Interview with Dr. Byrne (AI in GI): “There’s a tidal wave of AI coming and we should prepare for that now”
Dr. Michael Byrne is CEO & Founder of Satisfai Health, an artificial intelligence company that aims to deliver the future of gastroenterology. Dr. Byrne is also the founder of ai4gi, a joint venture with Olympus that focuses on AI solutions for colon polyp detection and differentiation. He’s also an interventional endoscopist at Vancouver General Hospital.
I spoke to Dr. Byrne to understand where exactly AI is in gastroenterology. What must private practice gastroenterologists do now? What does the world of GI look like post-COVID? Stay on top of GI trends by watching this thought provoking interview in full.
◘ Dr. Byrne’s various roles in artificial intelligence in gastroenterology
◘ “Will endoscopists who use AI replace those who don’t?”
◘ Patients will ask, “Is that group using the latest technology?”
◘ What if we are mandated to reach an ADR of 40% to get paid
◘ Remember EUS? Lot of people said it would be niche and it won’t take off
◘ Remember NBI? Initially, people thought it was a gimmick
◘ Would DNA testing or liquid biopsy replace endoscopists who use AI for detection?
◘ How many AI initiatives are underway in GI?
◘ Medtronic GI Genius. Which other companies are making progress?
◘ EndoBRAIN received approval in Japan 18 months ago
◘ How far are we before we start using AI in GI private practice?
◘ What would GI look like post-COVID?
◘ AI can help with clinical trial recruitment in IBD
◘ Can AI do a stand-alone diagnostic in optical biopsy?
◘ How can private practices pivot and take advantage of these technology trends?
◘ Whether you like it or not, it’ll be there in the not too distant future
◘ AI in GI space is beyond endoscopy. Lots of algorithms in IBD, liver disease, patient meta data, predicting cirrhosis and the various -omics.
◘ Scope Forward encompasses a lot of things GI physicians need to know
The Transcribed Interview:
Praveen Suthrum: Dr. Michael Byrne, I want to welcome you. Thank you so much for joining me today. You’re an expert in Artificial Intelligence (AI) in Gastroenterology (GI). You’re the CEO & Founder of Satisfai Health, a company that’s in the AI in GI space. You’re also the founder of ai4gi and an interventional endoscopist. I’m sure I am missing a few hats that you wear. So, it will be great if we start with you telling us about all the different roles that you play as far as AI in gastroenterology is concerned.
Dr. Michael Byrne: Yeah. Thank you, Praveen. I appreciate this invitation and the opportunity to talk about AI and endoscopy. So, as you said, we all wear many hats. So, my hats include being the CEO and Founder of Satisfai Health and I have also founded a joint-venture which we have named ai4gi that I think is reasonably well-known in the GI world of endoscopy. Our vision at Satisfai Health is to become the leading group in AI solutions for detection, diagnosis, and treatment of GI disease, particularly in endoscopy and that’s really where we are putting most of our attention.
So, I am involved in AI at an academic level as you know, I write a lot of leading articles in the GI Journals. I present at the symposia, DDW (Digestive Disease Week), UEGW (United European Gastroenterology Week), all the main GI conferences. Our group and I, we are working at the research and development level for providing solutions in the AI space and also at a commercial level where we are trying to forge increasing relation with the industry. So, I guess that’s a reasonable summary of where I am right now.
Praveen Suthrum: I came across a recent editorial that you had written talking about whether AI in GI particularly in optical biopsy is a hype or reality. You ended that piece by saying that “endoscopists who use AI will replace those who don’t”. So, you know I wanted to take you up on that and ask you to explain what you meant there.
