Dr. Charles Taylor would love to experiment with your arteries. He's a natural-born tinkerer, and the idea of operating on you only once and calling it a day seems less than ambitious to him. He's especially interested if you happen to be one of the millions of Americans who suffer from atherosclerosis: dangerous buildups of arterial plaque that hinder blood flow, causing pain, heart attacks, strokes, and in the worst case, death. Were you to hand over your clogged arteries to Taylor, he'd play vascular surgeon and gleefully perform as many as five, even 10 different surgeries on you. And before long he'd be wanting your heart, your liver, and your lungs in order to cut them, stitch them, and subject them to all sorts of pharmaceuticals. Ultimately, if his wish comes true, Taylor will have all your body— down to the last molecule— trapped right where he wants it: on his computer.
Taylor is not a medical doctor. He is a doctor of engineering, but he is also an assistant professor in the departments of surgery and mechanical engineering at Stanford Medical School. He has never cut into a live patient to fix blocked arteries, as his surgical colleagues do every day. Instead, he and his graduate students spend their days perfecting a software program they invented called ASPIRE— Advanced Surgical Planning Interactive Research Environment.
The program gives a vascular surgeon the opportunity to tinker in cyberspace with a patient's arteries, experimenting with different placements for arterial bypasses, different makes of artificial arteries, and a variety of other options to restore blood flow. In virtual reality, doctors are free to follow hunches, even to commit errors, with the luxury of trying again and again until they have explored every option. Then, with the click of a button, ASPIRE can predict which choice will have the best outcome. Right now, by contrast, each vascular surgery is a one-shot experiment performed in real time.
Christopher Zarins, the chief of vascular surgery and a surgeon with decades of experience, describes the current method of choosing the right vascular procedure as "guesswork." If ASPIRE fulfills its promise, Taylor will take that word out of Zarins's vocabulary— and profoundly change the practice of medicine.
In 1993, Charles Taylor was still a Ph.D. candidate in mechanical engineering at Stanford, spending his time developing computer software to test how fluids flow around different materials and shapes. His studies were funded by the U.S. Air Force, in the hope that his work would lead to more efficient airplane wings. That winter he attended a lecture delivered by Zarins on fluid dynamics and plaque formation because he was curious to see if the surgeon's research would echo his own work.
Most vascular surgeons concentrate on the repair of blood vessels, but Zarins is interested in the engineering principles underlying diseased systems. That day he talked about shear stress, turbulence, and pressure gradients, a vocabulary familiar to Taylor. As Taylor listened, he realized— in what he calls a personal epiphany— that the blood flow of the human vascular system was reducible to the equations of mechanical engineering and was, therefore, a legitimate subject for computer-aided design. He decided in that moment to make a virtual model of an individual's arteries that would allow surgeons to try out alternative operations before cutting.
The software program ASPIRE predicts the improvement in blood flow yielded by different vascular surgical options, giving surgeons new knowledge before they cut into the body. (1) This arterial model shows the clinical problem. The abdominal aorta branches into the two femoral arteries that supply blood to the legs, and the gray area at the top represents a dangerous blockage. (The blockage below is not a concern because it is causing no symptoms.) (2) New white branches indicate a bifurcated bypass graft. Here the graft is made to the side of the aorta so that blood flow will be split between the aorta and the graft vessels. This choice yields maximum blood flow overall but creates the potential for a clot because blood will travel slowly through the aorta. (3) The surgeon bypasses the aorta entirely with an end-to-end graft, sewing shut the lower portion of the aorta. This procedure yields less blood flow but carries less risk of clotting. (4) For a patient who cannot withstand the open-abdomen procedure required for the previous two choices, a less invasive femoral-to-femoral graft may be the safer option. Image by Charles Taylor
Zarins immediately grasped Taylor's vision and, along with Thomas Hughes of the department of mechanical engineering, offered to support his research. So Taylor abandoned airplanes and started adapting his computer modeling tools to build virtual vascular systems. "Not hard to do at all," he says. "Fluid systems are fluid systems, be they air over a wing or blood through a vessel. The coefficients are different, but the equations are essentially the same."
Doctors already use computers for imaging the insides of patients and for guiding procedures in real time. Programs exist that train student surgeons to operate on so-called average patients. But Taylor's is the first tool that uses actual physiological data from an individual patient to predict the success or failure of a procedure on that patient. The tools that Zarins uses today to plan a surgery— pencil and paper— look primitive to engineers. Zarins can sketch out his various options: angioplasty or bypass grafts? Real or factory-made arteries? What shape tubing? Which points of insertion? But he cannot predict how any of these options will serve a specific patient postoperatively. "The way we are doing vascular surgery is by building it and seeing if it flies," Zarins says.
