Latest / AI Meets MedTech / How AI is Transforming Cancer Prevention from Late-Stage to Early Detection
Transcript
- Anna Nesterova: Welcome to AI Mitsmet Tech. Today my guest is Chen Bendori. She's the co-founder and CEO of Afra Health, the AI Healthcare platform for at risk cancer patients. Gen, welcome to AM at SMETech. Chen Ben Dori: Thank you very much. Thank you for having me. Anna Nesterova: I want to ask you a few questions about your platform because it sounds very intriguing. Chen Ben Dori: Mm-hmm. Anna Nesterova: You're saying that OFRA is the AI driven healthcare platform for at risk cancer patients. So who's Chen Ben Dori: Mm-hmm. Anna Nesterova: your actual customer here? Who are you building this for? Chen Ben Dori: Yeah, so just a word about what we do. we our platform, our technology, helps proactively identify people that are at risk for cancer and then manage them long their longitudinal follow-up in order to reach that early diagnosis. so at the end of the day, every single stakeholder has a massive interest or in or urgent interest in our mission, right? To save lives by catching that cancer early. so we we are addressing our the patients, but we're doing it in a B2B model. So our actually our actual customers are health systems or clinics or risk-bearing entities. The unfortunate reality for these entities is that care teams are actually drowning today. They have so many things to do. to manage individuals is a burdensome, of course, task. And people are and and they're focused more on the sick care because, you know, these are the people that need urgent attention at this point. But there are a lot of individuals that are are healthy but are predisposed to get cancer, these at risk for cancer individuals that need to take care of. And when they fall out of the off the radar, you know, the system, health systems or the clinics, they lose the revenue, but Worse than that, the health outcomes is of course suboptimal to say the least. And the patients themselves fall through the cracks, they're anxious, they're confused, and sometimes just give up, you know, on their health. so at the end of the day, we address our customers are the the health systems and the clinics, but also the benef beneficiaries are the, of course, patients themselves. Anna Nesterova: So if I understand it correctly, you are trying to predict if a patient has a high risk to develop cancer. Chen Ben Dori: Exactly. Anna Nesterova: That mission is very clear to me. However, I also working in this industry for quite some time, Chen Ben Dori: Mm-hmm. Anna Nesterova: it almost changes the narrative about what does it mean to be sick or what does it mean to be healthy? Because all of us may have some chances to develop some sort disease. How do you go around that? And also the second part of that question would be how would healthcare systems treat patients like that really? Or they will just ignore them? Because to me, right now we live in the society where we put money where the sick patients are. I'm just trying to understand the shift of the narrative. Chen Ben Dori: Yeah. so let me let me try the first one first. I mean, to be honest, I I'm not sure definitions are actually relevant in our area in almost any aspect, right? I mean, we are beyond definitions. But what really matters is what a patient, what a human being needs at that point of time. So if historically medicine would be the absence of symptoms, right? Nothing hurts, you're fine. We're not there Anna Nesterova: Mm-hmm. Yep. Chen Ben Dori: anymore. So you have high cholesterol. you don't have any symptoms, right? But you probably would need need to be monitored or treated or whatever it is in order not to get to that really severe consequences. So if you have high cholesterol, you're healthy, are you sick? Does it matter? You just need to be something has to be done, right? So it's like that in cancer risk. Something has to be done. And and now more and more of, you know, of the governments are understanding that or the health systems are understanding that because they understand that if you don't identify that risk, you're probably going to get these people that are sick at a very higher later stage and an earlier age, which is also, you know, something that is new because we're used to being sicker as we grow older. But now there are chances to be sick when you are actually young at the age of 30, at the age of 40. so I think that the shift is becoming about that, is becoming about being active. In a continuous state of risk mitigation according to the risk, not according to your age, not according to if you're healthy or not healthy. It's about if if you're at risk, you need to be taken care of in order to stay of prov preserve your health. If you're not at risk, maybe the system should not focus on you at this point. So this is something that I think AI is changing, and I think it's for the best. I think that now. regulations, governments are actually moving toward that risk profile. I think that medicine or at least research is is creating these opportunities for genetic testing. And and now for for an example to understand a risk-based type of confirmation, right? Validation. And once it's becoming cheaper, you can actually use it. You see now