[00:01:39] Aleksandra Zuraw: Today my guest is Ben Cahoon from Techcyte. Techcyte is a software company providing digital diagnostic tools for medical, veterinary and environmental laboratories.
[00:01:51] Hi, Ben. How are you today?
[00:01:53] Ben Cahoon: Doing well. How are you?
[00:01:55] Aleksandra: Good. Good. Thank you so much for being my guest today. And obviously, we start with introducing our audience to you. Tell us about yourself. What’s your background, about the company, about Techcyte and what’s your role in the company?
[00:02:09] Ben: Okay, perfect. Well, I actually started with a computer science degree and then got really lucky to join Intel. We were the only software division at the time and we built desktop management software for IT departments. So Intel was a great company to start off my career and I was lucky enough to have them pay for me to go back and get my MBA. And after that, since about 2003, I’ve been purely an entrepreneur, building everything from a turbo company. People thought I had I lost my mind when I went and did that, but we built a turbo company and we grew up an ed tech company-
[00:02:48] Aleksandra: What’s a turbo company?
[00:02:50] Ben: Turbo company. We had a patent on a remote mounted turbo system. So you take your muffler off your car and you put a turbo in the back and you run a charge pipe up to the front. So we would increase your gas mileage, give you about 150 horsepower extra. So it was a win-win. We literally took the performance world by storm.
[00:03:15] Aleksandra: That’s interesting.
[00:03:16] Ben: Yeah. All my friends in the tech world thought I’d lost my mind, but it was really the first time I got the opportunity to run a company and I learned just a ton. It was great. But so after that, I ran an ed tech company that we eventually grew and sold to Lenovo. We then moved to Luxembourg and started a FinTech company. I came back from that and built 45 town homes in Moab, Utah. So if you like mountain biking, I’ve got a good place in Moab you could go to. And now, doing a med tech company called Techcyte.
[00:03:51] Aleksandra: How is med tech different from the other tech?
[00:03:55] Ben: The challenge of med tech, there’s two things. One is all of the studies that need to be done to prove the efficacy of your product, and then the regulatory burden. So getting your products cleared and through regulatory is a big deal.
[00:04:13] Aleksandra: So who are you at Techcyte?
[00:04:16] What’s your role?
[00:04:16] Ben: Yeah. After coming back from Luxembourg, I was introduced to Techcyte and I was immediately intrigued because it was combining the web and artificial intelligence and healthcare, which addresses part of the largest parts of our economy. So I knew this was going to be a great company and I’m currently the CEO, but I’ve been involved with every part of the company except for the actual coding. I don’t code anymore. And And one of the key roles in an AI company is annotating the data because you need thousands of images to be able to create good algorithms. So we call that clicking, and I have done probably more clicking than anyone in the company, but the real role that I want on my tomb stone is fecal innovator. Again, it’s so hilarious. Some of our first successes have been in looking for parasites in dog, cat and human feces. We have literally innovated on how to change the world looking for parasites in feces.
[00:05:29] Aleksandra: I think you’re one of the very few CEOs that actually did annotations.
[00:05:36] Ben: Yeah, I have done a lot of annotating. You really need to know how to use… My background is as product management, and so you really need to know how to use your product. And in our case, the users of our products are the labs and the doctors and the clinics, but also our internal people. Data is key to creating good algorithms.
[00:06:01] Aleksandra: so we’re going to go more into detail, but Techcyte is digital pathology, but it’s digital pathology for cytology and for smears and for wet preps, right? It’s not for tissue.
[00:06:14] Ben: That’s correct. Yeah. So we actually like to call it digital diagnostics because when people think of digital pathology, they think of histology. They think of finding cancer in tissue.
[00:06:33] Aleksandra: That’s why I wanted to clarify that.
[00:06:35] Ben: Yeah. So we’ve gone to all of the digital pathology shows, trade shows and conferences over the last five years and it’s pretty much digital histology, looking at tissue. We came out of the university of Utah and ARUP back in 2013. And one of the key relationships there that came out of that was a relationship with Dr. Mohamed Salama. Who’s now running the Mayo Clinic’s reference lab. He’s the chief medical officer there. He helped us very early on see that the market was going to sort of break up into two areas. One will be digital pathology or histology, and then everything else. So we really do everything smear, meaning a liquid like blood, fecal, which is smearable, bacteriology, which is smearable. So we’ve taken that liquid smearable approach and really focus on cells and that’s what we’re good at.
[00:07:38] Aleksandra: So from the computer vision standpoint, detecting the object of interest in your algorithms and for annotations, you called it clicking. Is it because you are using the bounding boxes?
[00:07:53] Ben: Yes. Yeah, that’s correct. So we’ll take a gigapixel image and we call it our Techcyte data pipeline, and you’ll start off annotating the object of interest. Then we put it through the pipeline, which generates a model, which then does a finder run, which generates more boxes. So eventually, you’re not doing as much clicking and bounding boxes as confirming and fixing what the model, how if it makes a mistake. So that way, you get really efficient on changing the things that are most impactful to changing the model.
[00:08:32] Aleksandra: So you’re using the reinforcement learning or so-called active learning for…
[00:08:37] Ben: Correct.
[00:08:40] Aleksandra: I wanted to clarify about the bounding boxes because annotation in the histopathology world has different meanings. One of them would be clicking on the cells, but another one would be delineating all the structures. So this is not the approach here. This is for the detection
[00:08:58] Ben: that’s correct. We count… Yeah. Many of the places we go to, they care about the quantity. So we’ll count. The first job of the AI is to box what we care about and then to classify it. Then once you do that, we can get a count.
