Latest / AI Meets MedTech / The Smartest Biopsy Scanner But Only $5,000??? Bringing AI to Pathology

Transcript

The Smartest Biopsy Scanner But Only $5,000??? Bringing AI to Pathology · AI Meets MedTech

• Anna Nesterova: What if the device that could screen your cancer biopsy could cost not $200,000, but let's say less than five? I would say like a cost of what two new foldable iPhones? Let's find out. Welcome to AI Meets MedTech. Here is Jerova. My guest today is Michael Fanus. He is the PhD researcher and founder of Fanus Photonics at UCLA. He's built a portable AI-enhanced microscope, which has a very attractive price point. And I can't wait to learn how it works. Michael, welcome to AI Meets MatTech. Thank you so much, Anna. It's great to be here. Thanks so much for having me on. Yeah, so let's start with A little bit with your journey. So you are a researcher. You worked, I would assume, many years in a lab. Absolutely. So what made you start a company and build your own product and how it's been going for you so far? So yeah, I have very much an academic background. I was happy in that bubble of just intellectual pursuits, arts and crafts. You know, it's sophisticated and you think maybe one day it'll help someone down the line. But a lot of academic projects are done purely for their own sake and they may never see the light of day in terms of implementation for like the next century. And that's fine and that's worthwhile. And I think those pursuits have merit. when we came up with this particular concept, just the theory, the proof of principle, it had a commercial aspect to it. Okay. it could have been licensed, but licensing is a little tenuous, it's kind of flimsy. I didn't feel very solid or Highly motivated to pursue it at that time. This isn't during my doctoral program. When I came to UCLA and a prototype was actually fashioned, specifically around an AI premise, which is important. Most med tech or hardware in general, that is so-called smart or intelligent or AI powered, it's putting the algorithm at the tail end of already established physical components. So we're doing the exact opposite and trying to optimize the hardware. By leveraging a very powerful initial AI concept. Ooh, I like that. Tell me more. Mm-hmm. So when the prototype was finally built, and initially it wasn't terribly attractive. It was gray, it looked like something out of Star Trek from the eighties. But I remember I brought it to Shark Tank and they dismissed it outright. They just thought this is a ridiculous looking contraption. And but the engineering I thought was very attractive and could be potentially enticing to the average pathologist if I just changed the casing, added a few components, made it put a screen there, you know, histology is naturally extremely beautiful, in my opinion. It's like a painting. So when I'm taking a picture of it against like a sunset, and it's like what is more beautiful, the tissue or the sunset? And I don't know personally. There's so many details, they're so intricate, and I think it's just downright gorgeous. So I thought, you know, and and the prototype is sitting there in the lab. And I'm thinking, I can continue doing this research or I could try and make something out of this. And at the time I thought, you know, I'll register a company. It was kind of a peripheral, half hearted endeavor. And I thought, you know, maybe an undergrad can take this. Someone who's type A, very extroverted, clearly. Could do well in an entrepreneurial setting. And I thought, that's not me. I'm timid. I'm bashful. I'm very introspective. I like just sitting around deliberating for hours. I'm not necessarily gonna go out pitching and that stuff, but I realized no one was gonna do it. I'm the one who really thought highly of it. No. And I was invested in the concept. Yeah. And so I just decided on a whim to take it to a conference. And that was again just kind of a tangential activity of mine. I took it there. I really didn't know what I was doing. I brought it to the booth and I didn't know what to expect. And I I had this like lingering fear that pathologists would just dismiss it. They would think it's silly. It looks like a toy. It's foreign. It's kind of gadgety. You're using AI. They would feel uncomfortable. Total opposite. I had drastically underestimated the ferocity of the desire for this kind of device and for new devices in general in the sphere. So can you describe the device then? Yeah. So the device is, and this is a bit tricky because it doesn't fit nicely in any category. Okay. It's a microscope, it's a scanner, it's a microscope scanner. It's a hybrid device. It's a category divine defining unit. So it defies a lot of what you might expect. In pathology, it's important to understand there are two main options right now. There's the well-established centuries old manual contraption, which is the Brightfield Compound Microscope, no digitization, maybe a camera is mounted on top, and then you have the scanner. Yeah. These are two worlds. They're disparate. And and just for the sake of me understanding, and pathologists use both at the same time, scanners and microscopes. This is important again. Yes. So for people who are not acquainted with pathology, the numbers In the West, in state of the art clinics, top-notch, whatever you like, they are staggeringly modest in terms of routine biopsy digitization. Okay, the vast majority of