predicting how a heart contraction will respond to a drug or other stimulus,” says Andrew McCulloch of the University of California, San Diego, a leading researcher in the field. “It’s allowing us to answer a lot of experimental and clinical questions.”
The virtual heart is a work in progress that does not yet mimic many of the intricate and still mysterious genetic, cellular, and mechanical processes that take place in real hearts. Nevertheless, as the project’s computer models improve over the next several years, they could revolutionize the diagnosis and treatment of heart disease by casting new light on the complex workings of the organ, and serving as tools for quickly and cheaply testing drugs, diagnostic devices, and surgical treatments that are still too risky to try on humans.
In a Heartbeat
Though the virtual-heart project is of global scope and has no official headquarters, it is widely agreed that its front line lies beyond the University of Oxford’s ancient, stately colleges, in a drab, modern building that looks out of place among its crenellated neighbors. Here, in a four-floor wing dedicated to cardiac science, is a research center equally uncharacteristic of its surroundings. Instead of stainless-steel tables, microscopes, and flasks of cells, this modest suite of offices is packed with computer workstations whose monitors are filled with strings of software code. This is the domain of Denis Noble, a man credited with almost single-handedly creating the field of cardiac modeling nearly 45 years ago. These days Noble, head of Oxford’s Cardiac Electrophysiology Group, is easy to spot among the graduate students and postdocs: a lean 67, he is the most hiply dressed and also appears to hold a solid edge in energy as he dashes among team members whose work ranges from hard-core computer programming to basic tissue dissection. Cardiac modeling, Noble says, necessarily combines the talents of researchers who might never otherwise come in contact. “This is a new form of biological science,” he says. “Being highly collaborative is essential.”
In a sense, the cardiome project began in 1960 when Noble came up with a set of equations that describe how the electrical activity of cardiac cells is largely controlled by the flow of potassium ions through their membranes, which leads to waves of activity that spread through neighboring cells and ultimately generate the coordinated beating of the heart. While the idea of describing physiological activity in terms of mathematical equations seemed groundbreaking at the time, Noble’s original model appears almost quaint compared to those his lab works with now-monstrous formulas with 23 variables accounting for 12 different types of cellular ion flows. Crunched on a computer, these models churn out a millisecond-by-millisecond simulation of a cardiac cell’s activity.
But modeling a single heart cell gets you only so far. Helping patients diagnosed with diseases from high blood pressure to congestive heart failure requires a model of the entire organ. Enter Peter Hunter, a former Oxford colleague of Noble’s. Where Noble works on individual cells, Hunter has taken on the task of modeling the heart’s large-scale structure and mechanics-that is, the beating of the heart muscle itself. When Noble visited Auckland in 1991, he found Hunter’s group making ultraprecise measurements of hearts extracted from dogs. “These people were shaving down a preserved heart a fraction of a millimeter at a time, like old-fashioned anatomists,” recalls Noble. Hunter’s intent was to build a model that would bridge the gap between heart science at the cellular level and the structure and function of the whole organ. In other words, he wanted to map out exactly how all those ion flows in cardiac cells teamed up to create a heartbeat, and in particular where things were going wrong in diseased hearts.
Today, the efforts of Hunter’s and Noble’s labs have been combined into whole-heart models whose behavior reflects the independently calculated activities of up to 12 million virtual cardiac cells. A real heart has closer to a billion cells, but even today’s fastest supercomputers can’t track that many cells in a reasonable amount of time. As it is, some of the Auckland models-which represent human, dog, pig, guinea pig, and mouse hearts-are so complex that it takes eight hours or more of a supercomputer’s time to crank through a single heartbeat. Explains Hunter, “The models show how electrical activity originates at the cellular level, how the activation wave spreads to other cells, how the electrical wave is converted to mechanical contraction of the heart wall, how the contracting walls cause blood to flow through the heart, and how energy is distributed through the whole system.”
Despite the complexity of these models, there was still a missing element: genes. It turns out that genes play an enormous role in heart disease; the inheritance of a single unlucky gene can raise the chances of early death from a long shot to a near certainty. Even genes that normally don’t cause heart problems can do so when they are switched on or off or damaged by environmental influences, such as cigarette smoke or stress. To make things even more complicated, heart disease itself can influence cardiac genes in ways that accelerate the disorder or cause new complications. To accurately model disease in a heart, researchers must account for these genetic factors.
Working with UC San Diego colleagues, McCulloch is using genetically engineered mice to identify genes that play a role in heart disease. He is then using that information to modify virtual-heart models. McCulloch’s lab uses mice with changes in a single gene that render it either constantly active or constantly inactive. These altered mice are then studied for differences in cardiac functioning and susceptibility to heart disease; any such differences can generally be attributed to the altered gene. If a mouse that has had a certain gene made continually active develops heart disease at an unusually early age, for example, then the computer model can be adjusted so that switching that gene on in the virtual heart will trigger disease processes. Such modifications can be critical in making the models more realistic. If the virtual heart is used to investigate a drug designed to prevent the onset of heart failure after a heart attack, for instance, then it has a better chance of predicting how well the drug will work if it includes the genetic processes that the drug might influence.