Taylor, in fact, sees his mission as strikingly similar to how aeronautical engineers design aircraft: "They build a model of the plane on a computer, digitizing its every component and quality. Then they fly it through cyberspace, subjecting it to various wind speeds and flight patterns, and calculate its performance. They are so good at modeling the plane, the atmosphere, and the flight contingencies that they can predict the success of their design and modify it to ensure perfection before they ever build it."
Doctors who have worked with ASPIRE are excited. "The immense variability of human disease makes things unpredictable for us," says Dr. Frank Veith, the chief of vascular surgery at Montefiore Medical Center and Albert Einstein College of Medicine in New York. "I can operate on 100 people with a similar problem, but with the 101st, I still can't be sure what will happen." A man at the top of his profession, Veith says he is still "constantly worried about making the wrong decision" when he chooses which surgery to perform.
ASPIRE aims to translate every characteristic of a patient's vascular system into computer numbers and thus create a cyber rendering of reality. It uses every bit of vascular data available from the patient: the three-dimensional geometry of his arterial system, his blood pressure, viscosity, and heart rate. It then feeds them into standard equations drawn from fluid dynamics that mathematically encode blood flow and resistance at each point in the patient's system.
Armed with this information, ASPIRE generates on-screen a three-dimensional graphic that the surgeon can rotate and assess from any angle. The graphic shows the surgeon how blood is flowing through the vessels, where the flow is good and where it is obstructed. From a tool bar the surgeon can grab any instrument or device he could possibly want: bypasses and shunts of different shapes and sizes and angioplasty balloons for clearing out plaque. He can manipulate the tools using the computer's mouse as a scalpel. After performing an operation, the surgeon can reset the program and the baseline so that the same set of clogged arteries pop up on the screen again, ready for a second, different operation. Having performed as many operations as he wishes (Taylor speculates that four will be average), the surgeon clicks a button, and the program begins analyzing the new arterial architecture each surgery would give rise to, calculating how much blood would flow through the repaired vessels.
This computer image of the abdominal aorta of a 67-year-old research patient models the pressure of blood flow on the interior of the vessel walls, represented by a range of colors from red to blue, red indicating the greatest blood flow and blue the least. Doctors monitor areas of blue because they are sites at which coronary artery disease is most likely to develop. This volunteer was chosen as a healthy subject, but the value of ASPIRE's ability to describe individual anatomy was demonstrated by this graphic, which revealed a potentially serious problem: The bulges just below the bifurcation of the aorta are aneurysms, and the patient may one day be treated as a result of his participation in this pilot study.Image by Charles Taylor
For Taylor, capturing the dynamics of the vascular system is just the first step. The system of blood vessels is essentially plumbing, and the amount of data one needs from a patient to replicate that system in cyberspace— about one gigabyte— is relatively small in comparison with the incalculable millions of gigabytes Taylor says will be necessary to reproduce the human body. However, even Taylor's modeling of the vascular system makes great demands on a computer. To run ASPIRE and model a single set of surgical choices now takes 24 hours. Taylor is waiting for computer power to double a few more times, so that results can be generated in about an hour, before he makes it available to surgeons.
It is also going to take time to win a difficult philosophical battle in the medical community. Not everyone is easily convinced that human beings can be modeled like airplanes. If we're organic, how can we possibly be digitized? Aren't we ineffably complex? To an engineer, nothing is ineffable in numbers. "It all comes down to our mathematical representation of the world," says Taylor. "Computers can solve the problems if we pose them properly." But if ASPIRE's problems are ill-posed— if Taylor's equations are inexact, if his program lacks a line of code— then his revolution might well founder. He knows it must be flawless.
So he hones his equations and edits his code. So far, using ASPIRE on pigs, Taylor has searched for discrepancies between the virtual version and what actually happens postop to the pig. He's close, he says. So close that he's ready to start shadowing human patients early next year, comparing the program's recommendations with outcomes of actual surgeries in order to see if its predictions are accurate. Taylor hopes he is no more than five years away from perfecting ASPIRE.
Once he has mastered arteries, Taylor plans to tackle hearts and then move upward in complexity. "It's only a matter of computational power— and knowing what the data set is for the model," he says. "If the vascular surgery model proves effective, we could make patient-specific models of the cardiovascular system, pulmonary system, even gastrointestinal. Chemotherapies and tumors. Ultimately, pharmaceuticals and cells. It's a hundred-year problem," he says. "A full human is the most complex system anyone might seek to describe, but we will construct a digital human."
If it works, it might be the ultimate reductionist achievement: not just a vascular system, a heart, an intestinal tract, liver, bones, muscles, or a brain, but all of them together, their relationships accounted for, a cyber avatar— you, on disk.
For more information about the ASPIRE surgical planning software, see www.med.stanford.edu/aspire.