AI and medical device manufacturing: How one affects the other
Ian Bolland spoke to Ashley Ross, regional business director at Medtronic, to discuss the many issues surrounding AI including its quality, its uses and the relationship between AI and hardware manufacturers of medical devices.
Artificial Intelligence has been trumpeted as a tool that can solve many issues but there is a lot of work that goes into its development. Ashley Ross started off our conversation explaining that AI and healthcare will evolve together, saying that despite many looking at it as something of a silver bullet that can solve all the problems in the sector, it is still in its infancy.
“AI is going to adapt as it learns more effectively from working within healthcare. From healthcare’s perspective it is just getting used to data management and what it does with data when you think about rules, laws, regulations or even just getting comfortable with it. Something we notice as a company is some people absolutely love the concept and are geared up for putting it in, and other people are rightly wary of it and are seeing how their world changes.”
As AI’s influence continues to increase, there is inevitably an effect on what medical device manufacturers – including those who had developed innovations prior to its existence – must consider when developing the physical devices, as well as the considerations around data security, along with the technology to ensure the devices are well updated.
Ashley Ross refers to Medtronic’s Reveal LINQ and LINQ II, implantable cardiac monitoring systems, which are worn underneath the skin, when explaining how both hardware and software can converge to realise their full potential.
“We are a hardware company historically, but we're bringing AI in.
“Reveal LINQ has been around in one iteration or another for the last 20 odd years, and the way in which we've iterated is, let's improve the hardware the way the hardware is being designed now is to allow the improvements to come from software, particularly from AI. So, it’s moving technology into cloud-based management of data, which then means we can roll out updates to the technology – in the same ways as you do a software update on your phone.
“We can change the performance of our device just by enhancing the AI on the back of it. So, we’re a hardware company that's seeing the benefits of AI to accelerate the development of how that device is used.”
Ross adds that it depends on the strategy of the manufacturer in terms of what the advance of AI means for the future of hardware devices, but also what medical device developers want to do with the AI at their disposal.
“We're taking the approach where the hardware gives us the capability to effectively update the AI or the software remotely. We never used to have two-way communication with the device like and they'd physically have to bring a patient back into a clinic to update the hardware.
“The rate at which AI is going to evolve will also be quicker than what we've evolved hardware because of the speed at which it can learn and the things that it can detect."
With companies increasingly trying to find ways to present a more sustainable image, Ross says the potential for AI in the sustainability agenda is huge, with physical devices lasting longer as updates to them means there is less need for parts to be disposed of, while allowing the company to capitalise on other opportunities within the space.
“It's when you start thinking about manufacturing lines and everything that follows. If we get to the point where we're trying to get to on the diagnostic side where you can have a baseline bit of kit that you're developing. So, the baseline bit of hardware, then as the AI develops, it allows us to then start to move into other areas, not just diagnostics but into disease state management and starting to understand morphology of ECG traces and what we can see within that underlying trace.”
Central to AI’s success, whether that be commercially or whether it’s helping the NHS tackle significant backlogs, is good quality AI – so how is that defined? Ashley Ross summarises by saying it depends what it is being taught on.
“For our latest iteration of AI and Reveal LINQ or LINQ II there's a million unique ECG traces. It's gone through a million different sort of arrhythmias that could pop up and it's learned off the back of that and then that's been validated. But then it also gets stopped. So, the learning stops and it's held at a certain point.
“It works well. It's determined to work well, and we can trust it because it has seen so many different variants. And to us, that's good quality. It's high volume of data, rigorous validation and it's been stopped from learning, so it doesn't start to learn bad habits.”
Champions of AI have often said it can be a way to address healthcare inequalities, so how does AI help redress the balance?
Ashely Ross concludes by saying: “It allows us to effectively update the kit in an instant. When we bring AI or AI iterations to our technology, now it comes automatically. There isn't necessarily a cost associated with that.
“That allows a hospital that is currently using our kit to then benefit from AI and future iterations of AI. It's cost neutral and if anything in the long term, it should make us more cost effective, which will then have a pass on effect on our healthcare system. So, I see it that way.
“The other aspect is volume. One of the biggest challenges right now is waiting lists and the ability of staff to interpret data and then work with that data.
“AI has the potential to massively speed up the diagnosis and interpretation of data; meaning a greater number of patients can move through a healthcare system at a similar cost to what the healthcare system is currently running at. That is a bit of a leveller.”