biopsies are not being digitized at all, or at least extensively. I know 90%. I did some research. Well, 90% is an aggressive number. If you add the camera on top, and if you also add the fact that there's maybe a scanner in a different facility, like but in the same institution and The pathologist can send it later for examination. You can play around with the statistics, okay? But the truth is the sobering reality, and if I'm being candid, and it's not glamorous, because the problem is again, there's a lot of very powerful AI pure software players. The algorithms are devastating and increasingly so, and for good reason, right? The algorithms can train themselves, yeah, and with more data, you're gonna get more reliable results and higher degrees of fidelity. Yeah. And so classification, dec detection, segmentation, these are all remarkable now. But if you're not digitizing anything, even at the lowest levels, and this is the barrier, this is the problem, you cannot use it. Okay. You need to take pictures of biopsies. It sounds trivial, it's not necessarily a very romantic notion, but this is a problem. Biopsies are not being digitized routinely. We need to change that. And I think there needs to be a new wave of technology. I hope. Scanimus will lead the foray into this new kind of category defining devices that are gonna be you can put it on a pathologist's desk right there. It's very portable, very easy to use. Yep. And we're trying to make it so that there are no more excuses. And then the initial excuse was price warrant and ergonomics in that order, typically. Yeah. So as I mentioned before, the typical scanner. can range in cost anywhere from a hundred thousand up to two hundred thousand dollars and you are trying I mean you've done it already you are changing the whole market. Exactly. So how's it been for you and how did you achieve that? So so yeah that's right. A lot of people are telling me to make it more expensive because our profit margins are not necessarily very attractive to investors. They're not the healthiest like gap between what we pay for and what we're selling. In fact we're not making much right now in terms of the profit margins. And we're doing that deliberately. We think adoption we want to increase adoption and then you know find you know viability through that. But let's say you're trying to buy a scanner, just put to put it in perspective for people, at the lower end you're talking about tens of thousands of dollars. It's typically like thirty all together with software and then the other integrated units like fifty thousand dollars. This is the lower end. So it's essentially like buying a vehicle. Then if you want the extra luxury item, the heavyweight, big, you can process hundreds of slides, all of this, state of the art, it's hundreds of thousands, sometimes going up to half a million. And only a low number of healthcare systems can afford actually having those. Well, you may as well buy a Ferrari. Like why would you buy this extraordinary, like cumbersome device when you can just Go and get a Lamborghini or something. that's what we're talking about. We're talking about taking pictures of biopsies. It should be as trivial as buying an iPhone or some gadget that's like you mentioned. So when pathologists digest like the full reality of the figure that we're proposing under five thousand, the demeanor changes entirely. Because normally you need to go have some sort of meeting and it's a collective agreement on this very large transaction. Okay. It's not something to do very passively overnight. You need to make sure that you're allocating funds for the right reason. And there is a lot of external AI pressure to do so. Yeah. And especially right now, there's so many companies, so many products, so many and so much pressure on budgets too. So I think that alone, you're in the right track. I can I can tell you that way. And especially talking about some Bigger markets, not all not bigger, sorry, different markets, not just the North American market, but we're talking about global market. If you are trying to define a new category, I think you global is important. And a lot of people have been telling me, you know, go to this continent, go to this. And I'm thinking, you know, I've seen the situation locally. It's dire. There's a great unmet need for pathologists. So Again, another misconception. There aren't enough pathologists. We need more of them. We're a very old population. We're the oldest we've ever been. We're the most numerous. We also have the highest purchasing power. There are too many biopsies. Not enough scans. Yes. Not enough pathologists. So we drastically need to improve the situation. There's gonna be a plateau with respect to how much you can do with AI and pure software plates and algorithms. They're gonna reach a wall there. You need more data. And certain software companies they'll boast about we've Trained on millions of slides. Yeah. In this field, millions is barely scratching the surface. Okay. The average human like volume, 60 to 70 liters, if you were actually to take all the sections of every organ and so on, you could get billions of slides. Easy. Yes. Yeah. Okay. Petabytes of data. And so we need more, we need more data. What are we like when we're talking about a microscopy slide of biopsy? This is a thickness of five microns. maybe an area of twenty five to thirty, forty, maybe seventy