Charles Taylor, an assistant professor in the surgery and mechanical engineering departments at Stanford Medical School, and his team are pioneering a method of modeling patients’ vascular systems to customize operations to each patient’s arterial system. He claimed that he will extend the worlds of computer-aided design (CAD) and computer-aided engineering (CAE) from the purview of engineers to that of vascular surgeons. A software program developed at Stanford University predicts improvement in blood flow yielded by vascular surgery options. A geometric model of an operative plan shows a contour plot of blood pressure at peak systole. With the Aspire system, doctors can compare the results of a bypass procedure with the surgical prediction modeled before the operation, as shown on this part of the system’s user interface. The Aspire system displays medical images downloaded from the actual patient slated to undergo vascular surgery, so surgeons can tailor the operation to the individual.
If you're an engineer, your days may be spent working with computer-aided design and computer-aided engineering programs, but your doctor's days are not. Charles Taylor predicts that will change within the next decade; he should know. Taylor, an assistant professor in the surgery and mechanical engineering departments at Stanford Medical School, and his team are pioneering a method of modeling patients' vascular systems in order to customize operations to each patient's arterial system. This, he said, will extend the worlds of CAD and CAE from the purview of engineers to that of vascular surgeons.
As a doctoral student at Stanford, Taylor followed a field of research that uses engineering methods to better understand the cardiovascular system. Many researchers are looking at how blood flows through the body's major arteries and how this relates to the presence of disease, mainly arteriosclerosis, the buildup of arterial plaque that hinders blood flow and can cause heart attacks. When he began his research, working with Thomas Hughes, Taylor used computational fluid dynamics methods to model blood flow through the aorta. But the resulting models, he said, were generic versions of the human aorta, much like a representation seen in an anatomy book-although in Taylor's models the blood flow was animated.
"The results are useful in that they're representative of the population as a whole," Taylor said. " It shows you what the system looks like, but in reality, everyone looks different. It's just as if you drew a stick person. Sure it looks like a body, but, of course, everyone looks different on the outside."
When he began pursuing his doctoral thesis, Taylor planned to use engineering software that would model blood flow in arteries. The software could show medical students, for example, how plaque buildup in a certain location affected the flow of blood around the buildup. Again, however, the models would represent flow through the vascular system of a generic patient. And no two people are the same, inside or outside.
But Taylor wondered why the same techniques that engineers use when designing, say, an automotive engine or the wing of an air plane couldn't be used to figure out how to best operate on someone with a vascular disorder.
Using Engineering Techniques
"I had been thinking, maybe there was a way to build models from medical image data. At the time, I thought we could use the information to develop more realistic models," Taylor said. "But then a light bulb went on. We could use this technique to model an operative plan."
After meeting Christopher Zarins, who came to Stanford in 1993 as chief of vascular surgery, Taylor conceived a way of modeling an individual patient's arteries. His method aimed to let surgeons decide how to best operate based on a patient's actual insides-not on what generally happens inside a patient with a similar condition. Taylor's technique would also let doctors try different virtual operations on a model of the patient's affected vascular system, in order to find the surgery most suited to the patient before actually doing the operation.
Currently, surgeons prescribe treatment based on diagnostic data like a magnetic resonance image, or MRI which shows the present state of a patient's body-combined with information about the patient's past gleaned from medical records and interviews with the patient. To make sure a patient's arterial plaque buildup doesn't mean a heart attack, for example, doctors compare information about the patient to typical patients with the same disease who have been successfully treated. A doctor might decide on a particular type of heart bypass operation because it worked well on the others.
"You can say, on average this is what happens to a patient in a similar condition, so this is what I should do. But the diagnosis isn't very personal," Taylor said. "Everybody is different. There's a lot of anatomical variability. In addition, you might have disease in one place and not in another, and you're going to be compared to someone with different clinical problems, of a different size, with a different lifestyle."
Zarins, in fact, has called the current method of finding the correct vascular procedure guesswork.
By the mid-1990s, medical imaging technologies and computer simulation tools were so far advanced that, Taylor figured, combined, they could adequately model a real person's vascular system. Surgeons could then take the models of a patient's arteries and simulate a number of virtual operations that might be appropriate, just as an engineer builds the model of an airplane wing, then simulates its flight under a number of conditions to watch how it works. And in the same way the engineer then chooses the best wing design for his needs, the doctor, by modeling the operation before actually carrying out the surgery, could choose to perform surgery with the best virtual result. In other words, a doctor could perform a virtual operation on a patient in order to see the predicted result before performing the real tiling.
"How do you predict the future? In engineering we deal with this question all the time," Taylor said. "The airplane wing has never been constructed, but you build it on a computer and run it before you make it to evaluate different design alternatives. And then you prototype it and make it.
"But in medicine, how do you prototype?" he added. "You prototype by operating on a person, which is the way it used to be done in engineering. You build and test. And after you've done surgery on the patient, you'll know after the fact that it worked well. You can measure how much blood is flowing through a bypass graft. But there are no medical tools like we as mechanical engineers use to build an engine and simulate how it will work. That's exactly what we're trying to do here. We're trying to predict the outcome of a surgery. Once surgeons can do that, they can design the operation. They can evaluate three or four surgical options and see which one results in the best situation for blood flow."
