Dean Bitan, Co-founder and CEO, Imagene AI 
Interviewer: Bryan Walsh, Editor, Future Perfect, VOX

Brainstorm Health 2022 - The Power Of AI: Revolutionizing And Democratizing Cancer Diagnosis 

Conferences

Revolutionizing And Democratizing Cancer Diagnosis 


Brainstorm Health 2022 - The Power Of AI: Revolutionizing And Democratizing Cancer Diagnosis 


thank you so much for being here. As we just heard precision medicine. Very exciting moment right now in the field. Lot of promise out there. Your company actually disclosed about $21.5 million dollars funding round backed by investors including Larry Ellison. But can you take us back to the origins of the company and what really motivated you co founder company that can use deep learning algorithms to speed the identification of biomarkers? Yeah. Um, so first of all I'm happy to be here and thank you for having me. My personal journey began several years ago when my mother cancer has returned. It was stage four aggressive ovary cancer and after reading the statistics, I think it was the first time in my life that I actually felt paralyzed. I talked to my best, my best friend a lot which told me then you're entrepreneur. Just just do what you do. Put your emotions aside and act like it's another project of yours. And this is exactly what I did have learned their cancer thoroughly. All of the potential treatment available clinical trials. I prepared the binder and went with it with the psychologist. My first meeting with the oncologist. I think that this is when I got that physicians and engineers are born on a different planet. While while for me as an engineer, it's all about having the precise solution to any problem. The physicians are much more experienced with uncertainties, probabilities and and unpredictable patients. My mother passed away two years later and during that time I've learned a lot about the gaps in the cancer diagnostics process, but mainly that we can and we should try to do better. And the key for that is bridging between those different ways of thinking, merging those different disciplines. And this is the story of imaging. The company that I co founded. We get our teams of oncologists, data scientists, software engineers, biologists and pathologists always the same mission to assist cancer patients and having the best possible treatment. So what can your ai do that the human eye can't do and how does that both change treatment options and potentially also improve clinical trials. So I think that um sometimes when you want to make a new science you just need a new eyes in order to look at the world is a great example for that is what the astronomers are doing with infrared technology that allowed them to discover new styles and amazing facts about our universe which was hidden from them up to that point, which we can do very much the same with Ai with cancer while looking through the L. N. We can discover new layers of information and implement them rapidly within the clinical practice and standards and and the research of course. And this is what we did at imaging. We developed deep learning algorithms that can take the biopsy image alone and in two minutes provided diagnosis regarding the presence of different cancer mutations and biomarkers and also whether a patient will respond directly to specific therapies. We mapped all of those professional guidelines such as the NCC and and others. Those are the oncologist treatment protocols and development still developing everything that is meaningful for the treatment. We were able to detect additional patterns that cannot be seen by human eyes. And and by doing so um we can provide a real time diagnosis that has no set of requirements. You don't need an expensive lab in order to provide it even not any other where by the way, and and basically can democratize the clinical trials and the cancer diagnostics for all kinds of questions no matter where they live. Uh The clinical trials part is interesting. I mean, you you know, there's a there's a lot of drop off on that. It's hard to actually complete those using this technology. I mean, how can actually improve the trial process which then of course can help us help help help get us to treatments and other effective means of actually dealing with cancer. Yeah, so I think that for many decades we treated cancer mainly based on where it secures the body organ where we found it to be long breast thyroid et cetera, which is not that fun for one size fits all. If you think about it applying the same treatment for those groups of patients. Um but thanks to the ability to read the human genome, we were able to move forward to precision medicine that basically searches for those cancer mutations that actually caused the cancer. Uh huh and provide those targeted therapies. The thing is that unfortunately the diagnosis for those uh the tensions are still costly time consuming. It can take several weeks. And and and there are multiple other problems as well. And this is why pharmaceutical companies are struggling with enrolling patients for drugs that are that targeted therapies in that case. Uh And and we see that there are very centralized in the big academic centers usually. And if we want to make a decentralized clinical trials that basically aligned the same opportunities to all cancer patients, we have to think and different approach of how we can screen those patients that will be eligible for the different quality. Just because I mean precision medicine at this point it's very promising but it really only reached I think like 15% of cancer patients right now. 15 and this is something went about the democratization power of Ai you see that connection where that can grow from 15 to 50 or more. You're definitely right about that. Um You know, precision medicine definitely significantly improved the response rates from many patients but it is still relevant to less than 15% and we have to ask ourselves what can we do about that. And and I think that um you know I really like Danielle's analogy of DNA sequencing to food, it's almost trying to take an ingredient of the dish and understanding the taste out of those ingredients which is more or less what we're trying to do with bob Steele while breaking it into a list of mutation and trying to understand the cancer out of that list of mutation. It can take us up to a point but we have a limited perspective. We are missing the big picture the context in the relationship between all of those ingredients. And I think that this is what Ai does best. It's encapsulates everything that is meaningful and relevant for the unique condition of the patient. It can provide procedural medicine that is tailored for each and every condition of those patients. So what we did at imaging is to build an infrastructure that can basically take only the bad teenage and and classify on top of it the micro environment. Those are the pathologists class features that usually can classify it under the microscope, but also the DNA RNA level ePA genetics and proteins, all of them. Again in a special way that we can understand where they are exactly in those relationships. So it can send us to a place where we can not only stay that patients will respond to the therapy but also to explain