This week I am talking to Dror Zur, CEO at Magentiq Eye where they are bringing the new field of deep learning and AI to the reading and processing of Colonoscopy screening.
Magentiq Innovation
Magentiq has developed an add-on system that enhances the efficiency of colonoscopies by aiding physicians in detecting polyps and adenomas. This technology is particularly important because colonoscopy remains the gold standard for colorectal cancer screening. The system processes video output from the colonoscopy procedure, using deep learning to identify and highlight potential polyps for the physician, significantly reducing the miss rate and improving overall detection rates.
The clinical validation of Magentiq’s system has shown remarkable results. Through extensive testing involving 952 patients across 10 centers in the US, Europe, and Israel, the system demonstrated a 37% increase in the average number of adenoma detections, a 26% increase in the adenoma detection rate, and a nearly 50% reduction in the miss rate. These outcomes make a compelling case for the adoption of Magentiq’s technology in medical centers, potentially revolutionizing colonoscopy procedures and improving patient outcomes.
Listen in to hear Dror’s view of the future and Magentiq’s focus on further refining and expanding its technology. They are currently working on a version that includes polyp characterization and size estimation, aiming to enhance the quality and precision of colonoscopy procedures. Additionally, they are exploring the possibility of expanding their AI capabilities to assist in diagnosing other GI issues and specific diseases. This innovative approach to colonoscopy procedures holds the potential to save lives and improve the quality of healthcare delivery in the field of gastroenterology.
Listen live at 4:00 AM, 12:00 Noon, or 8:00 PM ET, Monday through Friday for the next week at HealthcareNOW Radio. After that, you can listen on demand (See podcast information below.) Join the conversation on Twitter at #TheIncrementalist.
Listen along on HealthcareNowRadio or on SoundCloud
Raw Transcript
Nick van Terheyden
And today, I’m delighted to be joined by draw Zurich. He is the chief executive officer of magentic. I draw. Thanks for joining me today.
Dror Zur
Thank you so much, Nick. Thank you for having me today.
Nick van Terheyden
So if you would tell us a little bit about your background, how you arrived at this point, we’ll get into some of the details of magentic. But before we do, what’s your story to this point in your career?
Dror Zur
Yeah, thank you so much. So I have something I have a PhD in applied mathematics and computer science and before the Master of Science and Bachelor in electrical engineering, with a focus on what was called in computer vision, stuff like that, which is now called artificial intelligence. And then I have 30 years of more than Surtees experience in the industry, in technological companies in about 25 years of them, the medical, medical device companies. So I worked with companies who operate like Johnson and Johnson, working startup that was purchased with by St. Jude Medical. So I have experience in this area. And after some years, in this 20 years, I decided that I would like to fall into stablish, something by myself, it took me some time, at my own company, which was a consulting company, I looked for something that will have, you know, an unmet need a technological challenge, and also a business case and 2020. In 2014, I found that reducing the miss rate in the colonoscopy test, actually answers to the three main criteria, which I have.
Nick van Terheyden
Fantastic. So lots of experience in sort of building, vision, intelligence, you know, background in that, obviously, in healthcare and doing some of those kind of activities, you ended up on colonoscopy. It’s everybody’s favorite term and activity to get engaged in, we continue to struggle to get people to actually undergo screening that saves lives. Colon cancer is still a major killer in this country, and certainly around the world. Yeah, we can prevent it, but we’ve got to find it and find it early and colonoscopy is one of the ways of doing it. Tell us a little bit about the background to how you got here with magentic. And you know, what the story was in, in sort of creating that was that existing technology, did you create it what what was the, the process?
Dror Zur
Okay, so, as I said, in 2020, in 2014, it was an unmet need, like it was known that colonoscopy test is the gold standard for screening and preventing colorectal cancer. But yet, it was also known according to studies that were conducted, till then, that about 25% of the polyps are missed, in each Miss can be critical, because polyps can be dangerous, or dangerous, and if it is dangerous, and can develop to cancer, which as you have said, is can be deadly. So. So it was an unmet need, and there was no technological solution for that. But what happened this at that time, was that the area of AI, the era of AI began and began to be practical, no, for using the industry, especially a subcategory of it which called deep learning, to be practical, in companies begin to use it in the in the industry. And this is what I believe in that we can take this powerful, deep learning technology and apply it in order to try to use to try to solve the to reduce the mystery of politics by training these models of deep learning to detect politics and to be an additional eyes which helps to the doctor to detect more.