Beyond simply training students to operate on patients-for which much medical technology is used-Taylor's system uses actual physiological data culled from a patient's MRI to create a solid model of the part of that patient's vascular system of interest to the doctor.
Taylor and his team of graduate students have combined software and hardware to create a program called Aspire-Advanced Surgical Planning Interactive Research Environment. The program essentially links CAD and CAE software in order to model a person's vascular system and show how blood £lows through their arteries. If an area is affected by arterial plaque buildup, the CFD model accurately shows how the buildup affects blood £low. Doctors can then tinker on the solid model with different bypass locations or other options to restore blood £low. Instead of merely drawing a procedure n a piece of paper, doctors use a "surgical sketchpad" to model different blood-flow solutions, using the Aspire system.
"When you design an automobile, you don't know every situation it will function in, but you can model it in many different situations to see what the engine will do in them," Taylor said. "With computer-aided medical planning, you have a hard time getting the numbers because human physiology changes every moment. But you can evaluate one surgical procedure versus another."
The MRI, or other diagnostic tool, provides the hard numbers needed to make the solid model. Think of scanning the human body via an MRI used to create a solid model as a sort of reverse engineering.
Doctors Look at the Future
Aspire made its debut at a 1998 San Diego conference of the Society for Vascular Surgery, of which Zarins was the president. Before the meeting, Taylor and his team set one goal: to give conference-goers a live demonstration of the technology. "That really forced us to develop all the tools and the technology that enable Aspire," Taylor said
The National Institute of Standards and Technology provided a grant through its Advanced Technology Program to Stanford and to Centric Engineering Systems of Sunnyvale, Calif., to develop computer-aided medical planning technology. Centric has since been purchased by ANSYS of Canonsburg, Pa., and is incorporated in the ANSYS product line. NIST figured the surgical technology will have commercial potential that can eventually be mined. Aspire also made use of a geometric modeling application called Shapes from XOX of Minneapolis. The Aspire system is run through a Web browser and uses Virtual Reality Modeling language and Java programming language.
As part of Centric's cooperation, Dolf van der Heide, a Centric employee, became involved in the project as a technical contributor focused on the CFD aspects, with emphasis on modeling fluid-solid interaction. Later in the work, his role was expanded to principal investigator for the NIST project. Centric employees assisted Taylor and his students in solving blood-flow problems in time for the conference.
"Computational simulation-based design and decision making has already proven its merit in many engineering fields ," van der Heide said. "The introduction into the medical world, where those principles are applied to perhaps the most intricate system, the human body, was just a matter of time."
Aspire was a hit at the conference. Many vascular surgeons said they would begin using the system when it became commercially available. But the live demonstration (by four former presidents of the Society for Vascular Surgery planning cases on Silicon Graphics Octane computers) also made Taylor and his students realize that Aspire would have to give quicker results than it did at the conference, where the operative models were preconstructed and the blood-flow solutions precomputed.
Since the joint NIST grant ended, Aspire now runs with a flow solver developed by Ken Jansen at Rensselaer Polytechnic Institute of Troy, N.Y, rather than Centric's Spectrum solver. The Parasolid geometry kernels from Unigraphics Solutions of St. Louis and the Shapes geometry kernel from XOX are used in conjunction with a medical CAD system developed in-house . Rensselaer's Mark Shephard provided the mesh generation software used in Aspire. The Aspire system features customized visualization software.
Since the Aspire debut, Taylor and his team have carried out the first round of animal studies with the technology. They constructed a bypass within a pig and compared actual blood-flow results with the results predicted by the Aspire system. The system has accurately predicted blood-flow changes . In early December, his team began following a patient who will have surgical predictions modeled with Aspire retrospectively. Doctors will compare the Aspire predictions with the patient's post-surgery results to determine Aspire's accuracy in that situation.
Still, by today 's standards, Aspire runs slowly. To model one patient's potential operation takes 24 hours. Before the technology can become commercially available, Taylor and his team must improve the component software and wait for desktop computer power to inevitably double and, perhaps, double again. But, Moore's law being what it is, that shouldn't take too long.
Taylor predicts that in a decade, surgeons will be using Aspire to design operative plans. Within five years, he hopes the technology is in use at Stanford.
"It 's a long-term development effort because it's so fundamentally different than the way that surgeons work now," he said.
The technology may be slow to be introduced into the medical field, said van der Heide, because the stakes are high-doctors are dealing with human lives and they want to make sure the technological tools they use work perfectly.
Still, said Taylor, the U.S. medical industry spends about $10 billion a year acquiring diagnostic imaging data like MRI systems. Of course, such data would be necessary, even if Aspire comes into wide use, because the system depends on its raw patient data. But Aspire introduces a way to predict surgical results and to customize an operative plan to a patient. An MRI does not.
"Diagnostic imaging data basically helps you define what's there. But there are no tools for predicting what's going to happen or to simulate the future ," Taylor said. "This is an important problem for the physician. They want to know what's there, but they also want to know how to fix it. For that, they need a totally new tool and a new way of thinking."