the US Preventive Stack Task Force you know, actually funding or requesting to fund genetic testing in some circumstances. In Israel, BRCA testing for most of the population is actually free or reimbursed. So you see now these they're moving to that cost effective type of analysis because they understand exactly what I said just earlier. Anna Nesterova: Mm-hmm. But what happens when you identify the risk? So let's say someone got tested and they have a high risk of a particular cancer. What's next? So what does your platform suggest? Chen Ben Dori: Yes, so so you know the identification is is day zero, right? But you need a plan. that's that's right, Anna Nesterova: Mm-hmm. Chen Ben Dori: that's right in life, and that's right in in cancer risk. So the platform does not only say, okay, this person is suspicious as being at risk, but it also becomes an actual plan. So that the platform actually combines, integrates this risk profile or the health data from that patient that is at risk. With the clinical literature and the guidelines and provides an adaptive personalized care plan. But that's also not enough, right? You need a plan that actually somebody follows up on it. So the system also continues with that vertical and automatically creates that management or navigation for the patient and the health system. So it's like a two two interface, right? A dual interface that creates. an opportunity for the patient to know exactly what they need to do, what where they need to do it, how they need to do it, their education, their support, everything that they need within their the palm of their hand, but also reflects to the health system so they can monitor and see what's going on throughout that care journey. So everybody wins here. Anna Nesterova: That is a very interesting concept to me because I do know how sometimes patients also think if something doesn't hurt, why would I go and do something about it? Even though maybe it sounds scary, yes, I have a high risk of developing something. Well why would I go and maybe spend money and time on something that doesn't really bother me now? Do you think with the change of what with what you do and with AI, that will also shift? And how can we really shift that? Chen Ben Dori: it's it's a great question because it really depends on the space. I think and also depends on the risk. we see that individuals, you know, we talk to a lot of at-risk individuals. you see a lot of difference between individuals that had cancer death in the family. A lot of the times it's connected to that. If you have some Anna Nesterova: Mm-hmm. Chen Ben Dori: type of cancer in the family, you may be at risk for cancer and and you see people that are anxious they want to do something about it they don't want to reach that stage of course they want to do something about it when they're healthy this this is one type of population and they're actually they really want to do something about it but the health system is is you know sometimes has challenges or difficulties to help them and the other type of population they don't know it but if they knew it and they understand what it means because sometimes you don't understand what it means it's not about getting I don't know an heart attack at the age of 70. This is far away when you're 20 you don't think about that. We're talking about breast cancer at the age of 28, at the age of 35. This is now. This could be now. This we could we're talking about 80% risk to get breast cancer, right? in some type of genetic mutations. So I think what our space is a bit different and it looks a bit different and the engagement is a bit different when we see it in our pilots. We see we have a sixty five percent engagement, people that are wanting to join our pilot here in the breast centers in Israel. and we see it with the health systems that we're talking to. but I'm I'm sure also things would evolve in that space as well. People are more and more eager to take control of their health. Anna Nesterova: Yeah. Yeah. But how do autonomy and liability mix when it comes to care management? What about the liability, right? Like if your platform tells something, how do you go around that? Chen Ben Dori: Yeah, I mean for us the boundaries are very, very clear. we don't take any control of professional or clinical discretion. We don't diagnose. we don't provide recommendations for the patient that the doctor did not approve beforehand. s so it's the human in the loop, right? At the end of the day. We're like a chief of staff for the health system, like an elite chief of staff for the health system and a companion for the patient. so the act the care teams can actually focus on the medicine and not focus on, you know, the nudging and the isomic education and and all these all these things that can actually be very much easily facilitated by technology. That's where we come in. And for the care teams, we do provide that reflection of the implementation of the clinical guidelines and literature together with the risk profile to create that. personalized care plan. But this is something that has to be approved by them. we don't bypass, we don't intend to replace any any clinician. They're there and they're it's it's important that they're there. We help them. We help