LIFE SCIENCES

Intramedullary nails, or implants inserted into a fractured long bone such as the femur, help a bone heal faster by stabilizing fractured pieces of bone while they regrow. Intramedullary nails require that a person be up and moving in three days time so the healing process can begin. If the nail is incorrectly inserted, or the wrong model is used, though, it can begin to erode the joint near the hip, causing a much larger problem.

“What we are trying to do is figure out a patient-specific way to decide which implants are best,” says Ralf Schneider, HLRS computational scientist. “Doctors have preferences and certain ways that they perform these operations, and it is fine for most patients, but roughly 10 percent suffer from complications.”
Schneider’s ultimate goal is to help cut down on complications by bridging the gap between medical diagnoses and computational modeling. By designing and prescribing more personalized intramedullary nails, surgeons will be able to more accurately choose and place a design that matches a patient’s body type. Schneider has employed the HLRS Hermit supercomputer to find this solution.

Hermit is a Cray XE6 with one petaflops, or one thousand trillion calculations per second, of computing power. The machine has 3552 dual-socket computing nodes, and its AMD Interlagos processors give it 113,664 processing cores to calculate massive amounts of data.

By using supercomputing technology, Schneider hopes to evolve modeling so that it can be used on a commercial level. Supercomputing lends a platform to develop simulations to lower the time-to-solution, which is the largest hurdle. Current simulations take one week to complete, but medical professionals need to be able to act within 12 hours of receiving a patient with a broken femur, or serious complications can develop

From this C-CT scan, Computational scientists extract the bone's geometry and compute bone material data to give the bone real-life properties. Once a model of the bone is completed, researchers blend bone and implant meshes to get a complete model of the bone-implant-system.

The femoral head is particularly important to accurately model, as the success of an implant largely depends on its ability to keep the implant secure while a fracture heals. The femoral head is largely comprised of spongy, or cancellous, bone with a thin outer layer of strong, compact bone. Medical professionals most consider implant designs that work with a person’s body as well as placing the implant so it is unable to shift during the healing process.

Schneider’s goal is to bring these simulations down to the commercial level. He is collaborating with Lasso, a Stuttgart-based computer engineering firm, to find ways to make these simulations accurate and timely enough for medical professionals and engineering firms to collaborate. Once implemented, surgeons could send CT data to an engineering firm, and ask for simulation-based advice on which type of implant would be best suited for a patient, or if the risk of bone deterioration is great enough to merit an entire hip replacement right away.

Many smaller computer clusters are already capable of creating bone simulations with accurate implants and load trajectories. However, researchers must currently use “archetypes” of bones, as material composition of bone differs from person to person. The next task for supercomputers, though, will allow scientists to develop more accurate models of the material behavior in bone tissues.

Researchers currently simulate bones with isotropic material behavior, which assumes that material behaves the same way in all directions. Schneider is creating models to more accurately calculate anisotropic material behavior, which gives a more complete picture of materials’ reactions to forces acting on a bone.

By means of Micro-CT scans, computational scientists can directly model the cancellous bone's micro structure with very high resolution and its “global” anisotropic behavior on the scale of C-CT scans can be calculated. These high-resolution, micro-mechanical simulations require massive computing power though. A bone cube of only 4.8 millimeters edge length adds up to more than 52 million degrees of freedom.

“If you do not have correct material data, which we are generating here, then the entire process is impossible, Schneider says. “Material data of bone will be the basis for this project to reach the commercial stage.” Schneider hopes his process is ready to begin real-world testing in the medical field within three year’s time. For more information about this project, contact Ralf Schneider at +49 (0711)-685-87236 or schneider@hlrs.de.

--Eric Gedenk

July 2013

HLRS project ID: BoneImplant