millimeters. That is a derisory amount of biological tissue information. So we need more and quickly, and it's it's dire. Like right now, we we need we need to introduce something right now. So let's introduce that device right now. I think it's time. I think I've heard something. And I'm sure the audience all already is dying to know what it is. So can you tell me more? Sure. So Again, the the crux of our technology and what makes it so unique, where we're able to bypass a lot of the traditional heavy, difficult, expensive mechanics, is by using something called continuous scanning with deep learning, deep lowering. Now that's a mouthful. I understand. But it is important for people to understand that we always have the supplementary live feed and also a stop and stare method. So most scanners use what's called stop and stare. The specimen is physically s halted during the acquisition so that there it's it's perfectly sharp. Uh-huh. If you're gonna move it while you take the acquisition, you need a specialized camera, usually a line scanner, something called time delay integration. Mm-hmm. I don't want to get too technical, but it's a little boring. The point is it's expensive, it needs to be synced with the stage. We bypass all of that by taking a continuous video while the the the specimen is just blazing by. And then we could use camera, microscope, and on top of that, it's continuous scanning instead of just one screen. Simple shot. Simple, elegant idea. Execution is is more tricky. Okay. So nothing is free. Okay. You may think we're cutting corners. If you see how we train our data, you'll realize we pay very handsomely in terms of getting this done. This is supervised, meaning that each motion blurred micrograph has to have its crisp, sharp counterpart. That's easier said than done. Okay. Meaning that when we scan it, we need to scan it very rapidly, which is what we're s which which is what the end user is seeing, and that is the target for the model to process. But our target is the ground truth. Obtaining a ground truth is difficult. It's tiresome. It's boring. Okay, so so so that's behind the scenes. We don't need to show people that. But trust So so let's I guess a p role play. Let's Pretend I'm a pathologist. Yes. And I've been doing it for many, many years. I'm using a microscope on a daily basis. Now you come into my office and say, Hey, I have something I'm gonna show you. Yes. And can I what would be the first thing that I'll see that will definitely shock me, surprise me? I will be blown away by the technology. Well, right away when you're when you're looking at the pathology w the slide move by really quickly. Yeah, that is intriguing. Okay. To the average individual. But it's pathologists are no no except no exception here. Yeah. But the pathologists have a trained eye. Yes. So we're showing them the scan in real time before any reconstruction. They're already getting ideas of where they want to look at the image. This is immediate triage in their mind. So it's important to understand how the pathologist works and what they're doing when they're interacting with a microscope and a slide. Yes. And from what we've seen, no real technology has endeavored to emulate The the method to the madness, which is that interaction. And there are a lot of objectives. So why are there a lot of objectives? Why are they playing with the focal knob? What are they doing? The first thing they're doing they're searching for changes in cells. So they looking for they go very deep and kinda like see where the where the cell is different, why it's different and what it can be. Yes. So for the people who actually are not really familiar with pathology, that's the level of detail they have to, I guess, train their eye to and be focused on. So your machine can help them not only to see certain changes in cell structure, but what else it can do for them? It's trying to facilitate the whole process. Okay. So the when they're interacting with a with a manual microscope, what are they what are they doing specifically? And this is important to break down step by step in order to then digitize and make into a robotic system. Yeah. We're not trying to replace it. We're trying to facilitate their workflow. Okay. So what they're doing is a mechanic a set of mechanical and mental tasks cascading into increasing levels of difficulty and engaging higher order faculties until the they get to the final level, usually at the forty or twenty X magnification. Yeah. And they make their determination. Now between the between starting putting putting the slide on the stage and to that step, there is a great deal of laborious, menial tasks. Yes. We want to make eliminate them. Not not eliminate them, but make them more efficient, more attractive. And the other thing is like it's important to understand pathologists, they all have their own area of expertise. So they all have their own specific activity fingerprint. Yes. If you're of a certain organ, you may start with four X looking at the edges and then jump radically to 40X. Another person who's in a different organ may start with 10X and then go to 20X and look around a little bit. Yeah. And so what we do is we're monitoring that process. The other thing I was I would say to a pathologist is You can move this manually. This is a microscope. You can move this like me. I already rolled up my sleeves. I'm like, I'm ready to take a