In Silico Medicine for Modelling and Simulation of Medical Devices
Overview
The validation and clinical evaluation of medical devices can be performed using living animals or humans, but this is expensive, time-consuming, and sometimes even risky to the test subjects. Alternatively, the testing can be conducted with tissue-mimicking phantoms or in vitro, which can reduce time and costs but does not accurately reflect a real human.
The question then arises: is it possible to eliminate the drawbacks of time, costs, and possible safety risks while maintaining the advantages of in vivo evaluation? One way to do this could be In Silico Medicine, which refers to the use of virtual human models to replace their physical counterparts in testing of new medical devices.
In Silico Medicine uses computational models and simulations to develop medical devices and pharmaceuticals. This webinar explains how you can leverage in silico medicine in the research and development of next-generation medical devices while reducing development time and costs.
Highlights
- In silico medicine uses computational models and simulations to replicate human anatomy, physiology, and biological processes
- The virtual human models can be used in the design and testing of new medical devices
- Manufacturers using in silico medicine benefit from accelerated research and development cycles while ensuring the quality and safety of their products
About the Presenter
Dr. Visa Suomi is the Medical Devices Industry Manager at MathWorks. He has over 10 years of international experience in the life sciences and healthcare sector, with an interdisciplinary background from the medical technology industry, academia, and clinical research. He holds a doctoral degree (DPhil in Healthcare Innovation) from the University of Oxford, UK, with the focus on translating academic and clinical research into commercial applications.
Recorded: 22 Jun 2023
Hi, everyone. My name is Visa Suomi. I'm the medical device industry manager at MathWorks. I work with a lot of medical device customers throughout Europe. And today I'm going to present to you this topic of in silico medicine and how you can use it for modeling and simulation of medical devices so that you can accelerate the development time and use virtual prototyping instead of physical prototypes.
But first, a bit of background about MathWorks. So MathWorks, we are a software company. We are producing software for engineers and scientists to accelerate their product development and research. We have our own primary users in more than 180 countries throughout the world. And globally, we have around 6,000 people in staff in 34 offices around the world.
Our headquarters is in Natick in the US, close to Boston. And we are a private company. And that means we work with a lot of different customers in a partnership. And we also work across multiple different industries, including the medical devices industry.
Our two main products are MATLAB and Simulink. MATLAB is this scripting-based language that you can use to develop, for example, algorithms in signal or image processing. You can also put AI models or analyze any type of big data that you might face in biomedical field or in health care.
Simulink is mainly used for simulating complex systems such as medical devices together with the sensors and software. So you can use Simulink to build your medical device model in the simulation environment, see how it works, test and verify it, and then later on generate code for the real product as embedded software. Furthermore, we have more than 100 add-on products for different specialized R&D tasks. For example, communication, machine learning, image processing, or signal processing.
Our customers and partners, they come from private and public sector. We work with more than 90,000 business, government, and university sites, including the largest medical device manufacturers as well as the largest pharmaceutical and biotech companies. And we also do a lot of projects with hospitals and medical research institutions that make algorithms and AI models to improve the patient outcomes.
The agenda for today's presentation is shown here. I will first introduce you the topic of in-silico medicine. What does it mean? How does the market look like? What types of challenges and opportunities there are in this field, and what do the regulators tell about in-silico medicine and how you can use it for medical device development.
After that, I will talk about the computational modeling and simulation workflows, how you can apply these techniques in practice, what type of steps do they involve in R&D to use computational modeling and simulation, and how you can start using them in your own process. After that, I will show some case studies from the industry, some examples that we have been working with customers. For example, a dialysis machine example, an example of a cardiovascular system, and also an example of a artificial pancreas among others. Finally, I will conclude the presentation and show you some materials and other webinars where you can learn more about the topic.
What is in-silico medicine? What does it actually mean? In-silico medicine uses computational modeling and simulation to replicate humans and medical devices. So in theory, you can use these virtual humans or organs that are sometimes referred to as digital twins in order to access your R&D process for new medical devices.
So instead of doing things in vivo, in vitro, or on the bench, you can model and simulate your medical device together with the patient, simulate the interaction, and do some early feasibility studies of the device in order to make the right requirements for your medical device so that it's effective with the patient group that you are targeting. Some applications in medical devices for this type of technique include design verification, performance testing of the device, clinical evaluation of the device, and also providing a supportive record of evidence for record submissions with the FDA, for example.
