Computer imaging and web-based 3D tools help surgeons predict patient outcomes

People are as different on the inside as they are on the outside, making it difficult to predict how an individual will respond to a surgical intervention without resorting to statistics and educated guesses. Charles Taylor, PhD, assistant professor of mechanical engineering and of surgery at Stanford University, is using his engineering expertise to try to take the guesswork out of predicting surgical outcomes.

For the last decade, Taylor has been taking the detailed information of diagnostic imaging tools like MRI and CT scans and using it to form the basis of computer modeling programs that can help foresee the results of medical interventions.

On Feb. 21 at the annual meeting for the American Association for the Advancement of Science in Washington, D.C., Taylor presented his latest accomplishment: factoring in the flexibility of veins and arteries to his model of the cardiovascular system. The realistic response of blood vessels adds more predictive ability to earlier versions of his simulation, which assumed rigid vessel walls for simplicity.

"The physics of blood flow is so complicated that it is impossible to guess what will happen during a surgery," Taylor said.

Medical professionals now have access to powerful tools for acquiring and visualizing data, Taylor noted. "It is incredible what we can see, but what we need to be able to do to make a good decision is to project into the future," he said.

In other words, he said, "the diagnostic data tells us what is there, but we need to know the answers to the 'what if' questions." These questions include, "What if it is better to do nothing?" And, "What if the patient got a little better but there might have been something else that could have made him or her a lot better?"

Taylor's computer model incorporates imaging data into a Web-based tool that includes 3-D views and surgical sketchpads. Millions of complex equations involving fluid dynamics model an individual's physiology and demonstrate what might happen under different circumstances.

Funded by a grant from the National Science Foundation, Taylor is testing his model in retrospective analyses, when a surgeon is going to do a procedure anyway. His team gets data before and after surgery and determines how well their model predicted what was actually seen. In large-animal studies, he said, they can predict flow after an aortic graft within 10 percent.

"The point is to intervene first on the computer before going to the patient," he said. "We don't care if something bad happens to the computer model."