Dr. Michael Byrne: Yeah. I guess you have to be careful with the statements that you make so that you don’t make too many enemies in this space talking about replacing endoscopists! But you know tongue and cheek, there’s a lot in that statement. I do believe that people have to embrace new technology. There are several layers to this answer. There’s a competitive spirit among professionals so, we all want to use hopefully the latest and the greatest technology. There’s curiosity from the public who know all about AI now and it’s benefits and maybe the pitfalls of what AI can bring. And I think going forward, we will probably have mandated performance metrics from the various payors and the regulatory bodies, to look at how we are performing on a day to day basis.
If you use CADe or Computer-Aided Detection as an example, there are several studies that show that this AI-type tool improves the ADR or the Adenoma Detection Rate or in other words looking for precancerous polyps. And several groups in the US, a few years ago, I think it was what was advertising in a competitive fashion, that their group was using the “FUscope”. If you may remember in your previous coverage that the FUscope is the one that had some side lens as well as a forward lens and the idea was that it increased the field of vision of the scope so that you could see more. And, there were some studies that showed that it can out the regular forward-viewing scope from the competitors that could increase the ADR or the Adenoma Detection Rate. So, there were some groups that were advertising to the public: ‘Our group uses the FUscope, this has been shown in studies to improve the pickup rates of polyps, you should come and have your procedure with us’ or in other words, it brought on a competition between groups. So, I think the same will happen with AI. Public who have been availing these services and particularly in healthcare systems like the US will look and see, ‘Okay is that groups using the latest technology? I have heard that AI can help improve the human mysteries’
As I hope and as I expect in the next few years, it is mandated that we reach a certain minimum threshold for performance. So, again I have used the example of the obvious one, colon polyps. That’s the one that most people can relate to. If we are mandated to reach an ADR of say, 40%, I’m going to pick a figure. If you are not reaching that level, by whatever means maybe you won’t be reimbursed and/or your privileges for that procedure will not be renewed by your healthcare institution. And that is maybe the way to drive increasingly better performance. To the naysayers, I’d say remember EUS or endoscopic ultrasound? When that first came out, properly 15-20 years ago, a lot of people said, ‘oh it’s black and white’ ‘it’s going to be niche groups’, ‘it’s going to be very hard to interpret’, ‘we’re not radiologists’ and ‘it won’t take off’. Now you can see it is in every major healthcare institution, driving lots of interventional endoscopy.
Remember the same with the NBI or the narrow-band imaging, the blue light on the endoscope that Olympus got initially and all the other companies like FUJI and Pentax have developed a version of virtual chromoendoscopy since. Initially, NBI was almost felt to be, how can I put it politely…a gimmick or something that most people wouldn’t use because they didn’t know how to interpret what they’re looking at. That’s not the case now. It is driving this visualization of pathology. So, back to your question, will people who don’t use AI get replaced with those who do? Maybe not replaced, but I think, people need to realize that very quickly they have to adopt, get on the train. It’s moving, right? It’s leaving the station.
Praveen Suthrum: Would DNA testing replace or liquid biopsy replace even endoscopists who use AI for detection or for screening? You know, it could be, because if a blood sample is going to tell you whether the patient has cancer, does not have cancer and it’s not doing just for GI-related cancers but say we get to the point where it does for 15 different types of cancers. Then, would they even do an endoscopy on such a patient? And where would that leave some of the AI initiatives which are underway right now?
Dr. Michael Byrne: Now, I think it is a very good point. I think that the scenario that you mentioned that the liquid biopsy, the blood test being very accurate and predictable of certain types of cancers is quite a way away. I mean, I know we are getting lots of biomarkers now for cancer, but I think expecting that next year we are going to have (for example), a blood test that is very accurate for determining your colon cancer risk is aspirational at this point. That doesn’t mean we’re not going to have better and better non-invasive ways to look, whether it’s with imaging, capsule endoscopy, blood tests, or the genetic tests looking for your own risk of colon cancer for example and that, that’s not going to come. I’m sure it is. And maybe that will also help to streamline endoscopy.
