Why medical device manufacturers must become tomorrow's software innovators

Today’s medical devices are far more than their hardware components. And even small mistakes can have big implications.

The software built into medical devices is arguably one of the most important sources of any manufacturer’s competitive differentiation because it governs the increasingly complex functions and processes behind increasingly commoditised hardware. Mistakes such as errors in code can necessitate warning letters from the FDA or even lead to widespread product recalls with substantial costs to both a manufacturer’s brand reputation and the bottom line. Even error-free software must conform to the compliance and regulatory regimes in every distribution market - a process that can send even the most technically rigorous developers back to the drawing board.

Medical device manufacturers must embrace the tools of software development to overcome these hurdles and make real innovations possible. The FDA references one such problem-solving approach - model-based design - as an effective software engineering method, including for medical device applications. Some Australian device manufacturers have already adopted model-based design within their core processes, but many more tend to stay with their traditional development approaches and face the risk of being left behind by their competitors.

Rapid prototyping: a software developer’s scalpel

Model-based design involves using modelling and simulation software to test and compare how different algorithms will work. Rather than having to build physical prototypes to gauge the effects of their algorithms, clinical personnel can do so within modelling platforms like Matlab and Simulink, developed by The MathWorks. With model-based design, developers can create mathematical models of their process dynamics and then simulate how their algorithms perform while interacting with these modelled systems. Within this same model-based design environment, the software implementation of these algorithms can then be tested in real time, allowing embedded software engineers to further test and verify their code and strip out any errors far earlier and faster than ever before.

The time and cost efficiencies of such an approach are obvious, but model-based design’s real value to device manufacturers stems from the ability to engage in rapid prototyping. Because tests no longer require expensive physical prototypes (a single one of which can cost in the tens of thousands of dollars), developers can test out every design option that’s tabled instead of being able to trial only one or two. And they can do so within minutes, allowing them to cut through all the options and find the most effective solution in the least amount of time.

Rapid prototyping is not new to Australian medical device manufacturers. Cochlear’s engineers have for years relied on Matlab and Simulink to quickly design, test and reiterate different algorithmic options for their iconic implants, using the modelling software to test up to six times more options than with traditional prototyping methods. This means that more innovations can be converted into real-world applications and be deployed to market more sustainably and more quickly than ever before, allowing Cochlear to maintain its competitive edge in an increasingly contested market.

No need for manual coding

One of the fundamental benefits that Cochlear experienced from a model-based design approach was that it allowed the clinical professionals with the algorithmic expertise to also stay engaged during the prototyping process - an advantage made possible via the visual nature of the Simulink model-based design environment. Rather than being hindered by the need to manually code in a lower-level language, clinical staff were able to quickly create and test algorithms using a graphical programming interface. The model-based design environment then automatically converted these algorithms into the machine-language code that ran on the prototype devices, ensuring that errors were not introduced between the design and prototyping implementation stages.

Some platforms like Matlab also offer predesigned application centric toolboxes and algorithms to further accelerate the development process. Instead of having to ‘reinvent the wheel’ - and potentially introduce code errors that may go unnoticed until it’s too late - manufacturers can focus on new innovations and ideas that use pre-existing functions as a foundation to accomplish clinical goals.

When developing its AirSonea app for asthma sufferers (recently released in Australia), engineers at iSonea used Matlab to automatically generate the C code that would eventually be used in apps for iOS, Android and the cloud. Not only did model-based design speed up the design process, but it also dramatically reduced the complexity of debugging and optimisation: the team knew that any performance issues were a result of design rather than errors in code conversion.

Collaborative innovation

Designing a successful medical device is, however, more than simply a matter of technical efficacy. Model-based design’s most powerful application could be as a common platform for the myriad silos involved in the development process.

This is particularly relevant as medical devices become reliant on patient data and predictive analytics, both of which require close collaboration between clinicians and data scientists. The ‘softer sciences’ of biology and chemistry are increasingly demanding platforms that can learn and make predictive diagnoses based on previous cases. The latest pharmacokinetic simulations rely on analytics technology to simulate the results of different medicine dosages or their effects over varying timeframes on a specific patient. But, if any insights from data are to benefit patients in the real world, they must be translated by algorithmic or human analysis into applications and physical devices.

Model-based design environments bring together these insights and applications, providing full visibility to teams at every stage of development. Clinicians can access analytics results and feed them into algorithmic models in real time. Data experts can not only use them to design analytics algorithms, but also to review how their insights are being applied and to assess if the algorithms’ results are sound. By providing what is essentially an open ‘sandbox’ for modelling and design, these frameworks provide a central platform between the different disciplines involved in medical device workflows.

This increased collaborative potential, coupled with the ability to test ideas faster and more rigorously than ever before, promises to turn medical device manufacturers into an industry of software innovators. The resulting devices will benefit communities in Australia and around the world.

*Lanfrey is a senior application engineer with The MathWorks. He holds a master’s degree in electrical engineering from the French engineering school ENSIEG with a specialisation in control engineering. Prior to joining The MathWorks, he worked for five and a half years in the automotive industry: at PSA (France) his role in the Engine & Powertrain Control was to design, test and improve control laws; at Air International (Australia) his main focus was the development of embedded software, especially the algorithms controlling the heating, ventilation and air-conditioning system. Then he worked at Parrot (France) for one and a half years where he was responsible for the design of the embedded software that controls a quadrotor helicopter.

*Horton joined The MathWorks Australia in April 2006 and currently holds the position of Engineering Team Manager. He has spent the last 15 years helping clients adopt and implement MathWorks products over a broad range of application areas. As a principal engineer, Horton has supported and consulted for clients on projects ranging from process control engineering, power systems simulation, military operation research and business intelligence analytics. Before joining The MathWorks, he spent five years as a systems engineer with the Defence Science & Technology Organisation (DSTO) working as an operations research analyst. Brad also worked for the former Australian MathWorks distributor as both an application and consulting engineer. Brad holds a B.Eng. in mechanical engineering and a B.Sc. in applied mathematics.