Nick van Terheyden
So So, I’m right in saying you said 2014, right. That was when we started I’m going to go. And I’m going to say, I’m casting my mind back. But I do remember the first instance of when what we talked about was a virtualization of the colonoscopy. And of course, there was huge excitement in the population, because now suddenly, no more bowel prep. I mean, that’s the reality for most folks that the actual procedures, you know, relatively mundane, really, very safe. But the bowel prep is quite unpleasant. I don’t think it’s gotten any better, at least not in my personal experience. And there was the emergence of these faster scanners. And my recollection of this historically was, the first thing we saw was a virtual colonoscopy that was done using a CT, similar to the sort of CT angiograms that we see. And that was a function of speed of acquisition and movement and so forth. Was that the sort of genesis of this? Or was there something that I’m missing?
Dror Zur
Yes. So I think, you know, it was really something new and something something very promising, and it’s still externalities. And it has, and it still will have, you know, its market share. I mean, no, no doubt about it. And it has its own advantages as your question, I mean, no need for a bother preparation or is. So this is, this is something different thing advantage, yet, I think, the most accurate, I think there’s no doubt about it, that the most accurate test for once again, for cervical cancer screening and prevention is the colonoscopy test, because you have to come around in the colon close, you know, to the tissue, and you get the best the highest resolution, and eyes quality image, which the doctor can contest. And more than that, if needed, if something is detected, then immediately it can be treated, right, in contrast to the these fields or tests that if something is suspicious, then you can meet up with them, we are more talk about suspicion, then the passion has to be referred to the test.
Nick van Terheyden
Yeah, I mean, when when, when we talk about screening, obviously, we’re talking sensitivity specificity. And, you know, the challenge between those two and the impossibility will of getting 100% of either I mean, you can get it, but then you essentially upset the other side of that particular coin. So, you know, finding the right test. And, you know, historically the the major test was obviously the fecal occult blood. So looking for blood in the stool, very insensitive, had all sorts of, you know, potential error rates, I know that they sort of work to find better markers or better indicators that that, but really, it was a very sort of generic, unsatisfying, I think the colonoscopy has proven to be, you know, it still is the gold standard. We’ve seen some additional testing. You know, I think the biomarkers that are now being used cologuard, I think is the brand name for screening, but still has the problems that you described, which is we don’t get to actual treatment, and all we get is a suspicion. And if you get suspicion, then you’re down into a path. Do you think there’s some utility to doing this in a staged approach? So that we start with, you know, depending on where we are, and the risk levels, you know, moving down the pathway to more and more intervention, because obviously, in your particular instance, and we should get to that in a second, it’s a less invasive activity. But you know, you can’t treat it if you find it. Is the some scope to sort of defining this sequence of appropriate nurse. And you know, how we do this? I mean, have you thought through this at all?
Dror Zur
Well, I think, first of all, yeah, I completely agree with you. I think. I think what you have mentioned in the beginning is exactly the point. I mean, first of all, it’s a matter of sensitivity and specificity about the performance of this. And in this, this new stool test with biomarkers are really sounds promising potential and I think if they will be proven ready in order to have a high performance, sensitivity and specificity then think it can be thought to have a gradual death but as you’ve mentioned, it will be something into do it in steps that you have to close up tests still It definitely will stay, you know, the the main test or the test that you know that everything is directed to you know, because you tend, you have the end test, which is in which you see, actually what is going on the thing is can be treated, there are countries, you know, out of the US, by the way, which has bowel work, they have bowel screening programs, stuff like, which really said, the more today that they’re using, they’re using stool tests, like, you know, even blood test, and only if it is positive, it is recommended to go to colonoscopy. So in some places, it is already in this way. But yet, for example, in the UK more, I think it is like that, but yet there are a lot of colonoscopy test
Nick van Terheyden
entry. Interesting. So, um, tell us a little bit about magentic I am what it is that you do, you know, what’s the process? What’s what’s been the sort of pathway to get to this point? You’ve obviously some recent news with the FDA, which is obviously exciting, but tell us a little bit about the background, if you would,
Dror Zur
yeah. So, so, actually, what maybe I will, I will tell you a few things about the technology itself. So the magentic product is the magenta color, it is an add on it’s a junk to the to the colonoscopy device, it is any gets the video output of the console like this video, which goes directly to the monitor, at which the physician look and then instead of going directly to the monitor, it goes through the computer a unit of the magenta color, this breaks it into frames, and we switch AI and try to detect symbolic and if a product is detected, it sign it with the bounding box as an overlay. And then the video with the overlay is sent to the monitor. So in the monitor that so there is a monitor also with the video and these bounding box which draws the attention of the physician directly to the to the to the public in a way that she or he cannot miss it in this is the way to actually improve a you know, the average adenomas, adenoma, it’s dangerous type of the average number of other normals which are detected per test, and also the percentage of patients with at least one of the normal which are detected. And we did a huge, big clinical trial International, which was running the US, Europe and Israel, the company headquarters in Israel, and it really clinically proved the efficiency of using the using the magenta color in order to increase the diversion and number of other numbers per test which are detected that to reduce the miss rate and to reduce. And this was also one of the bases for the FDA.