In-silico medicine also has several benefits to all health care stakeholders. For example, regulators, they see computational modeling as a way to provide supportive evidence. So in addition to your real clinical data from the clinical trials, you can also submit virtual patient data to support the evidence for regulatory approvals.
For companies making medical devices, they can use these virtual human and organ models to accelerate device development and testing. This eventually is shorter time-to-market because you do all of these early feasibility studies in a virtual environment defining the right requirements for your device, and do some early clinical phases with the testing.
You can also reduce R&D costs because you are not relying that much on hardware products. You can develop your device in a simulation environment and when you are ready to go to hardware, you are already down far away in the final product and how it should look like.
Patients in the end, they will benefit from safer medical devices. For example, in silico clinical trials. They can be conducted with larger and more diverse patient cohorts, including different age groups, different genders, and also different ethnic backgrounds.
The market for in-silico medicine is also growing. So in '21, the market was around $3 billion US dollars. It is expected to grow to up to $7 billion in '28. So more than double the growth in just seven years.
And that means more and more companies are coming to this market, providing their solutions to medical device companies to use virtual prototyping and virtual human models to accept devices worldwide. And we have also seen already a lot of startups, innovative companies providing software services where they can use these virtual models online for product development.
There are, of course, also many opportunities and challenges in using these kind of in silico techniques in practice. And this organization called Medical Devices Innovation Consortium in the US is trying to drive this adoption in the medical devices space and they have asked their members where do they see the biggest opportunities and challenges being in this space.
As for the opportunities, there are a number of different mentions for the opportunities of using in silico medication, but I just want to highlight the two biggest ones. So device development, developing new medical devices using these virtual models and virtual patients, and secondly, conducting these early feasibility studies making sure that your device performs well and has the right requirements in the beginning and later on then reducing that risk by running clinical trials and then you have the wrong type of device, or the device doesn't perform well with the patients.
As for the challenges, there are also a number of challenges that were mentioned. But the two biggest ones were uncertainty and regulatory affairs from regulatory parties. And secondly, the expertise of organizations. They necessarily have the skills, the knowledge, and the expertise. You see this kind of techniques in practice, and therefore it is difficult for them to adopt them and start using in-silico medicine in product development.
But I would like to address these challenges in this presentation and show that actually there is some communication already from regulators on how to report these virtual trials or results in records submissions, and also that the expertise doesn't have to be that high in the organization if you use ready-made models on high-level languages to build these models. And you can actually quite easily, then, start using them in your product development.
Now let's look at the standards and guidelines. What kind of guidelines and standards that already exist for applying in silico medicine for medical device development. The standards guidelines, they are built or created to ensure the credible use of computational modeling and simulation in medical device development. In other words, so that the models that they are using have been verified and validated for the purpose of device development.
There are some industry standards. For example, ASME V&V 40 about assessing the credibility of these computational models to verification of ideas and techniques in application to medical devices. From the FDA, they have also published a number of guidelines for reporting these results. So already in 2016, they published this guideline on how to report these studies in medical device submissions for record approvals, and also in '21, they published a guideline on how to assess the credibility of these models and simulation techniques for device submissions.
So as you can see, there is already some communication among the regulators. So there are standards and guidelines how you can start using these techniques. So that shouldn't hinder the use of computational modeling and simulation because the message from is quite clear. You can start using this and also the guidelines on how to apply them in practice.
FDA has also reason to publish this report on success stories using computational modeling and simulation for recording the evidence. And in this report, they highlight several use cases where in-silico medication was used to support the clinical evidence in regulatory submissions. For example, there is this Virtual Family project whereby different type of human bodies, virtual humans, were created that can be then used for validation of the medical device and making sure that the efficacy is good for the target population.
MATLAB and Simulink were also used in multiple studies in the report. For example, there was this Tobacco public health study whereby MATLAB was used to analyze big data from health care and see what is the effect of tobacco consumption on the public health. For now, let's have a look at the computational modeling and simulation workflows, how you can apply these techniques in practice, and what kind of steps does it involve in order to use in-silico medicine for medical device development.
So what do you need for studies in in-silico medicine for your medical device development? So first of all, you need a medical device model. For example, that could be a model of a ventilator, a model of a dialysis machine, or maybe an insulin pump.
Secondly, you would need a corresponding physiological or anatomical model. For example, if you are developing a ventilator, you would have a model of the patient's lungs. If you are developing a dialysis machine, you would have the model of the kidney. And if you are developing an insulin pump, you would have a model of the pancreas and the blood circulation.