the why. Together with that infrastructure. I think that they get a good example for that is um you know just aired Albert dr Albert bourla talking yesterday about immunotherapy which isn't you know potentially can cure cancer. And he said totally agree about that. Um But the thing is we don't have good enough biomarker yet to understand whether patients should receive those drugs or not and and and this is exactly the type of technology that can answer that and can also explain why we think this is a, you know, it's the newest technology. So what role does regulation does the FDA play in actually getting this rolled out in a larger way? Yeah, so there is always kind of attention between are wanting to move forward and and um they need to keep the safety for the patient and I think that the FDA is of course the gatekeeper, they need to find that balance uh and we need to provide them the evidence that we have an accurate places declared methods ah repeatable and robust products and models that can answer those and and if you can also add on top of it, you know, the approaches have just talked about it can do the transparency and they explain the why behind it. So I do expect that in a few hours. You will see that over the clinical paddling, you're a Tell Aviv based company. I'm curious because Israel is a country that really punches above its weight in a i in a lot of fields including ai medicines, cures, what you know what your perspective is on exactly why that is the case. Um Well it's small ecosystem and I think that um I think it's kind of probably assisting us one with the charger, those entrepreneurs and then you're also close to all of those medical centers that assist us with having the relevant data in order to train our models. Um Probably that's the reason that's a small ecosystem that's doing great. You mentioned at the start with the story about your mother learning engineers in the world one way position to see the world a different way. Obviously, you know, in this experience you've had with the company, have you gone to see kind of a synthesis between the two? Or do you perhaps see physicians beginning to shift a little bit? They can actually see things like engineers and engineers a little bit seeing things the way physicians do as well. I think that cancer is is super complex disease. And we are starting to understand that you know, it's affected by so many parameters, our genomic gender microbiome, it's part of our revolution and and we want and we want to embrace that complexity and trying to tackle it. We need different ways of thinking. And this is very interesting to see on the same table those uh you know, those disciplines and and it's creating great things. That's that's what we saw so far. Is there also a cost issue here too. So when I think of how ai has influenced other fields, I mean the power to scale to do what you know would take many many many humans obviously to do Ai has that power obviously when we think of medicine with cancer care cost is definitely an issue there. I mean just not just now but in the future as a I gets more involved. Is that something that can really bend the cost curve as well where there's cancer with each other treatments as well? Yeah. Um By the way I'm glad that you asked about that because I think that there are you know two levels for that. So we're talking about the diagnosis basically we're talking about the technology that it's a piece of software where we're uploading an image and received the the answer immediately. So of course we need to invest resources in order to develop it. But in terms of running those tests we're talking about a few cents of computing power so that's where we can go with that. Um But but of course there is also the treatment and and we need to think about that. So I think that you know embracing that new standard for diagnosis and also assist the farmer companies in producing sorry in reducing the R. And D. Costs for their drugs, accelerating the clinical trial eventually. Also to improve the success rate of those drugs and push more drugs to the market. Um So so yes I think that we also need to think of our or we can make it more easier for those drugs from three companies to provide more affordable prices for the drugs but that's the mission of the oil industry. It's a global mission and this is the perfect form to discuss about those issues and those topics because we have to find a creative models together with the regulation authorities and those companies and the rest of the industry in order to provide those treatments also for developing countries and also for areas where and we talked a little bit I think backstage and when people say I still sometimes images they get concerned. Is there is there bias for this work on all peoples equally? How have you sort of dealt with that? Have you have you addressed that issue? Yeah. Um I think that um it is very important to see that the algorithms do not discriminate between different populations but biased issue. I think it was introduced to the genomic field much before we even started to think about if we will work on all of those studies have been conducted up to that point in genomics, you will see That's around 80% of the samples were taken out of European Caucasians which of course do not represent the true diversity of the population. And this is also the reason why we keep testing that with our models. So so far we did multiple trials testing thousands of patients from different populations and try to see whether there are differences and I'm happy to say that from what we saw until that point there is no significant differences between population when it comes to diagnostics based on an image. But we should not stop there. We also need to understand whether there are differences in terms of the outcomes, the response of those patients from different populations. And this is the study that we're currently designing together with the University of Miami and additional groups and hopefully we'll have the results. Let me just lastly, you know, let's think 10 years into the future. Um Well, ai not just again within cancer, diagnosed within that, but but really in all kinds of medical care will be encountering all the time as as patients. It's something that will just be kind of wrapped into our care in a way that's sort of in the background, the way that AI has in a lot of other sectors as well. Yeah, I do expect that, you know, in 55 to 10 years we will see um, we'll get to a point where the patient can receive at the point of care, the optimal diagnosis and the treatment immediately after the vaccine. And and you know behind all of those numbers and graphs and statistics, we have to remember that there are real patients and for them it's much more than just like life or death. It's also about the quality of life. Just imagine the pressure that the patients are dealing with while waiting a month for the results and their families by the way. Um, so I think that it's our duty to do whatever we can to provide them the peace of mind, knowing that they receive the right care at the right time, at the right place while fighting the biggest fight of their lives. I think it's a great place to end Dean. Thank you very much. So what?