Built up from the workings of individual cells and genes, the virtual heart presents a vivid image of the vital organ. But is it a realistic one? The models provide what are essentially predictions of how a real heart would behave, and researchers need ways to ensure the accuracy of these predictions. Chris Johnson, a computer scientist who directs the Scientific and Computing Imaging Institute at the University of Utah in Salt Lake City, has created one solution: a way to gauge the models against data from living volunteers.
The main tool for measuring the heart’s electrical activity, an electrocardiogram that takes readings from 12 electrical leads, yields only a relatively crude analysis. But a “jacket” developed at the Cardiovascular Research and Training Institute at Utah that employs 192 leads, along with a standard MRI scan, gives Johnson a much more complete picture. To translate the jacket’s measurements and the MRI data into a detailed picture of the heart’s electrical activity, Johnson first takes into consideration how bone, blood, fat, and muscle distort a signal traveling from a particular point on the heart to a particular point on the skin. He can then infer an electrical map of the heart at any point in time. “We’re taking voltages from the outside and determining what they would be on the surface of the heart,” he says. This allows the modelers to determine whether their millisecond-by-millisecond predictions of the heart’s electrical activity are accurate-and to fine-tune their calculations to bring them closer to reality.
Johnson’s models and the electrode jacket are also in experimental use to help cardiologists spot heart disease. While electrocardiograms of hearts with potentially fatal artery blockages often look completely normal to all but the most expert eyes, the jacket-based system generates almost MRI-like images that can reveal blockages and other defects with such stark clarity that even a layperson can spot them. Johnson’s team has also created software that allows the simulations to be viewed in 3-D with special stereoscopic glasses. The improved view could, for example, allow doctors to initiate drug therapy or perform artery-clearing angioplasty earlier than they might otherwise, which could help prevent heart attacks or avoid the need for more invasive coronary bypass surgery.
The virtual heart has, in many ways, come alive over the last dozen years. But it still has a long way to go. “We can model a heartbeat over a period of 10 minutes,” says McCulloch. “But we can’t yet model the natural progression of disease-how a cardiac cell gradually proceeds from normal to injured to failed.” One barrier: although hundreds of researchers around the world are exhaustively deciphering the workings of the heart, most biologists haven’t been trained to gather and present data in a rigorous, quantitative way that can feed into the mathematical formulas used to build computer models. “When you talk to them about describing their results as formulas, some of them get very turned off,” says Paul Herrling, head of corporate research for pharmaceutical maker Novartis.
Yet the cardiome is already making contributions to medicine, and one of its biggest may be as a tool to help researchers discover better heart drugs. Novartis, for one, is already using cardiome models to develop drugs by programming in the changes that a compound has been observed to make in a cardiac cell, and then letting the model project how those changes will affect heart rhythm and blood flow. “We’ve been able to make predictions of which ion channels in heart cells to tweak with drugs to reduce arrhythmias,” such as those found in patients who have suffered heart attacks, says Herrling. He emphasizes that the cardiome needs a great deal of additional development before it’s capable of providing detailed, complete, accurate predictions of how the heart would respond to a wide range of potential drugs. “But we’ve had a sufficient number of elements come together to allow getting a good start,” he says. “That tells me it’s worthwhile pursuing the models, even if they’re not perfect yet.”
The virtual hearts are also advancing surgical therapies. For example, about five million Americans suffer from congestive heart failure, and one relatively new treatment that is gaining popularity involves implanting two pacemakers in patients to counter the abnormal heart rhythms typical of the disease. But doctors can have trouble determining the sequence of electrical stimulation that best ensures a stronger heartbeat. So McCulloch has adapted one of his models to simulate a diseased heart with two pacemakers, allowing him to experiment on a computer to find the right placement and timing for the two jolts. “There’s intense interest in the work from pacemaker companies,” he says.
As exciting as these early applications are, the modelers have far greater ambitions. Eventually, biologists and physicians hope, modeling research will give life to an entire virtual patient, with a full complement of simulated organs. That would enable, for example, studying how an experimental heart drug affects the kidneys, or identifying the long-term effects of a high-fat diet within weeks, rather than following human volunteers for years. Taking one small step toward this lofty goal, Hunter is helping to oversee the development of an open-standard programming language called CellML, based on XML, the Web page development language. Over the next two or three decades, CellML and other such standardized tools will give modelers the world over a common language and enable the integration of the cardiome work with computational models of other organs. “We’re all asking ourselves what sort of infrastructure we need to make sure our work is expandable and extensible to other applications at other levels,” says Johnson. “We don’t want the cardiome to be a one-off.”
The flurry of modeling is leading to a promising trade-off: the better we get at creating virtual heart disease, the less we stand to see of the real variety.
Virtual Hearts in Operation
|Artesian Therapeutics (Gaithersburg, MD)||Cardiac models to support drug development|
|Immersion Medical (Gaithersburg, MD)||Whole-heart models for training surgeons|
|Insilicomed (La Jolla, CA)||Whole-heart models for medical-device design|
|Predix Pharmaceuticals (Woburn, MA)||Cardiac cell and tissue models for drug discovery|
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
Meet the people who use Notion to plan their whole lives
The workplace tool’s appeal extends far beyond organizing work projects. Many users find it’s just as useful for managing their free time.
Learning to code isn’t enough
Historically, learn-to-code efforts have provided opportunities for the few, but new efforts are aiming to be inclusive.
Deep learning pioneer Geoffrey Hinton has quit Google
Hinton will be speaking at EmTech Digital on Wednesday.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.