look. So let's introduce a little guy. Let's go ahead and let's stop this. Let's stop this scan and let me show you a little bit how this works. So the first thing I like to do is I like to just show real quickly. This is a slide. Okay. So here we're looking at a piece of tissue. This is moss skin skin skin. section about twenty five by twenty five millimeters. Okay. So what does the pathologist do when they see this slide? They're already looking at it. This is what's called one X magnification. They already have an idea of this. Yeah. So the scanners are not doing this. They're not trying to emulate this in any way. They may get a delineation of the boundaries of the the tissue so that they're not scanning empty frames and and blank yeah. This is another problem. There's a lot of it's an inordinate amount of data that is not being adequately processed because you're not filtering it intelligently beforehand. So that's number one. Number two, put it on the stage. It's as easy as this. Okay, there is no special insert here. Yeah. You're just putting it on a glass. She knows. Okay. Mm-hmm. Now, when you want to manipulate it from the screen, you just drag your fingers. Now, what's interesting about this, this is synced with the stage. I'm not moving a still image. I'm moving the stage. Yeah. This is already like a why a wow factor for pathologists because they're thinking they have to move it manually very carefully. And yeah, so this is yeah. So you can switch between manual movement. And digital movement. Yeah. Either way, we're tracking this. Okay. So either way, the however you you like to the style that you like, we're tracking it. Yeah. And we expect pathologists will be playful with this. We hope they are. We hope they'll enjoy the process and we'll segue quickly between digital and manual. Because they already have this manual stuff. Yes. Okay. So then they may find an area of interest that they like and then What they want to do is increase the magnification to then improve the resolution. Here is just the push of a button. Mm-hmm. And this particular unit has only three objectives. The ones that we're gonna deploy in October, they'll have four. They'll have four X, 10X, 20X, 40X. The reason for this is just to address as many pathologist needs as possible. Now you may think Why not just scan everything at 40x, you get everything that you want? It's not that simple. There's something called depth of field. The higher in magnification you go, the slimmer the amount of vertical information that's in focus. So here we're at 20x, okay. Same idea, same idea. You get you get an idea, and then you can use AI analysis directly on the frame. And then you can go back, you can switch very readily. Let me show you what a scan looks like real quick. Yes. so again. How do I convince the average pathologist? What I'm showing them now is just a live feed. There is no reconstructing reconstruction. There is no AI. But let's say that we want to start with a continuous scan, right? So this is a small continuous scan, meaning it only stops at the edges of whatever region that we're looking at. It's very small. But it'll give directly the pathologist an idea of a certain area. And then they can look in this is a stitch that's already made. So it's a little disjointed. We would need to align it. And we can do that post processing. The idea is to deliver as much information as quickly as possible to the pathologist to streamline that flow because there's just too many delays in the average workflow. So question. So what is on average the time spent at each specimen with each specimen, by the third? I I qu it it depends again what it is. Yeah. Some some pathologists, again, a lot of these determinations, and this may surprise some. Of viewers, they're made at two to four X. So a lot of pathologists they'll come to me because this is the scanner now, and they're thinking this is a different world. We want the max absolute maximum out of a scanner. They're going from the manual microscope, something that's just 150 years old in terms of the staining. It's more than that. It's like 17th century, really. and then they want the absolute most out of a scanner. They say we want 40x 25 by 75 millimeters in 30 seconds, or else I'm not interested. And I'm thinking to myself, my friend. You're going from the absolute minimum, zero digitization, and now you want the absolute maximum. We're pushing this as much as we can with our price point. Meet me in the middle. Let's be a little more amenable. And so that's the idea. I think there's great opportunity between the manual microscope and the traditional scanner. And again, some pathologists, they'll look at it at 2x, 4x, they'll make a determination, they'll throw it away. They're they're dealing with stacks of slides. Can I ask you a question? Why would a pathologist Use scanner or not use the scanner at this moment of time. So at this moment of time, we have to understand that there's a great deal of pressure with respect to AI. All of these softwares are devastating. Okay. There's a lot of talk about AI will never replace you with edge cases and extreme scenarios. Yeah. And I think there is a kernel of truth to that potentially. But there is something, there's a problem here because there is a correct answer. There's something called prognosis. This allows you to use reinforcement learning and ultimately potentially with time, challenge