Once you have those two models, you can then start simulating and see how they interact. Does your device have the right requirements? Does it work correctly? Are there any clinical conditions that they account for and start testing the device early on and do these early feasibility studies? And that way, you will have a working in-silico medicine model that you can start using in your research and development of medical devices.
There are several steps involved and is end-to-end workforce for computational modeling and simulation. So first of all, you need to acquire some type of data in order to build your models. You might want to acquire data from sensors, say, whether it's an EEG or ECG, for example. You might have already files containing some clinical data, or you might have some databases of larger patient populations which you can then use in order to build realistic models of medical device.
Secondly, you would then model and simulate these different types of virtual humans and medical device models. And additionally you would, for example, simulate the algorithms or AI models that you want to incorporate into your medical device.
Thirdly, once you have that model in place, a working simulation model, you would then verify and validate that model, making sure that it meets all the requirements in the standards, such as the ASME V&V 40, do some functional testing, and making sure that it complies with the standards.
Lastly, once you have verified and validated your model, you can then deploy it and start using it in your product development. For example, you can use it as a desktop app if you are an engineer, say, working on a desktop environment. You can deploy it on the cloud as a collaborative platform for many researchers or engineers using the same model. Or you can also use the model directly on embedded devices if you want to see how your medical device model works on a hardware device by generating code.
MathWorks has also a number of different products that they can start using for building your in-silico medicine models for medical development. For example, with Simscape, you can build electrical or fluid models, say, of a cardiovascular system, or if you want to build an electrical model of the heart, for example.
With MATLAB, you can do, say, thermal, structural, or electromagnetic models with FEM models that you can then use, for example, developing new implants. And with Simulink, you can integrate everything together and also build medical device models. For example, in this case, there is an example of ventillator model together with the patient lungs, and you can then simulate and see how the ventilator works together with your patient model.
Let's then have a look at a couple of case studies that we have done together with our customers in order to use in-silico medicine for medical device development. But this example is about simulating a medical ventilator. Medical ventilator, it's a device that replaces respiration by automatically moving air in and out from patient's lungs.
It is a FDA class II device. For this type of device, the simulation cost would be, for example, to create and tune a closed-loop control system for ventilation, and secondly, once the system has been verified, to generate certifiable embedded code for the production device.
The simulation model of a ventillator that would look something like this. So first of all, you have a patient model. In this case, it's a passive patient with some body temperature, some moisture in the lungs, and also muscle pressure.
You also have the respirator mask here for ventilating in and out and you have the valves controlling the inhaling and exhaling of the air. Finally, you also have the control system here that recollects the valves in order for the patient to breathe safely.
This way, you have a patient model connected to a medical device model and you can start testing how the ventilator operates in different clinical scenarios and with different types of patients. And this model is available to download also on File Exchange by searching for the medical vendors example in Simscape.
This example is about simulating an ECMO, or a heart-lung machine. So ECMO, or heart-lung machine, it's a device that removes carbon dioxide and sends oxygen-filled blood back to the tissues. So basically bypassing the cardiological pulmonary system of the patient.
It's FDA class II device. The simulation goals for this type of device, for example, would be to build a complete representation of a cardiovascular system, and secondly to verify and validate the ECMO machine using this patient model.
So the simulation model of the cardiovascular system together with an ECMO looks like this. So above, you have the patient model, which is, in this case, the composite cardiovascular system modeled in Simscape. And below that you have the ECMO system that is connected to the patient, to the cardiovascular system to remove carbon dioxide and providing oxygen to the blood circulation. This model is also available to download on File Exchange.
This example is about simulating an artificial pancreas. So artificial pancreas is basically a glucose monitoring system that is combined with insulin delivery. So basically, you have this glucose monitoring system monitoring the glucose levels in the blood circulation and then you have an intelligent algorithm in the device recording the administration of insulin in the blood circulation.
This is FDA class III device. It needs to go through the premarket approval in order to be sold on the market. And there are already several devices like this to be sold to the patients.
The simulation goals for this type of device would be, for example, to create and tune across your control system for insulin delivery, and secondly to generate certifiable embedded code for the production device that can be then sold to the patients.
So the in-silico medicine model of an artificial pancreas looks like this. So first of all, you have a diabetic patient that is eating some meal. And based on that meal, the blood glucose levels are rising. Secondly, you have the artificial pancreas device that is a combination of continuous glucose monitoring system and also an insulin pump that's administrating insulin in the blood circulation.