them be focused on what they need to do. Anna Nesterova: Yeah. So do you combine the genetic testing with other testings, obviously all different diagnostic tools the doctors are using to create that plan? So this is what the AI behind your technology does. Chen Ben Dori: It's it the the AI of the identification actually scans the health records or the electronic health records. And we have identified from the literature and with our of course multidisciplinary team around we're starting with breast cancer. I should have said that. So we're starting with address for breast cancer. We we identified, I don't know, around twenty risk factors from the literature that could be out there within the records that are like signals within The health Anna Nesterova: Mm-hmm. Chen Ben Dori: records, the data, and we identify them and then we calculate the chances of this person being at risk, right? so we have an amazing type of machine learning algorithm that shows you, okay, this person has connection to this higher risk because of ABC, right? And then the Doctor can see that and say, wow, okay, yeah, right. I didn't notice it because there was another doctor and it was in another clinic within these network or whatever it is. And now you can and and if there was a genetic testing, of course it's even better validation, but now you can know what you need to do in order to continue the validation. Sometimes it will be a validated calculator within what you already have. It could be a genetic testing. whatever the doctor feels or the care team feels that is necessary to actually validate it. Okay. and then Anna Nesterova: Mm-hmm. Chen Ben Dori: you continue with the, you know, process of following up only after you validate, Anna Nesterova: Yeah. Chen Ben Dori: of course, the risk. Anna Nesterova: Yeah. Actually the topic of breast cancer is very close to me because I have a strong history of breast cancer in my family. And the question I have for you is probably it comes from my experience when I went to hospital here in the United States in California and I asked for genetic testing for breast cancer. And I was denied by the healthcare provider and by the insurance, even though I have a strong family history, including my mom. And the reason being, they said that my mom needs to be tested, not me. But my mom has no opportunity right now to do so, which I found was Chen Ben Dori: Mm-hmm. Anna Nesterova: very interesting. And the way how I interpreted it was all cost related. They didn't want to spend more money on me being tested, because I could probably have some results from my mom. So what Do you think your system could change in this scenario? Can your system flag to a doctor and and can it suggest to a doctor someone like me would need that type of testing? I wonder how you can go around this barrier of the economics of of the testing. Chen Ben Dori: Yes. Unfortunately we cannot change you know the all the world yet. But Anna Nesterova: Mm-hmm. Chen Ben Dori: I think that providing the when you integrate a a lot of data within and you have these explanations why this is needed, it's much easier to persuade, of course, the system. Sometimes it will pass, sometimes it wouldn't. I think that We are moving towards an era that, like I said before, that governments or payers understand that it has to be done in order for them to actually lose less money. early diagnosis Anna Nesterova: Mm-hmm. Chen Ben Dori: creates a lot of cost opportunities and it's becoming cheaper. So more people can actually pay. I don't know, it becomes like $100 or whatever it is that you you can pay and you get that genetic testing. Sometimes it's not required. If if the calculators, if there is hereditary cancer, you can use calculators and then you can see without the genetic testing. Of course it's better because then you can have additional follow-up recommendations, but you can actually do it also without the genetic testing. So again it's it's a cost is always a challenge in health systems, Anna Nesterova: Mm-hmm. Chen Ben Dori: but we're moving toward a space that it Everybody understand that you need to know about the risk earlier. Anna Nesterova: Yeah. Based on what you're building, how far away do you think we are from AI fully making treatment plan decisions for patients? Chen Ben Dori: Well, I don't think if if we're talking about like AI signing off like a chemo regimen or something of the sort, Anna Nesterova: Yeah. Chen Ben Dori: we're not there, in in my humble opinion. Not from a tech aspect, but we're we have a long way to go. But if we're talking about creating a treatment plan, we're definitely there. Anna Nesterova: Mm-hmm. Chen Ben Dori: I think the challenges today are mostly on implementation type of of space, right? You see so many brilliant things going on, right, with the tech and with the research facilities and in academia, but they're not infiltrating to the health systems. Anna Nesterova: Mm-hmm. Chen Ben Dori: especially not the health systems that don't have enough money or are not wealthy enough to actually, you know, and create that integration within them. so This is the main challenge. You see it in Israel as well, and you see it a around the world, creating that integration into into a very somewhat