even the greatest, most ingenious, most expert. I I don't wanna be and again, there's an unmet need with pathologists. This is kind of like counterintuitive. We need more pathologists and we're building technology that's trying to imitate their tasks at a much higher level and qu and much faster. But can I for the audience, the people who are maybe quite far away are far away from pathology, can I make comparison of what you're trying to achieve with your technology, something that we already seen with radiology, where we have softwares that can scan, I don't know, humangous amount of CT scans, make predictions and find certain things that radiology cannot really find. Yeah, yeah. So do you think that's that's this is a comparable It's it's about one thousand times more the case. Why? Because it's a thousand times order higher in terms of resolution. This is a two edged sword. Okay. Uh-huh there's gr a much greater amount of detail. We're talking about hundreds of nanometers, they're talking about millimeters or so. Yeah. Not to be critical or anything. Radiology is important. But also radiology is seeing a more need for radiologists because they're able to have more scans. With pathology, this will be even more the case. If radiologists are safe, pathologists are living in a fortress. Mm-hmm. Okay. This is a good career. It's going to be a remarkably like luxurious one in my estimation. So if I had children that wanted to pursue pathology, I would say Go right ahead. You have my full blessing. This is gonna be an exquisite career. And it's gonna be so luxurious, in my estimation, with all of the advancements, especially the infiltration of hardware. Okay. Because AI, we don't need to worry about it. The avalanche is already here. Yeah. but when you integrate more AI capability in terms of hardware, now the pathologist is gonna be doing so many scans, they're they're only gonna be engaging The really exciting mentally tasks. So the higher order, they're gonna be looking at extreme edge cases. They're gonna be able to share it seamlessly with colleagues. They're really gonna have a nice intellectual puzzle to solve and be challenging the AI and then go further than that. Yes. If you have a number of these devices working in tandem, coordinating with one another, now you can reach for the stars. Yes. Now you can try to make training each other maybe. Absolutely. There's gonna be train each other. There's gonna be self training, there's gonna be automation. Yeah. Absolutely. And you actually I wanted to pick on a word you said, sending the information scans to each other. Let's move to that a little bit. So to me, this technology will allow like is like we just said, train each other remotely, but also doctors to be connected at different parts of the world. Absolutely. Exchanging the information. Yes, maybe you can demo there's a portal here. Yes. Let's see. To me that's the value, that's the future. And my show is about the future of healthcare. And I think this is the future when you can actually do this at your maybe a small hospital somewhere on outskirts. I don't know where it's Perfect great great example. Great great question. Yes. Absolutely. So frozen sections, rapid on site evaluation, what's called Rows. physicians can't usually go to the site. They need a technician to send them there. So the technician would load it up. Obviously very easy to use. Yes. We've showed you how to do it. I'm gonna try after I I I look forward to that. And so this is live now. You can access it through the internet. This is this. And you can control this to this. If I wanted to move it a little, let's move it a little. there. So we moved it. And so you can interact from here, controlling this. Yep. If you move the slide, it's gonna be updated on the portal. We worked very hard on this. So the other thing is all of this all of the scans and the stitches in different formats. Yep. So we have pyramidal TIFF is standard for now. They're trying to make DICOM more prevalent, which is the radiology. yeah. Well format. Yeah. Talk about formats. Format now is not a big deal. We can we can reformat it. But the point is, all of that information is on the cloud and all the deblurring happens on the cloud. We tried doing it on the device, it just wasn't convenient. So now we we want the pathologist to have No nuisances. Okay. You're just putting it here. You're scanning it. It goes directly. You have the private keys. This is all safe. It's all for you to see. Nobody else can can hack it or anything. If you de-identify it and give it to us for further training, we reward you for that because that's helpful for us to retrain more algorithms. and so the idea is you don't want to have all of this data stored on whatever heart whatever computer that you have. It's too much, right? This goes into the terabytes to petabytes really quickly. So we want to save the data and only put it in specific pathology softwares that you want. And so you can do that from any tablet, anywhere in the world, just logging into your portal and you'll have your own set of keys. So that's the idea. That's awesome. So let's talk a little bit about the future. So to me this is the future of pathology, but what would be your vision of the ideal pathway of a specimen from a patient who potentially may have, let's say, let's say cancer. Let's use that as a as an example. what is the ideal pathway that you