Vascular surgeons plan most arterial bypass grafts on the basis of experience and intuition: only after surgery is there a way to tell whether an artery has been grafted to the place best suited to restore blood flow with minimal complications. But a group led by Charles Taylor, a Stanford University mechanical engineer who studies the human vascular system, is proving the power of an experimental software system that lets surgeons "sketch" several possibilities and preview the likely results before making a single incision.

The process starts with the collection of three-dimensional nuclear magnetic resonance data that describe the patient's anatomy-the exact shape of the coronary or carotid arteries, for instance-along with snapshots of blood flow at various points. Taylor's software converts the anatomical data into a numerical "mesh" that represents the vessels, and surgeons use computer-aided design to add hypothetical bypass grafts to the mesh. Next, a supercomputer applies fluid dynamics equations to the original flow data, producing a color-coded simulation of blood coursing through the newly configured mesh. By using the simulations to test a variety of graft placements, surgeons can plan operations that don't inadvertently create areas of low blood flow, a cause of clotting that can lead to heart attacks and strokes.

In retrospective comparisons of data from two human patients who received bypasses in their lower extremities, the system's suggested predictions "probably would have changed how the procedures were performed," says Taylor, who hopes that within five years he'll have made the simulations fast and reliable enough for regular use in surgical planning. He says that "10 or 20 years from now every single [cardiovascular] patient is going to be treated this way."