conservative medical system, which makes sense, right? It's this lives we're talking about, is the greater challenge. so tech-wise, we're there. Other our other other aspects were not really there. Anna Nesterova: But AI that you have and you are using for your platform, do you think it has a capability of not just assessing the risk, but also deciding if this patient needs s maybe some drastic treatment to start, let's say preventative early on, or we're not quite there yet. Chen Ben Dori: I think it it can, you know, raise the flags for the care teams. It doesn't do it Anna Nesterova: Mm-hmm. Okay. Chen Ben Dori: autonomously because again, we don't want we're not diagnosing anything. not at this point. But it does it Anna Nesterova: Mm-hmm. Okay. Chen Ben Dori: can raise the flags and create that Anna Nesterova: Okay. Chen Ben Dori: that I would say information coming from the patient to the doctor that sometimes is missing in order to raise these flags. Anna Nesterova: Yeah. So far with your experience in healthcare systems you're working with, what was the response to this type of assisting AI healthcare platform? Chen Ben Dori: I think a lot of the systems that I talk to seemed very eager to have help in the at least in the navigation and management aspect of things. Some of the system are doing their manual or human type of identification. Whoever comes in, it's reactive most of the time. So whoever comes in that is at risk, they're handling them. But then you see a lot of challenges with the downstream longitudinal follow-up because again, not enough teams and we can't scale that thing. so we see a lot of positive feedback, but it's it's it's slower because this is the this is the type of industry we're working with, right? We're working with an in the health industry, takes time. there are a lot of things that you need to be careful about. This is personal health information. there are a lot of things that you need to take into account and and that's fine. Anna Nesterova: Yeah. And you said you spoke with some patients as well, right? Who used this platform before. So what was the feedback from patients? Are they happy to use AI? Are they feel concerned? Because what I'm trying to uncover is what is the reality right now? When people are facing AI powered tools in healthcare, how do they really react? Chen Ben Dori: I think people are now, especially in the US, are eager to take control of their health. They want to do it. At least the w we are talking to women that are at risk for breast cancer. They want to take control. they want the tools. I think the issue is more about being sustainable over time. And that's why we build a system that does not only address the patients. We didn't do it a P2C type or D2C. to C type because we believe that when a clinician or the care team is in the loop, it creates that sustainability. I think that will create the change that we want to see. and and that actual engagement of the patients because the patient want the cli there a lot of the times they want their doctors or their health system to be, you know, to help them. we need to help both sides to the story. Anna Nesterova: Mm-hmm. Mm-hmm. Yeah, because I also am w I'm wondering, like even like you said, I'm one of those people. I'd like to be proactive when it comes to my health, specifically related to breast cancer. So I'm wondering if this platform could change the path for someone like me. I wonder. Chen Ben Dori: that's a good question. I don't I don't have an answer for that. There's there Anna Nesterova: Yeah, yeah. Chen Ben Dori: may be people that will not benefit, unfortunately, but I think with Anna Nesterova: Mm-hmm. Chen Ben Dori: time, it will allow to optimize more and more patients that can benefit from it. so Anna Nesterova: Yeah. Chen Ben Dori: there are I would say lower hanging fruits for the people that can benefit from it tomorrow and maybe somebody can benefit from the next year and and This is something that will grow as we go along. So it will optimize Anna Nesterova: Mm-hmm. Chen Ben Dori: that, you know, the throughput of of of what of the people who can benefit from it. And yeah, some people will probably not at the at at the beginning, unfortunately, right. Anna Nesterova: Yeah, yeah, yeah, of course. But that's a nice segue to my next question. I want to talk about the future with AI in healthcare. what would be your ideal way of a patient going through the system with help of tools like you you're building in terms of diagnosing cancer? What is the ideal pathway for a patient, let's say in the next five, ten years, when AI and tools like that are already adopted. Chen Ben Dori: I think, you know, when I look at it and I say, wow, if if I had you know, we build our system with that vision. What would be the ideal way, right? Anna Nesterova: Mm-hmm. Yeah, Chen Ben Dori: if a person comes to a doctor and the doctor has all the indications that they already have, right? the it shows up very clearly to them, you know, what this patient, like a s a report, what this patient had. what are the symptoms today? What are we looking at? How are what are the plan that could be relevant to this patient? And then the patient is coming, but they already know they have that all that