Finally, you have the control system, the closed-loop controller that is monitoring the values of the glucose levels from the glucose monitoring system, and then calculating what should be the right amount of insulin to be administrated in the blood circulation. This control system could be a different type of model. It could be, for example, a traditional PID controller. It could be an artificial intelligence-based model, or like in this case, it's a quasi logic-based controller. And this model is also available for download on File Exchange.
This example is about the dialysis machine. So the dialysis machine, it removes unwanted waste products from blood circulation. It is an FDA class II device. And in order to simulate this type of device, this would be, for example, to create and tune across the control system for ultrafiltration, which controls that in order to remove waste products from blood circulation. Secondly, again, to generate certifiable embedded for the production device.
Here's a similar version of an in-silico medication model of a dialysis machine. So you have the patient model here with some blood plasma filling rate. And then you have the ultrafiltration system that recreates the pressure levels for predefined base products from the blood circulation. And again, this example is available to download on File Exchange.
So here we have an example of an infusion pump. So infusion pump delivers drugs and nutrition in a controlled manner in the patient's body. It is an FDA class II device. In this infusion simulation, for example, one could verify and validate the pump behavior in different clinical scenarios and in as well. And then secondly, to generate certifiable embedded code for the production device.
The in-silico medicine model of an infusion pump is shown here. So this is a high level overview. There are more details below because this is a higher level system, so more details below of this initial craft. If you open the Simulink model, which you can download from File Exchange.
But basically what we have here is the patient model and infusion pump together under one system. And then we have the controller, the control system for administrating the nutritions and drugs into the blood circulation, and then we have also a command center for manually controlling the parameters of the controller and overriding the other automation to the closed-loop controller.
This example is about simulating an orthopedic implant. So orthopedic implants, they basically replace a missing joint or bone, or they support a damaged bone. These devices, they are either class II or class III in FDA classification depending on the place where they are implanted. The simulation goals for this type of simulation would be, for example, to analyze the stress on the bone or the implant and the wear over time. And secondly, based on these results, one could then determine what are the right materials and the design or the final product.
What we did in this case was to focus on a spine implant. So in the first step, we took some clinical patient data of the spine in CT scan and we used Medical Imaging Toolbox the segment a vertebrae from this data. After the segmentation, we extracted the vertebra from the spine and we did a meshing on it so that we can use finite element modeling to analyze the stresses and wear over time.
And the last step, we simulated what kind of stresses does the spine affect on the vertebrae, and based on this analysis, we can then determine how would the design be for the implant, what would be the right materials for the implant, and also be positioned on the vertebrae, or on the spine in order to have the maximum efficacy for the patient. This example can be found from our documentation on MATLAB website.
So next, let's jump into conclusions, some key takeaways, and show some others where you can learn more about applying in-silico medicine for your projects in medical device development. So as a summary and key takeaways from this presentation, I would want to highlight that in-silico medicine, it uses simulations to verify and validate medical devices using virtual humans or organs. The simulation models can be used in product design, which will result in shorter time to market, reduced R&D costs and time, and safer medical devices to the patients.
Barriers to get started can be lowered with end-to-end workflows relying on the relevant examples that I showed earlier and also technical support from organizations such as MathWorks. The examples that I showed in presentation can be all found on MathWorks File Exchange. So please go there.
Search with the name shown here and you're going to find the example and start using the example in your medical device product development. And that's where you can get started easily by tuning and adjusting the example to your own project needs.
We have also written this white paper on in-silico medication, which shows some case studies from the industry, how other companies have used in-silico medicine for the medical device development, and also lists some best practices for getting started by using these techniques in your own projects. So if you want to get hold of this white paper, we are happy to serve it to you. Just email us at medical@mathworks.com we will show you the white paper.
And for your projects, we don't only support you with software, we also have other services that can be taken advantage of in order to start with your medical device development. For example, we can offer trials and evaluations for you to start using these techniques in your projects. We have also consulting services if you would want to focus on a data part of the project, and then maybe implement that using consulting.
We have also training services available if you want to train your skills in a certain area. Or we have also on-ramps online that you can leverage for free to ramp up with the skills needed, so for example, for Simulink and Simscape. And we have also technical support available online if you face any challenges in using these products in practice.
Finally, I want to thank everyone for listening to this presentation. And if you are interested to learn more about how you can use MATLAB and Simulink for your medical device development product supporting AI, machine learning, signal processing, image processing, please have a look at the website at MathWorks.com/medical. You can find different examples and case studies from the industry on how to apply these techniques in your projects in medical device development. Thank you very much and now we have some time for the questions after the presentation.