Nick van Terheyden
So for those of you just joining I’m Dr. Nick the incrementalist today I’m talking to Dr. Xu, he is the CEO of magentic. I, we were just getting into the details of the technology. So essentially, this is an add on to the colonoscopy process, the gold standard of diagnosis. It takes the images that we get from the colonoscopy, as it’s being performed, and essentially puts in additional information for the practicing physician who’s carrying out the procedure. Who would do this unaided, typically, you know, he does it visually he looks through this is all based on experience, you know, the identification of these lesions, obviously, the identification of polyps and adenoma is in particular, critical to the success. And you know, even with a colonoscopy, there is a small risk that somebody might miss something. And what we’re essentially doing here is using technology to surveil that imagery. And then I assume some level of the deep learning that we talked about earlier, is being applied to say that you’ve trained this through, you know, large amounts of data to then say, well, I’m seeing something even if you don’t, and drawing the attention of the surge as the surgeon to it. Tell us a little bit about the process and the data that you’ve sort of accrued and you know what you’re getting into and The results because it sounds like you’ve got some progress here, right?
Dror Zur
Of course. So as you have said, we’re talking about millions of frames, we deal with, you know, for taking from recorded colonoscopy tests which we use, which we use in order to train the system. All the data, by the way, was collected under a clinical studies for data collection. With, of course, IRB approval, passion concepts and stuff like that. But we’re doing it since almost the beginning of the companies collecting this data, and we continue to collect. And this is essential for this type of this type of technology. Because what is we have a team called the the clinical team. And they are, well, although they have a medical background, and all they’re doing is taking these videos, checking, you know, using the help of fasten your gastroenterologist and reports from the procedure, and they are really verifying where the polyps appears in these videos. And they tag it, you know, and this is what we call the grant rules for this type of technology. And then as you have said, the model, the API model is trained on millions of frames, for which frame there is an answer, which is the ground truth if there is a polyp and where the polyp is, and accordingly, the relay system learns where it is. And then when it can counter in real time, in the field, within encounter a new frame of new video, if there is a polyp it knows to say that he there is a polyp. And based on what you’ve learned, this is like we train the brain of the model to be ready for new things that will appear.
Nick van Terheyden
So, you know, millions of images, lots of training going through this, what are the results? Like I know you mentioned them, but just tell us what you’ve managed to achieve. And, again, I think there’s also this FDA clearance, which means that this is now potentially available here in the US.
Dror Zur
Yeah, so maybe the clinical validation, I think maybe it is the most important. So we have seen a increase in the APC, which is the normal protocols, like the average number of adenoma dangerous curriculums, we have seen when using the system, an increase of about relative increase about 37 percents in that annoying number. And then the ADR which is the normal detection rate, which is the percentage of patients in which at least one at least one of the normal was detected. Essentially the mother of all the questions that were tested, and increase when using the system of 26 percents, and then the miss rate, because our clinical study, part of the patients went through double pros. First one and immediately a second one. And what is found what was found in the second one is actually a miss of the first one. So we could measure miss something, which is called the miss rate. And miss rate was cut almost by half, which is think and all these results were significant, statistically significant in this what actually proved the efficacy of using the system. And it was done over like nine 952 patients were recruited in 10 centers, in 31, colonoscopies to participate.
Nick van Terheyden
Yeah, I mean, let’s dive into that a little bit. Because obviously, you know, people are listening to this across a spectrum. And, you know, certainly for those folks that are listening to it and saying, Well, you know, what about me? What about my colonoscopy, you know, that miss rate, even today is relatively low. But obviously, for the one patient that happens, that happens to that’s a huge deal. Because if you miss it, you don’t get screened again, you know, potentially have a diagnosis that’s missed and then develop a later stage or you’re discovered at a later stage. That’s a huge negative. And what you’re saying, What I heard was the miss rate decrease, and I think I heard you say it was more than 50 over half, almost, almost. Well, you know, anything based on you know, that’s a big deal. So reduction, you know, that alone, and then obviously, the opportunity to identify things, you know, through that process of the sea. You know, that’s exciting. So you must be pleased. I mean, this is good news. I think, you know, there’s real potential, what’s the sort of current use use of this is this in play what where are we seeing this?