education, they're coming prepared. They have they have their own data and they're less anxious because they know what they're going through, or they know what to expect with that stressful, you know, interaction with the doctor. and then after that interaction, everybody you know, goes to their ways and the doctor knows exactly that this patient is on track and the patient knows, hmm, I'm on track. And everybody is not confused in their ways. So I think Anna Nesterova: Mm-hmm. Chen Ben Dori: so I think that, you know, the this type of medicine would be the ideal type of medicine. People w sometimes you get that when you go to, I don't know, a private clinic and you get that, you know, all the services that you need and all that you combined you know processes and roadmaps and if any everything that you need within that clinic. So it looks like that, but for everybody. Not Anna Nesterova: Mm-hmm. Yeah. Chen Ben Dori: just for you know the wealthy or the people that can afford it. It just looks like that for everyone using tech because that's how you scale. Anna Nesterova: Yeah, it sounds like you're trying to democratize this ideal way of putting the patient in through the proper path from like finding out the risks and obviously doing all the necessary steps to mitigate the risk. And if they start presenting some symptoms, what they should do so they don't have to wait. Chen Ben Dori: Yes, I mean that's that's a a great pro byproduct of what we're doing. Anna Nesterova: Mm-hmm. Chen Ben Dori: actually we started this is this is very important because we started here in Israel where at risk for breast cancer is very, very prevalent. And and we have at risk clinics that came about, right? That they that that were established in order to solve the situation and they're great. They're really great. But then at the end of the day, you know, in a few years, it a few years back, we see that they're backloged. There's their waiting list, the two hundred percent, three hundred percent waiting list for these clinics. And we said, wait, maybe we can take this, model it into technology and create that scalability for everyone. So That's how we started it. so we when you talked about Anna Nesterova: Make sense. Chen Ben Dori: that ideal situation, it it you know, that was the ideal situation for us and and we wanted to mimic it into a scalable type of platform. Anna Nesterova: Yeah, that makes sense. I actually wanted to ask you about how did you come up with this and why Chen Ben Dori: Mm-hmm. Anna Nesterova: you thought that AI is a good idea, but now that now what you just described makes sense. Have Chen Ben Dori: Yeah. Anna Nesterova: you seen any pushback from doctors? Because I feel like with AI there's always a bit of a challenge in adoption and perception. Do you feel doctors get threatened by it, saying, Okay, now AI can decide for me? Or they see it now as an assistance? Tell me more. Chen Ben Dori: That's that that's that's also a great question because we remember, you know, you probably also remember the Dr. Gru will effect. and the doctor Anna Nesterova: Yeah. Chen Ben Dori: pushed back on that and I must say it's rightfully so. it it it it brought a bit of chaos, a bit of anxiety, misinformation. So there's there was a lot of things to push back on. But things have changed since, you know, twenty and ten years ago. And when I talk to clinicians and we build our system together with clinicians, and I talk to clinicians literally every day, actually married to one as well, they don't push back and they actually want validated structured AI. Today every doctor uses ChatGPT, Gemini, they use it. Okay. They are customed to use it. So it's not in a world that they're not they're not using AI within their everyday life. So they want to use it, but they want to trust it. And that's the point. They want explainable AI, they want structured AI, they want validated AI, they want to trust it. And they should trust it. And they they they're asking for something that makes a lot of sense, right? I don't want to trust something that is not good enough in a professional field that is very, very you know, that deals with lives. So I think that doctors actually really want this and they're also swapped, right? They have they need help and they know that. so I think it's it's a very big change since the LLMs came in into our lives that also created that. Anna Nesterova: Mm-hmm. now with what you said, like you can validate your system, you can build that trust, but what helps you do so? Is it the certain approvals you have to go through? Is it certain algorithms you have to, I guess, test against, I don't know, the gold standards? How do you usually go through that process? Chen Ben Dori: Yeah. So if you're within the regulatory process, then that's the best way to go about it in many ways. If you're outside the regulatory pr process, which we are at this point at least, then you have to show, yeah, against the gold standards, you have to show it on real world data. you have to pilot it in order to Anna Nesterova: Mm-hmm. Chen Ben Dori: prove that it actually works. and then you do like any research, right? You go you think about it as a study in academia in many ways. So you do the study, you see the percentages, you