Professor Marc Noppen explains the value of AI and an aiOS in clinical systems
In November 2021, the European Commission released a landmark report outlining a framework of policy that member states can implement to accelerate adoption of healthcare artificial intelligence.
Since the report’s release, a number of institutions have embraced AI as a solution to help tackle challenges plaguing healthcare in the continent’s 27 member states – including specialist shortages and increasing patient data volumes. In the process, many of these institutions are becoming AI pioneers in their own right.
One such institution is Universitair Ziekenhuis Brussel (UZB), a 729-bed facility that admits more than 50,000 patients and treats over 400,000 patients annually. In January, UZB’s CEO, Prof. Marc Noppen, signed an AI enterprise agreement with Aidoc to integrate the aiOS™ (AI operating system) into UZB’s clinical mainframe, enabling radiologists and neuro and cardiovascular specialists across the health system to benefit from AI.
During a visit to the Aidoc offices in Tel Aviv, we sat down to chat* with Professor Marc Noppen, CEO of UZB, about European healthcare challenges and the opportunities of AI.
What do you consider to be your institution’s main challenges, and how do you think AI can help?
I think the main challenge for now, and in the years to come, will be the human factor in medicine. There is a hyperinflation of all kinds of data that should be processed by clinicians, which poses a real challenge as there is a limit on the human capacity that we have in hospitals.
Everything that can help to reduce the workload for the clinicians, and at the same time, improve efficiency and quality, will be helpful. What interests me, for instance, is a product that can eliminate all the normal images so radiologists can focus on the abnormalities, because the number of radiologists we have is limited. AI may be such a tool.
There are many other domains where AI can support, including in population health (health promotion and disease prevention) and disease management, which is also very interesting for the sustainability of healthcare.
But with AI, whatever it is, it depends on what purpose it will fulfill. And it’s up to AI providers and healthcare providers to work together to define the horizon for the foreseeable future, and where we want to go with AI in healthcare.
As you said, radiologists have been typically early adopters of AI in healthcare, but you also mentioned that AI has potential beyond radiology. What is that potential?
It’s in this field of diagnostic support where AI has played a critical role. But what comes after the diagnosis might be even more important in creating collaboration between radiology and the other disciplines.
Looking at cardiology, we were working with Heartflow for measuring fractional flow through the coronary arteries. Just recently, they finished a study where the decision whether or not to go to a coronary artery bypass graft surgery would be based on this data instead of coronarography images. Right now, there is an evolution going on in many disciplines, even with dermatology, for example, where AI could theoretically assist with the diagnosis of skin cancer, as well as in pathology.
So to put it all together, a radiological finding will likely be paired with biological findings, patient history, and NLP on medical files to draw broader conclusions about a patient’s medical condition. Or at the very least, we can say, “let’s come to a proposition of next steps for the patient,” which is built on data, and hence, something that remains to be proven, can improve outcomes. And isn’t that the ultimate goal of healthcare? To improve outcomes?
It’s arguable that, for AI to become impactful at scale, there needs to be support from management in a hospital. How do you see the balance between a bottom-up approach driven by individual departments or service lines who champion adoption of AI, and the top-down approach driven by what’s called the “enterprise” approach?
It starts first with the movers and believers. In UZB, the radiology department has always been on the forefront of new developments. For instance, the concept of PACS was developed at UZB, which is where a radiologist used PACS for the first time, 30 to 40 years ago. So radiology was always on the forefront of innovation and finding new use cases. From there, I believe the openness and championing of AI will bubble upwards. Then it’s up to the hospital management and policymakers (and payers!) to help promote adoption.
The bubbling upwards doesn’t happen independently; it happens in tandem between healthcare providers and companies providing AI. If the two cooperate well and learn from each other, it’s the only improvement cycle – the Plan-Do-Check-Act (PDCA) cycle – you can really afford to go by. But it’s up to governments or payers, and let’s say also the management at healthcare facilities, to allow for the framework to empower this cycle to take place, without interfering too much.
Ultimately, if AI becomes an element of care or the standard of care, it is the responsibility of the care payer, whomever that may be. Whether it be a government body, such as the NHS in the UK, or private insurance or a mixture of both, like in Belgium.
For instance, you could compare this technology to imaging machinery. We have a LinacMRI radiotherapy machine. You could ask, “Who should pay for this?” Because it’s a very expensive machine. In a way, social security will pay for that as a tool that’s innovative and new.
The same goes for AI. AI will have to be included in the package of technologies for clinical care the same way MRI and CT machinery is. Of course, the healthcare payment model varies from country to country in Europe. But at the end of the day, I see no difference between who would pay for AI and who would pay for a CT machine.
*This interview has been edited for length and clarity.