Dror Zur
So we are now getting prepared to deploy our system and medical centers in The US will begin relatively slow, like later on this year. And we hope to reach more and more centers in the US, really in order to, you know, to help saving more lives of people in the US, who can really add according to these results, you know, we can really have that. And at the end of the day, I think I can recommend, you know, that we will try to be present as many as possible centers. And I really recommend, you know, to everyone to think about asking to have a colonoscopy with a, even if you have great doctor, you know, it’s just can help, maybe it will no, it’s just an additional eyes.
Nick van Terheyden
Yeah, I mean, one of the challenges, obviously, in the US is, who pays for it? And, you know, we continue to sort of struggle with this, you know, one of the key things will be the validation of this from a billing perspective, to sort of include it. You know, there’s also, I would imagine, I mean, these things are all recorded. So surely, there’s some potential to actually do this retroactively and say, Hey, we could take all of these recordings, even though it’s been carried out, if you, you know, we’ve seen this with Radiology Departments where they will run, you know, some background detection to say, Hey, what did we miss in that process? Thoughts about that?
Dror Zur
Yeah, well, I think it’s a great, it’s a great point that you are touching. And we have thought about that. And actually, from our point of view, like as a company, technologically, you know, that we are ready for that, because we are running all these training that were mentioned, from the cloud, you know, we are running many things for the cloud. And in order to and we are actually ready, almost ready, like to be something almost immediate. And, you know, to give the service also in offline from the cloud, I think that either, let’s say, the medical center, you know, we’re looking a little bit deeper, you know, some vision a little bit more to the future, know that the passion can can get a recording and upload it to the cloud and getting additional additional results from the AI that will, that will, that will test this video and give the give the report. But I think really, this is a matter of relief to see how it will. It will work with the regulations and with the robustness methods and stuff like that. But in my opinion, this is something that will happen. It’s just a matter of time, you know, but not very far future. Yeah.
Nick van Terheyden
It’s interesting. I mean, I’ve certainly folks that listen to me know that I’ve spent, you know, my career, my life pretty much collecting my medical record, I insist on getting information historically was extraordinarily difficult, you know, images, I still have old films, I’ve digitized them, although, you know, questionable the value of all of that. Now, I sort of, you know, anytime I have imaging done, I’m insisting on getting my DICOM images. And now you’ve raised an issue, and I’m thinking for the colonoscopy that I’ve had, and I’ve had certainly more than one. I’m thinking back and going, Gosh, I didn’t even think to ask for the video, I’m sure that that’s going to create no end of you know, resistance initially, why do you want to whatever, but now I actually want it because at least if I have it, I can retrospectively go back and look to say or even, you know, potentially upload it and say, Here, here’s an opportunity. So I think it raises a number of opportunities for folks to sort of think about and obviously for you. Tell us what you’re excited about for the future.
Dror Zur
Yeah, okay. So for the future, we have we have several things that I would say in our pipeline, and we will include first the next version, which is already being being tested. In the field. For example, in Israel, Europe is a version which include also polyp characterization and size estimation, characterization, whether it is dangerous and decisive measure which is also important because it’s about the quality of the more mature it is the more it is, the more closer it to potential, unfortunately, cancer and they and then we will put additional things like a cleansing level automatic calculation of the clinical level of support because the patient should come prepared. The more prepared the better. And there is a standard and want to be more standardized in additional features, we collect data also for upper GI issues, a lot to treat and also for specific diseases, not only a polyp to help with AI to try to diagnose additionally, additional things. Yeah, so we see I would say that’s really in future we can give some kind of Gi is with AI. So AI suite for the gastroenterologist.
Nick van Terheyden
Fantastic. So the future is bright. I think another example of artificial intelligence, deep learning, you know, just generically, those terms being applied in healthcare in a way that’s meaningful, delivering positive change. Unfortunately, as we do each and every week, we’ve run out of time, so just remains for me to thank you for joining me on the show. Dr. Thanks for joining me.
Dror Zur
Thank you so much, Nick, Nick, for letting me have this opportunity. Thank you.
