Barriers to deploy AI in Healthcare
This week, we share an interview with Michalis Papadakis, CEO of Brainomix.
Panacea: In the last few years, we see an influx and hype around artificial intelligence (AI), medical imaging and diagnostics. Do you see this as a bubble or is it still healthy within your domain?
Michalis: At Brainomix, we are solving a real clinical problem: the millions of stroke patients who are missing out of life-saving treatments because they get admitted to a hospital where there isn’t an expert available to correctly diagnose the stroke. Medical imaging is a core part of making a diagnosis and identifying patients that can benefit from life-saving stroke treatments. To address this problem, we have developed the world’s most comprehensive AI imaging software, the e-Stroke Suite, which can improve the diagnosis and treatment of stroke patients. Our technology has already been used in more than 200,000 stroke patients in real clinical practice, and it is continuously growing. We have a huge clinical, and commercial opportunity to expand this AI platform to other imaging modalities and disease areas and there is still a lot of value to be realised.
A bubble, however, may lie for some newcomers in the AI imaging space who are underestimating not only the technological barriers, but also regulatory hurdles as well as integration challenges within a hospital setting. We are fortunate that we have gone through and understood the difficulties, but they are not trivial and maybe overcasted by the hype.
Panacea: Do we need to wait for it to be a big enough bubble before AI-based software becomes the norm and be regulated?
Michalis: AI-based software is already well-regulated and we don’t necessarily need to make it more difficult, but we need more structured approaches for getting market clearance for AI software as a medical device. There is a continuous revision of the medical device regulatory guidance and we are expecting the new EU regulations for medical devices coming in May 2020 to make the landscape even more challenging.
“…we need more structured approaches for getting market clearance for AI software as a medical device.”
Panacea: In a recent BBC interview, you mentioned that “the biggest challenge is getting doctors to adopt the new way of working”. How far do you think it is for medical imaging software to be included into the clinical training phase?
Michalis: We are fortunate that our technology is used in routine clinical practice. When we first launched the product as a decision-support software, there was skepticism from physicians. It is a long process to build trust with physicians, reaching out to key opinion leaders and large university hospitals and building the data supporting the performance and clinical utility of the technology. The infrastructure for training is ready: we have collaborated with the pharmaceutical, Boehringer Ingelheim, through the European Angels Initiative and we have developed a tool that is used to train physicians, from junior doctors to expert consultants, to read brain scans of stroke patients.
“It is a long process to build trust with physicians, reaching out to key opinion leaders and large university hospitals and building the data supporting the performance and clinical utility of the technology.”
Panacea: How do you feel being an entrepreneur in Oxford?
Michalis: AI medical imaging is an exciting field. What we do transforms lives, makes a difference to the patient and gives them a better chance of leading an independent life. Being part of the Oxford ecosystem means that we have excellent resources, technology and investment, in addition to the international access through the credibility we have as a spin-out from the University of Oxford.
About Dr Michalis Papadakis:
Dr Michalis Papadakis is the CEO and co-founder of Brainomix, a University of Oxford spinout company which has developed medical imaging software (e-Stroke Suite) to automate the assessment of stroke damage on CT scans, which can improve the selection of patients for life-saving treatments.