So, endoscopy becomes more therapeutical, almost as if you have assessed that somebody has the risk, you’ve assessed that somebody has non-invasive imaging to have polyps, (again for example) on a capsule and then you chase it with your endoscopic procedure to remove those polyps. So, yeah it may take away some of the excessive volumes of screening in endoscopy. It may promote even more appropriate intervention and therapeutic endoscopy or colonoscopy. Will AI play a role there? We still need to see these things when we’re looking for them so, humans still need help. And I feel that the thing we haven’t touched on it in this talk so far is the field of optical biopsy or doing a virtual or real-time pathology evaluation with your eyes or the AI. That will be needed to be aided by technology. Once we find the lesion, we need to know what we’re looking at. So, yeah, very good point but none of these things should be threats. These could all be cooperative interventions, right?
Praveen Suthrum: Okay. So, last year when we spoke for my book, Scope Forward I had asked you how many AI initiatives are underway in GI and you had mentioned that at least a hundred. So, I wanted to ask you how accurate have you been on that prediction?
Dr. Michael Byrne: Well if you look at the publishing, those who are publishing or trying to publish in the main GI journals, then the number must be way more than a hundred. If that’s how we define a group. In the last month alone, I think I have reviewed maybe ten articles to do with AI in endoscopy for the main GI journals and I’m just one reviewer. So, if I’m reviewing ten papers in the last month on AI alone, and they’re all from different groups! You can imagine the number of groups globally who are actively working in this space. I would say probably several hundred groups, looking at this from an R&D or at least from an academic perspective. How many of those are true companies or groups with all the facets, that’s a little hard to define.
Praveen Suthrum: I saw the announcement from Medtronic related to GI Genius. They launched in Europe and they were supposed to launch in the US this year. I’m curious to know if there are other companies that are ready for primetime. Have things been launched already? Can private practices particularly take advantage of any of the AI solutions which are available out there?
Dr. Michael Byrne: So, you correctly mentioned about some approvals for AI in endoscopy in Europe. With the FDA, it’s still a little slow but we’re getting there for sure. There was a thing… I think it was called the ‘First Global Proceedings for AI meeting’, it was in Washington DC in September last year I believe and there were representatives from the FDA, and from the NIH and global thought leaders in this space were there. A lot of these big tech companies – Amazon, Google, and Microsoft people were at this meeting. So, it only got some attention (there) and with the FDA I think things are going to move quite quickly going forward. But as you said, there are some groups that got their CE mark in Europe for detection. So, CADe or Computer-Aided Detection, again in colon polyps, that’s a disease where lots of efforts have been made so far.
So, they include Medtronic with GI Genius, FUJI recently announced that their CAD EYE or their REiLI system has also got CE approval for colon polyp detection or CADe. As have Pentax with their discovery system. I know that Olympus is working incredibly hard in this space too and with optical biopsy or rather CADx but for confidentiality reasons, with my own involvement as I mentioned maybe at the beginning that ai4gi has a co-development agreement with Olympus in polyp and AI. So, I do know Olympus’s plans to a large degree, but I can’t really say any much more right now. But it’s definitely moving very quickly. For CADx or Computer-Aided Differentiation or doing a virtual biopsy or virtual pathology, there is no FDA approval yet to my knowledge. Other than this group by the name of NinePoint Medical, I’m sure you know them. They are doing some great work with VLE or Volumetric Laser Endomicroscopy, mainly in the esophagus (for example) and they have an FDA approval for using an AI tool, the image feature segmentation. But it’s not true optical biopsy. So, more advanced CADx or optical biopsy is still awaited emulation for an FDA approval actually even for a CE as well.
There’s a group in Japan, called EndoBRAIN, who have done a lot of work in the last number of years on CADx or optical biopsy and they did receive an approval from the Japanese version of the FDA called the PMDA (Pharmaceuticals and Medical Devices Agency) about 18 months ago for AI support, for optical diagnosis of colon polyps. So, hopefully, that also sets the scene for other jurisdictions. I think it is important Praveen, to point out that many of these tools, CADe, CADx, detection, differentiation… at the beginning appropriately they should be and will be clinical decisions support tools to help the physician, rather than standalone diagnostics. Because right now we’re not in the space where we can say for certain that, that’s what AI can do. It needs to be an aide to the physician right now. But I think things will change quite quickly.