validate them or you see that they're actually correct. And then there's no hundred percent, right? When when we're using machine learning, OAI, there's no hundred percent. But if you're going to the 90, 95% of accuracy or precision with real world data, that's great. Even 85% considering the situation, right? And the gold standards today with the calculators today is is also great. so it really depends on the situation, but I think the validation process is is clear. It's just you need to go through it. Anna Nesterova: Yeah. I guess one last question about Chat GPT and LLMs the doctors are using. I wonder if someone would ask you what's the difference if I would just put the same information into I don't know, Collader Chen Ben Dori: Mm-hmm. Anna Nesterova: Chad GPT versus your system. What if you know Chat GPT would give me the same result? What would be your answer? Chen Ben Dori: the it could happen like any statistical fact, any statistical Anna Nesterova: Mm-hmm. Yeah. Chen Ben Dori: event, but you don't when you build tech for health systems or for health care, you work through a lot of processes of reinforcement learning and you teach the model how to think, you fine-tune it. so there is a process that we do at least in the in the care planning model or engine and the care management there is a process with clinicians that we work with with data that we work through so chat gpt is great for some things but it has it at this point it cannot be a a validated tool for health care especially not in certain verticals because it's in a horizontal tool at this point. And we are working vertically. So we are the specialized doctor, I would say, for this type of AI and healthcare. And in addition to that, ChatGPT putting you know, PHI p protected health information within Chat GPT is not something that that a doctor would do Anna Nesterova: Yeah, definitely. Chen Ben Dori: or a health system would do. And we of course take all of that into consideration with HIPAA and the standards and everything that is required. So there are a lot of processes of digital health and and AI and healthcare that have to be taken into consideration. It's it's not I would say not it's not easy. Anna Nesterova: Yeah. I have a fun question for you. So has anything about AI in healthcare surprised you recently? Chen Ben Dori: I saw I idol in psychology, which is is close. that I was very much surprised that people relate or can actually talk to a virtual or avatar psychologist in Anna Nesterova: Mm-hmm. Chen Ben Dori: a in a in a in a way that they would talk to a human. they feel even more relaxed or more free to say what they think or what they want to say. I was surprised. I I thought it would actually be, you know, we'll feel maybe it's a bit cold or high-techy. But nope, the exact opposite happened. So there are a lot of opportunities there that a human in the loop is is very, very important. But probably maybe some of the things that are done by humans can also be replaced by in AI in in a different ways and forms. but that definitely surprised me. Anna Nesterova: Would you go and talk to a psychologist like that? Chen Ben Dori: Well, I think you can understand from my question that no. I I was I I probably less relate to that, but just Anna Nesterova: Yeah. Chen Ben Dori: th that's me. You know, I I assume that there are not different type of people. Anna Nesterova: I'm actually surpr I mean, I'm surprised and not surprised at the same time. I think this is where we're heading, but it concerns me. Chen Ben Dori: it it it is concerning. I mean it is I I would like to believe that we are like to believe in human beings and to to stay in the loop and that's how we build our company with the thought that human is always in the loop. but there are many things that concern me about, you know, AI and and the integration of AI tools. This is one of them. Are we replacing human beings? how do we do it? If we do it, what's the line? ethics, but also health equity. who is receiving what? we see it in the US a lot. and innovation with that access, in my opinion, is not a victory. so it there are many, many questions, and I think now you see more and more. of governments or health systems addressing these questions and trying to put in the boundaries but these will change this something that is is an ongoing process of that we need to think about but I think once we if we are thinking about it we're in a great place. If we're ignoring it it's not good. great. Anna Nesterova: Yeah, no, I agree with that. And I do agree with your overall statement that it's I hope that the human part of us will stay, and we using AI for better and for assistance, but not replacement. I have one last question for you. If you would have a magic wand and you could Chen Ben Dori: Mm-hmm. Anna Nesterova: ask AI with that magic wand to do anything for you. It doesn't have to be healthcare related. It could be anything. What that would be. Chen Ben Dori: Stay healthy and happy. Anna Nesterova: Do you think AI can do that? Chen Ben Dori: I don't know, but you but it's a magic wand. Anna Nesterova: That's awesome. Thank you so much. Thank you for joining me on AI MidsMetTech. Chen Ben Dori: Thank you very much for having me. It was a pleasure talking to you.