Could Simulation Advance Your Testing Programs?
Computational modeling and simulation testing could help expand testing options during medical device development.
At a Glance
- In silico testing could be used during early feasibility and design stages.
- Simulation could help accelerate design and eliminate some physical testing.
- An FDA guidance document shows how to assess the credibility of computational modeling and simulation for submissions.
Computational or simulation testing (also known as in silico testing) allows engineers working in product development to test products virtually. In medical device development, for instance, virtual human models could be used instead of human subjects during various testing phases of product development. Such simulation could help engineers expedite design iterations and therefore potentially accelerate product development. Virtual testing could also increase opportunities to evaluate medical device designs ahead of human clinical trials.
Design News asked MathWorks Medical Devices Industry Manager Dr. Visa Suomi about this emerging means of testing products and how engineers could put it to use. Suomi has more than 10 years of experience in the life sciences and healthcare sector, with an interdisciplinary background from the medical technology industry, academia, and clinical research.
MathWorks offers software for computational or simulation testing and develops algorithms for such testing models.
During what stages are design engineers testing products either for the patient or for meeting requirements?
Suomi: There are roughly four stages in which engineers work in product development. The first stage is the concept and feasibility stage. At this stage they are trying to define the proof of concept. They have some early requirements from stakeholders, such as what the products would look like. They don’t yet know how the device interacts with patients or the human body. They are beginning to balance the design in terms of how innovative or how many features to implement compared to what is actually practical and fulfills early requirements.
The next stage would be the actual design phase, so this is a more formal process. It's at this stage they take those formal requirements that have been written down as specifications and design the medical device. At this stage verification becomes important. They need to make sure that those design aspects meet formal requirements, and their processes must be compliant with regulations in managing iterations.
Next are clinical trials, which entail ensuring device efficacy and safety, followed by manufacturing, which is more about quality management before products are released in the market.
You mentioned verifying that the product meets requirements—is that perhaps where in silico testing could help?
Suomi: So, I would say it helps in the early feasibility stage and in the design stage. You test different types of concepts to see how well they work. You might use simulations at that stage and some basic human physiological models. And once you have a concept ready, you would then move to the design and development stage. You can use some kind of physiological or anatomical model to validate and verify that design, and you can test that in the simulation environment. It also depends on the device type. If you want to then take that into lab testing or bench testing where you use physical prototypes, that's after simulation. So, there's the transition to the physical device at some point after the development stage.
Could computational or simulation testing eliminate some of the physical testing?
Suomi: Definitely at the beginning of conceptualization. You limit your concepts, so you don't have to build so many physical prototypes. And, when it comes to verification of the actual device, there are some examples where you can actually skip physical testing of the device. For example, there are some examples of testing pacemakers for MRI compatibility. During an MRI scan, a pacemaker cannot exceed certain temperatures. Simulation could replace testing so that you test a model with the implant to see how much it heats up.
So, is the benefit saving a lot of the physical iterations? Or perhaps a lot of the early testing when you don't know exactly what direction to go in?
Suomi: Defining those requirements early on, that's definitely one aspect of it. You can test those early prototypes in simulation to make sure that you have the right audience for your device and that your requirements are within the set limits depending on that target audience. At that stage you can do feasibility studies to ensure that the simulation model behaves as expected before you move into the design phase.
Are there examples of how such modeling can help advance medical devices or progress design?
Suomi: Yes, there are examples of novel applications that have recently come to the market in the medical device space, such as continuous glucose monitors combined with insulin pumps. This closed loop system constantly monitors insulin levels of the body and then administers insulin. In those kinds of applications, we have actually worked with some companies where they used cloud-based in silico trials to test their continuous glucose monitors to ensure the response of insulin delivery is in accordance with those models.
Does that mean that it helps with some of these advanced products that have a digital element? Is it suited for novel products that are more than just a physical product?
Suomi: For example, the algorithm that responds to the glucose levels could be an AI model or a new algorithm like a fuzzy logic controller or something else. And you can test different algorithms against the pancreatic model.
Let's contrast this example with traditional physical devices like ventilators. We have worked with many ventilator companies that want to test ventilators for different types of patients and under different fault conditions and what might happen during ventilation. It’s kind of like a closed loop system. So, you have the ventilator and the patient, and you test the ventilator design against the in silico medicine model. For instance, we worked with Cambridge Consultants and a UK vendor that was supposed to build new prototypes of ventilators for the UK government to address the shortage of ventilators and they used this simulation model to rapidly iterate the ventilator design that works with different patients. Once they had the simulation concept ready, they went to the hardware design, and that took only 47 days. So, it was a very rapid iteration of the design.
Has FDA weighed in on the use of such simulations? Are there any concerns that engineers would have from a regulatory perspective?