Praveen Suthrum: So, just to clarify, can a private practice gastroenterologist in the US play with some of these tools as of today or no?
Dr. Michael Byrne: Not right now, because there’s no AI tool in the endoscopy space other than the one that I mentioned with NinePoint Medical for this very sophisticated technology, this VLE device where the AI allows looking for certain features, other than that for colon polyp detection or colon polyp differentiation, there is no FDA approved device as it stands.
Praveen Suthrum: How far are we before the day where they can start using it or at least testing things out?
Dr. Michael Byrne: You know, I’m sure I don’t have an inside track to the FDA processes and what all the groups are doing in terms of their regulatory pathways. But I think, given what has happened in Europe and looking at the huge amount of work that is being published in the journals and the clinical trials that are listed on the clincaltrials.gov website, I think it’s only a matter of short time that the FDA will approve some of these AI devices. So, my best guess is that AI guided detection for polyps will probably be on the market for physicians to use sometime next year.
Praveen Suthrum: Okay. So, as you know we’re in the middle of a pandemic that’s not over yet. And what I have noticed since the beginning of COVID is that a lot of trends that were already underway have accelerated. And these could be business trends or technology trends, be it the number of AI initiatives. So, I want to ask you what kind of a GI world do you see post-COVID?
Dr. Michael Byrne: Well, the very obvious one is what you and I are doing today, we’re having a zoom meeting. And I think with respect to many of my colleagues in my group here in Vancouver, probably didn’t even know what zoom was three months ago. So, you know many physicians are not particularly tech-savvy but it’s very incredible to see how quickly endoscopy for large meetings have gone virtual. So, people are getting more familiar with technology.
As you said, COVID had a lot of significant downsides, a lot of heartache, and disease burden which is very sad, and hopefully, we’re coming out of it now. But a lot of groups focus as you say, in GI, on the COVID effects from the clinical standpoint, to name a few – ‘How COVID impacts patients with Inflammatory Bowel Disease (IBD)’ ‘should we stop or start new biologics?’. ‘The PPE use in endoscopy’ you know, it became fairly quickly clear that endoscopy was a risky procedure for transmission of COVID because of aerosolization for example. ‘How do we triage patients after lockdown?’ So, all of this came out of the recent pandemic. Can AI help in these COVID related situations? I’m sure the answer is yes. We have now huge backlogs for colon cancer screening. There are several papers coming out now showing that the burden of disease in the last three months is really quite alarming. We have to know how best to come out of that. How do we triage these patients? Who gets done first based on urgency and disease likelihood? I’m certain that AI can help to decipher some of that for us. It can see patterns that are hard to see from a traditional regression type model.
Live endoscopy causes which used to be of course mostly people at that facility but transmitted globally that’s all virtual now and is going very well. So, I think this is just going to increase the appetite. We need help with improving clinical trial recruitment overall, in IBD, in other disease states. We need to maybe rely less on human interaction all the time. So, we talked in the last few minutes about CADx or optical biopsy, if in time AI can do a truly standalone diagnostic in an optical biopsy, maybe we can take away the need for an expert pathologist to do some of that very high-volume but very low-hanging fruit work and leave them to do the more difficult work and more important work for cancer diagnosis.
This is more of your expertise than mine in the venture capital and the private equity world. Of course, there has been a huge economic downturn in the last few months and is likely for the foreseeable future and my impression, I’m not sure of yours, is that there is an appetite from such VC and PE groups to look for safe havens or growth havens for their investments. And healthcare has often been a safe haven even in crisis and I think now, groups are seeing that healthcare technology including AI is a safe haven and actually almost certainly a growth haven. So, I think all of these factors will just help promote a quick adoption of AI into our practice (hopefully).