hope to see in let's say next five, ten years for that specimen to go from the patient all the way to pathologist and back with results obviously with a diagnosis in the future treatment for that patient. Yeah. So time will be massively compressed. Mm-hmm. it's gonna be mostly on the pathologist side. So the patient is there and then they have to wait. They have to do the waiting. So that time compression, right now, time the timeline is like logarithmic. Everything is being compressed. Every day things are getting more and more advanced, and it's exponentially more intricate and more powerful. Yeah. So one day may as well be like 10 years. you know, that's an exaggeration, but that's what I'm trying to get at. Yeah. So everything will be accelerated heavily. Not only will it the the patient get the result quicker, but it's not only about, you know, you have an extreme scenario, you have a very big problem. Now people will try to monitor just for the sake of it. Most people have never explored their cells and tissues. Yes. Thinking about this just in terms of like a luxury, I think that is the future. Mm-hmm. And and just being able like you won't be able to have a brain biopsy every day. But there are certain Biopsies that are less invasive. Especially on the skin, I would say that skin, gastric tract. Yes. Blood smears are big ones. Every time you go to the washroom, a great amount of intricate, highly valuable biological information is being flushed down. Yes. And you could have prevented it. Yep. Prevented something potentially malignant from transpiring. I like that. I like that. That's almost it it's becoming like a preventative medicine in terms of easier, faster. And you don't have to wait for those biopsy results for weeks now. Sometimes weeks, depending on where what we're talking where we are. And so you would say the patient will come in, obviously the biopsy is done, the cut is made, the slice is ready, and potentially can be done right there. Right there and there. And also I think this notion where patients can get a a sample of their biopsies, 'cause usually a core. Yeah. So there should be no issue in giving the patient and if they want to have a a device like this at their house and then further I'm I I know this is a fantasy that might not be FDA cleared in the next to be determined. Yes, I agree. It is, yes, I should mention it is currently a research use only device. Yes. I I think we're gonna make headway very quickly. Yeah, because to me it's also it has such a utility for surgical applications. The surgeon who opens up I don't know, again, part of the body and they're trying to understand the margins of the tumor, they can also use it right there, right now. And they don't have to be pathologists. they can just do that, press the button, use the whatever nurse uses the iPad, sends it to all remote pathologists in specific pathologists too, very specific to that particular exactly case. Exactly. And you can have multiple coordinating working on the same if there's any questions. Yeah. There's no excuses now with the power of AI and The greatest minds leading the field. Yeah. You have now an an inordinate amount of ability to analyze this like nothing else. Well, Michael I really want to test it and try it out. I know we're kinda running running out of time, but I'm Did you want to try dragging your fingers on the screen just to see if it would move? I think that might be amusing. I'm actually probably covering the camera. Am I? Am I? Lice and easy. Nice and easy. Okay, yeah. Yeah, a lot of people they keep their fingers on you you wanna you you wanna f just flick it. It's it's a lot more delicate. okay. Yeah, see, it's easy. Yeah. And then you can try to move the slide physically if you like. But I have to tell you, like, it's been what fifteen, seventeen years since my last pathology class when I actually worked at microscope. So this is much more advanced. It feels so much better. 'Cause I remember that I've struggled with microscope and trying to find resolution now. It's like my god. Yeah. And so you can also move the slide physically here. I know it's I guess it's hard for you to read. really? If if well Give them my regards. If if my professor is watching, I think this is what you definitely would love to see. awesome. Yeah, so that can all be recorded. Yeah. And the way that you move around, yeah. It's like self driving. The way that you do it, the more data we have, the more we can replicate it. And huh. So you're training. We're training training. And so eventually, you know, within two or three weeks, you could just say autopilot. And then you're looking at the thing. Hmm. You know, it's a s it's a nice spectacle. Drinking your coffee. If anything looks amiss, then you interfere right away. like that. No problems. Okay. So it's like autopilot. Yeah. Sell driving car for pathologists. Multitask. Yeah. You can have you can have a dozen of these because the price point is so accessible. You may as well buy more than one. True. Well, I know we are getting close to the end of the show, even though I have so many more questions for you. But let's let's maybe finish with a couple blitz questions that I've done for you. I'll give you a short quen question and you give me a short answer and we'll see how it goes. Do my best. Working in the lab or building a startup? building a startup. Mm. I would love to know what it's more painful, but