Visualize MED: Building Trust in Computer Modeling and Simulation

Medical device makers lag other industries in their use of computer models. One of the big issues discussed at Visualize MED is how much can researchers, regulators, and engineers trust their models and simulation?

by Alan Brown
June 11, 2018

Last year, 220 of the 1500 new medical devices submitted to the U.S. Food & Drug Administration included computer models and simulations as evidence in their regulatory submissions.

Yet despite their high-tech patina, medical device makers lag other industries in their use of computer models. Those models, as well as simulations, bring substantial advantages to the table. They can help companies design better, safer products while reducing the number of animals and humans needed for testing.

Still, one big question looms large for medical device makers: How much can researchers, regulators, and engineers trust their models and simulation?

That question was at the heart of Visualize MED, a forum for senior leaders from the biomedical engineering community held last month in Minneapolis, MN. The meeting coincided with ASME’s Verification and Validation (V&V) Symposium at the same location, which provided a more technical slant on many of the same issues.

The consensus seemed to be that although there are still hurdles as far as the trustworthiness of models, large companies are slowly broadening their use of modeling and simulation across an array of applications.

Markus Reiterer, a technical fellow at Medtronic, a medical device company based in Minneapolis, led off the forum. His company had previously used modeling and simulation to prove to FDA that its Advisa pacemaker was safe when patients underwent full body scans in MRI machines.

For You: Bioprinting Better Artificial Joints

In the past, FDA placed restrictions on pacemaker recipients in MRI machines because microwaves heat metal. Medtronic research showed that pacemaker lead wires were vulnerable. It redesigned the leads to reduce heat buildup inside an MRI. The company then ran a clinical trial of 275 patients for the leads in its Evera pacemaker.

At the same time, it used a computer model to simulate more than 2.3 million pacemaker-MRI scenarios. This enabled the company to verify its wire heating models. The large number of scenarios also provided more information about those interactions across a much wider range of body sizes and types than any possible clinical trial.

Because Medtronic had a verified model of lead wire heating, FDA approved the new leads for the Advisa pacemaker without requiring clinical testing.

Medtronic is building upon this success by pinpointing design issues early and using models to unify the work of its many different specialists.

“Computer modeling and simulation reduce time and costs and increase the performance of product,” Reiterer said. “The more we observe, the more the more we learn. We learn early, and we learn quick.”

Modeling also brings two important strengths to development. First, it highlights potential problems, pinpointing where teams need to conduct experiments. Reiterer finds, for example, that his group now does more materials characterization and validation experiments at the end of the design cycle.

Second, models provide different specialists with a common language.

“We typically cut problems into smaller and smaller pieces until we can understand them,” he said. “But now there are so many specialties, it’s hard to put the puzzle together again. Modeling and simulation bring everything back together to create a bigger picture that different disciplines can share.”

Medtronic uses modeling in several ways. One is to accelerate R&D. It uses simulation to test strain on device interconnects due to deflection. The model calculates changes in conductivity and battery discharge of milliseconds—as well as years—of use, so designers understand the impact of their decisions.

Medtronic also models biological systems. For example, it used its heart model to study arterial fibrillation and develop new therapies, including catheter ablation, antiarhythmic drugs, and device therapy. Medtronic also conducted animal testing to validate their approaches.

“We want to ultimately skip animal testing and go to human trials,” Reiterer said.

Better Models Lead to More Profits

Medtronic’s goal is to be paid for the value of its devices. To reach it, Medtronic must predict health outcomes. This is especially difficult with medical devices, whose final assembly takes place inside the patient during surgery. To reach that goal, Medtronic needs models that better capture anatomical and physiological functions.

Another challenge is evidence. Models, after all, are only models of reality. They are based on “unreal” data: clinical trials designed to produce optimized outcomes by including carefully selected patients, physicians, and perfect clinical environments.

To capture a broader range of clinical evidence, such as the results of underserved or very old or very young populations, requires larger trials or more reliable models that can capture the variability in human patients and surgical skills. The payoff from more extensive modeling comes in lower costs and fewer trials that expose patients to unproven therapies.

Reiterer mentioned several other barriers. On the technical side, they involved developing a better understanding of patients and disease pathologies. On the institutional side, they include general skepticism and regulatory uncertainty.

“We need to create more use cases, including verification and validation and economics, so management will invest time and money in this technology,” he said.

He also called for a modeling and simulation development protocol, since large companies “cannot rely on one person doing it all, from beginning to end.”

Stryker, a medical device and equipment manufacturing company based in Kalamazoo, MI, uses modeling and simulation mostly for design, said Scott Taylor, senior director of engineering for the company’s Joint Replacement Division. The company relies on its library of 3,500 computed tomography 3D scans plus 15,600 bone segment scans, to which it added muscle and soft tissue elements. The company can then dissect those by age, race, disease, and height, and use them to create variations in density or orientations of defects.

The company uses modeling to understand failure, to avoid under-estimating complexity, and to minimize the risk of late-stage verification and validation failure.

It also uses modeling for engineering tasks. One example is simulating steam flow and temperature rise through trays holding instruments during sterilization, though the models are “very far from actually replacing sterility testing,” Taylor said.

Other examples include finite element analysis of bone-implant interactions and looking at the displacement of media during vibratory polishing.

While the FDA is more open than ever to models and simulations. It allows them to be used as evidence in for Class I devices, which often involve new sizes or substituting a previously approved material, which don’t usually require clinical testing. Yet Taylor found it difficult to work with FDA on a protocol for MRI safety.

“Stryker’s implants are nonmagnetic,” he said. “But it was painful. We’d identify a worst-case scenario for our components, and the FDA pushed back and wanted to compare our models with in vivo and cadaver data. That’s why standards like ASME’s V&V40 are essential.

V&V40 uses statistical and other techniques to measure uncertainty to improve the credibility of computer models and simulations. It does this by verifying that the code represents and properly solves the underlying mathematical models. It then verifies that the code is an accurate representation of the system it seeks to model.

Read More: Designing Autism Friendly Clothing

Soft Wearable Exoskeletons Challenge Status Quo

Restoring Sight to Blind Rats