Praveen Suthrum: My final question Dr. Byrne to you is – So, for practices who understand these shifts, they’re seeing that AI is coming, they see these technology shifts coming but then they are currently tied to an older business model where they are doing a lot of endoscopic procedures, seeing patients in the office, but they also see a shift saying all this is going digital. Now, based on everything that you know, what advice do you have for them, how do you pivot and take advantage of what’s happening?
Dr. Michael Byrne: So, again that is a very insightful question. It is tough. There are naysayers out there or there are other people who are slow adopters or may seem to some degree threatened or that they don’t need to pivot their practice quickly to technology such as AI. And we do need to get physician acceptance, physician familiarity with AI so that we can feel comfortable and pass on that message to our patients, right? So, do people need to pivot straight away? Probably not, I mean it will take leaders in the field to have clinical experience in the field in the next year or two or three with all of these tools that will be coming to our hands quite quickly. Beyond that, do people need to prepare right now? I’ve used a phrase in one of my editorials that there’s a tidal wave of AI coming, and we should prepare for that now.
I do believe that’s the case. AI is pervasive in all aspects of our life. Every time you pick up your phone or use Siri or Alexa, every time you search for a flight it is all AI guided as you very well know. AI in genomics, what used to take humans months or longer to find, an AI model can decipher in minutes or hours. If you read any GI journal in the last 18 months or any medical journal in the last 18 months, I would challenge you to find an issue without an article on AI. So, it all shows that this is all coming very quickly. Look in particular at the Pharma and the device medical industry, they are all focusing on better optics, better visualization, new light technology, better resolution, knowing fully that the human eye cannot appreciate all that new data being thrown at them and so that you will need some sort of intelligence built into their endoscopes. So, the OEM’s, the device manufacturers, know this is coming and they’re investing heavily in this space. So, if you’re looking from the outside in, and you are feeling threatened and you feel like you don’t need to bring AI into your practice, maybe that’s okay for now, but in the not-too-distant future, in the next year or two or three, it’ll be there whether you like it or not.
Praveen Suthrum: I wanted to ask you if there was anything else that you wanted to share?
Dr. Michael Byrne: No, I think we touched upon all the main points. Of course, you know, AI is beyond just endoscopy. AI in our GI space for my GI colleagues listening or he medical colleagues, there are lots of predictive algorithms for re-bleeding risk in patients with peptic ulcer disease, for who will respond to certain biologics, in IBD, in liver disease, without using invasive technology looking at bio-data and looking at patient metadata, clinical data, predicting with great accuracy how likely is somebody with viral hepatitis is to get cirrhosis in 12 months’ time. Lots of AI models are showing huge potential in these spaces and we could spend the next half an hour Praveen, if we had that time today, talking about the various omics – the genomics, the transcriptomics, the proteomics, the metabolomics, and the microbiome and the effect of that on GI disease and how AI can help us decipher all those interlacing signals. But again, that would be a very separate and long conversation.
Praveen Suthrum: Yeah. That would make a good part 2 of this interview. But thank you so much for taking the time today.
Dr. Michael Byrne: Praveen, I would just like to say that I have been following many of your publications and your writings in the last couple of years. Your blogs which have been really helpful to show the business side of GI but also where technology is going. Scope Forward which I know is coming out this summer 2020, seems to encompass a lot of things that we as GI physicians need to know. What is the business angle in GI, how does the current crisis that we have come through affect our practice, where’s technology going? All of these things are very important to know and I think you’re doing a great job to encompass it in one space for busy physicians like me to digest it quickly.
Praveen Suthrum: Thank you so much, Dr. Byrne. Thank you for saying so and for taking out the time today.
Dr. Michael Byrne: You’re very welcome. Thank you.
By Praveen Suthrum, President & Co-Founder, NextServices.