more rewarding. There are highs there there's more highs and lows. So you need to be comfortable with a great deal of fluctuations. And we're in California, so it's a lot of hills, and so I'm used to it now. Highs and lows. Okay. Fair enough. It's exciting, you know. Yeah. One technology that will change healthcare in the next five years. So I'm terribly biased and this may seem overly obvious, but I think the microscope. I think the microscope is a very underrated piece of engineering. And you know, when most people think microscope, okay, we've seen this before in biology class, you know, look at a piece of look at some cells, look at a tissue, next. But I think when you digitize it, when you couple it with AI, and then when you have self-improvement automation. It's gonna flourish nothing else. Yeah. And I do want to remind people that microscopes are so important still today. To be sure. I think people kind of forget how much they definitely do. They just don't realize that. Yeah. Okay. One thing that AI will never replace in medicine. Okay. So I I I I always thinking about this. There's a word imagination. I think that's an important word. There's another word, inspiration. That's an another very important word. And I think that at the highest levels, the 1% of the 1% of the 1%, the greatest AI, in my opinion, will never be able to touch it. You have this quote by Edison success or genius is 1% inspiration, 99% perspiration. The idea is hard work is what matters execution. I think with AI and all of the new hardware, et cetera, the robots, and this is a kind of robot, I think that that is going to be inverted. Completely toppled. It's gonna be ninety-nine percent inspiration, one percent AI. And actually, the useless daydreamers, the profitless visionaries, people who just spend time thinking about elaborate, crazy, grandiose schemes, this is their time to shine. Yeah. Because the AI will do all of the power for you. But that really that gem of an idea, what what Einstein had when he said Gluglische Gedanken, that initial spark of genius, I don't think that. c AI can ever emulate. And if you do, I'd I'd be very happy to see it. Because we'd get more Mozart, we'd get more Vivaldi, we'd get the greatest paintings, but I'm not seeing it. And I'm not sure if I want to live in a world where we replace the humans with you know if there's more Mozart. I know, but mm yeah. I need more melodies. Well we can talk about that more. Okay, so one thing that AI will never replace in pathology. Same thing. So there's a pathologist called Santiago Ramoni Cajal. He won the Nobel Prize for just looking at dry brain biopsies. Okay. And being able to he he did it he did a number of discoveries, but he was the one who interpreted neurons as dynamic separate entities. And so I think now pathologists pathologists tend to be into research. Like there's a great there there's there's a a a significant fraction of pathologists who are inter interested in in research and publishing and so on, I think now they can really reach for the stars and try to make some absurdly ambitious breakthroughs. Because when you have all this AI capability, you really just need the the germ of an idea that's very promising. And then you could just let the AI to do it. Carry it out. So be I think pathologists should be a little bit more ambitious in terms of intellectual pursuits. Yeah. And just Taking the whole field of pathology to a new level. I think that's maybe fanciful, but potentially realistic. It kinda answers my next question, but I'll still ask you. So will AI replace pathologists? Yes or no? Absolutely no. There's gonna be there's first of all, let's deal with the current sobering reality. Unmet need. We need more pathologists. Not enough pathologists for the population are being reintroduced to the field. We need more pathologists. It's a great career. It's gonna be a marvelous one. No. Okay. And the last one, if you had one wish for How AI gets used in medicine, what that would be. that's a general question. Mm-hmm. You know, I think that the infiltration of AI in every sphere is very good. Why? Because it's derivative in nature. So it's taking out what is derivative. It's taking out the menial. It's taking out the drudgery. It will force us as humans, as a whole civilization, to aim higher and to go for what we used to, you know, highbrow that i a little bit s it could be snobbish, but try to It try to achieve something of a high intellectual pursuit. Yeah. I think that's good for medicine. And so as much AI as possible. That may not be a popular opinion. I think it's good. I think it's a boot. As much AI as possible. Yeah. Awesome. Thank you, Michael, so much. Thanks for bringing this little guy. let's maybe the last thing I want to ask you, just tell the audience where they can find you. Yeah. How they can find this. And you said it's a research only use right now, but it's available to purchase. So please tell tell. Well it it's available well we have many commitments. So we have an initial deployment in October. I'll be at if any if any pathologist is watching Cap twenty six in Vegas and then Path Visions in San Diego. Please come by to my booth, bring your own slides if you're hesitant about how this algorithm works. And yeah, scanamus.ai. You can follow me on LinkedIn. And I'll share all those details underneath video and on the social media. Thanks, Michael. Thank you.