Suomi: I think for a long time there has been support of this type of computational evidence for regulatory approvals. In 2016, FDA released a guidance document on how to report modeling studies for medical device submissions. After they released other documents for manufacturers on how to assess computational simulation models. For example, just last November they released a guidance document on how to assess the credibility of computational modeling and simulation for submissions. They also published a report showing examples of how manufacturers have been using computational modeling and simulation to support regulatory submissions. And in many cases, the simulation data from these models is complementary to the clinical data. But there are certain aspects of the design that can be verified and validated only in simulation, and one example of that is the MRI safety assessment of the pacemaker that I mentioned earlier. You can prove the MRI safety of the pacemakers just by using simulations if you have a properly validated model for this testing.
For engineers working in this space, the regulatory aspect is still kind of a concern—Is it allowed or not? When I look at reports from the Medical Device Innovation Consortium in the US, every five years they do a study in which they ask the industry for the biggest concerns for device approvals, and the regulatory uncertainty of using computational modeling and simulation always comes up. But FDA has been very clear that they support it, and they have evidence and examples that companies have been using it. There are several documents showcasing how it can be done and how to assess the credibility of the computational models.
And on top of that, there are standards being developed on computational modelling in medical device development. For example, ASME has a standard on computational modeling and simulation, so they're credible to use in medical device simulations.
So yes, there is a lot of support from the standardization point of view and the regulatory point of view and that shouldn't be a bottleneck or a concern for companies to use this type of technique in product development.
FDA has a couple of programs in which they proposed solutions or tools to be used in the simulation studies for medical device development. For example, they have a program in which they assessed certain tools to be used in medical device development, and there are some medical device simulation tools available.
And the Office for Science and Engineering Labs has a regulatory science tools catalogue, and it includes tools to support in silico testing of medical devices at an early stage. And by the way, many of these tools in the catalog are also made with MATLAB.
Are there any skills that engineers need to develop if they're not used to doing simulation?
Suomi: One key aspect is that you should know the scope of the model that you want to apply to your device. So, if you want to model a ventilator in the simulation environment, you should always focus on the physiology and the anatomy of the lungs. You need to know what aspects of the human lungs are relevant for testing. Is it the humidity of the lungs or the pressure coming out of the lungs? Breathing or respiratory frequency? These aspects would be modeled so that you can test your device in this environment.
There's a learning curve if you haven't done computational modeling and simulation before, but a company can hire someone who has that experience or use training to ramp up in this domain.
So, it sounds like there's a lot of support if engineers are not used to working in virtual modeling.
Suomi: There are already a number of very good models available. We have a library of in silico medicine models on our website that can be used for design verification of certain type of medical devices.
For example, we have artificial pancreas models that can be used for verification of continuous glucose monitors and insulin pumps. We have cardiovascular system models that can be used for verification of heart-lung machines. We have just recently posted electrophysiological heart models developed by the University of Auckland that can be used for testing pacemakers.
All these models are already available, so you don’t have to start from scratch. You can take something ready-made and then start building on top of that.
Are there any future models you're planning?
Suomi: We are, of course, looking at the applications, what's going on in the industry, and what devices manufacturers are developing. Based on those needs, we then develop models that could address some of these needs.
There's a lot of talk about design for manufacture, such as ensuring that designs can actually be manufactured. Does any of the modeling software help in that regard?
Suomi: Some topics around manufacturing include what material to choose for the medical device. You could simulate material interaction with the human body and see that it's safe and it meets the purpose. Another aspect could be individualization, customization, or personalization of medical devices. Certain types of medical devices need to be personalized such as prostheses or some surgical implants. So, you could use simulation early on to test different designs and then move those into manufacturing.
Can in silico testing or computational modeling be used in other industries? Could those other industries like automotive or aerospace look to what the medical device space is doing and then replicate it for their industry?
Suomi: Automotive and aerospace are using a lot of simulations for different aspects of testing. The big difference between the medical device industry and these industries is automotive and aerospace have a very limited set of different types of products and testing. But in the medical device industry you have hundreds of different applications, hundreds of devices, and different types of therapeutics or diagnoses. You cannot apply the same computation modeling techniques from those industries to medical devices. You need to operate models for all these different aspects.
So, from my point of view, we want to accelerate adoption of computational modeling and simulation, so therefore we have resources available for training. We also have ready-made models as I mentioned before. They can speed up development, and if companies need help in implementation or any of the work, we have engineers and consultants to support in that aspect. It's not only about the software, but it's also about leveraging the community when developing these models for medical device design.
Editor's note: An earlier version of this article misstated that electrophysiological heart models for testing pacemakers had been developed by the University of Oakland when in fact they had been developed by the University of Auckland. We have corrected this error.
About the Author
You May Also Like