[00:09:13] Aleksandra: Where are you located?
[00:09:15] Ben: So we are located actually, all over the world. We have head in Utah and Luxembourg, but we’ve adopted a remote first approach to building a company. I don’t know if you’re familiar with the GitLab and how they have really open sourced how to build a great remote first company. But remote first basically mean means that we can hire people all over the world, and that has changed our culture, how we communicate. I had previously built a company using Skype back starting in 2006 to about 2013. We used only Skype to communicate. We never had an office, so I knew it could on, but then when we came across GitLab and their remote first culture, it really just was… Then when COVID hit, it just became obvious that for us to get the talent we need, our solutions are worldwide. So we need to be able to reach talent worldwide. So we literally have people from Luxembourg to Las Vegas to Mexico.
[00:10:24] Aleksandra: So obviously, you also offer your product all over the world. It doesn’t mean where your clients are located.
[00:10:32] Ben: Yeah, that’s correct. Well, our clients, they eventually are clinics or labs and they’re going to have a scanner, a whole slide scanner in that location. But then the images come up to us, and as long as you’ve got a web browser, now you can access the solution.
[00:10:51] Aleksandra: So you mentioned Techcyte came out of the University of Utah in 2013, right?
[00:10:56] Ben: Mm-hmm [affirmative] .
[00:10:59] Aleksandra: Can you tell us a little bit more about the creation story of Techcyte?
[00:11:03] Ben: Sure.
[00:11:03] Aleksandra: How did it happen and what was the driving force to spin off?
[00:11:08] Ben: Yeah, absolutely. So Ralph Yarro, he’s a successful Utah investor, he hired a computer scientist/MBA. His name is Rick Smith and they worked with the University of Utah and transferred some technology out of there concept. When they got into it, it ended up not being what they really wanted. So they started from scratch and sort of built up a prototype in Ruby on Rails on a blood classifier. Then in 2016, myself and some other friends who had had successful exits in software, we invested and joined the company. Then we rebuilt everything from the ground up in Google Go’s in our backend. We use React on the front end.
[00:11:56] Then we also changed the strategy of the company to go after veterinary, human and environmental markets because we knew that human was going to take a long time. Human are massive markets, but they’re going to take a lot of money and a lot of time. So we started off going to environmental, which is mold testing, and veterinary, knowing that they’re unregulated and if we created good high quality solutions, that the customers could adopt them quickly and that would help fund the company. So we don’t have to give up so much to the VCs.
[00:12:33] Aleksandra: What is Techcyte’s mission then?
[00:12:35] Ben: So our mission is to digitize and automate diagnostics through AI. It’s important because in healthcare today, it’s too expensive and there are too many mistakes are made. So we can make a significant difference in both of those areas. At a fundamental level, everyone on our team wants to make an impact on the world. So by building a remote first organization, we can really start it with ourselves and how we interact with our family and then really reaching out to our customers. So even though our stated mission is to digitize and automate diagnostics and that’s what we’re building, we really do want to make an impact on the world.
[00:13:18] Aleksandra: You said there was several times you had to build it from scratch or pivot. So what were the things that were failures at the time that set you up for success? What did you have to change and how did it influence the company and where it goes now?
[00:13:37] Ben: Yeah. So at the first, I mentioned that I’d love to have the words, “fecal innovator,” on my tombstone. The reason why, we feel a bit like Edison who tried 10000 ways how to not build a light bulb. We feel like we literally tried every single way and failed so many times on how to take a fecal float, get a good image, and then create good AI. It was way harder than we thought. So one of the key problems that we had to solve there was around image acquisition. As you probably know, all of this scanners do great at finding tissue. So if you put a tissue in a whole cyte scanner, it’s going to find the boundaries, it’s a mono layer, so it’s going to scan it in. It’s got lots of material there to focus on.
[00:14:30] If you put a fecal float in any scanner, all you will get is a blurry image because there’s so little actual material there to focus on. The eggs are microscopic. So we tried everything from wheatgrass powder to microbeads, and we ended up finding a very elegant solution on printing dots on a cover slip so that the scanner will focus on those dots and then we know the eggs are below those dots and we scan those areas. So now, we get beautiful images, which allows us to get amazing AI results, which allows us to go after a pretty big market. It’s surprising. If you take production animals and companion animals, there’s probably over a hundred, 150 million tests done every single year. So it’s actually a pretty big market.
[00:15:18] Aleksandra: This is always mentioned as a challenge for cytology, that the focusing is a problem that you have to do z-stacking. I think this is the first time I have heard about this kind of solution to find a fixed point, focus on that, and then scan below. It’s really interesting. Any other challenges that were in cytology versus pathology versus histopathology?
[00:15:47] Ben: In most of the cases, the reason why our technology has diverged from other AI companies that are doing tissue is, again, as you mentioned at the beginning, they’re trying to find boundaries of cancer and we are trying to find and count very, very small… For example, in a mold test, you’ve got 250000 to 500000 objects that we’ve got to count and classify. So you’ve got to look at your precision or recall, and with recall, getting everything boxed and then called correctly is very, very difficult. So getting the technology, our data pipeline, the process, all of the hyper parameters that you use in AI, to get that all to work well and produce really good results, and then prove to yourself that you’ve chosen the right model and that the model’s getting better, it’s a huge challenge. Then to take that and make it in a worldwide platform that’s always available that’s HIPAA compliant, it’s CE marked, it’s definitely harder than we ever expected.
[00:17:01] Aleksandra: Can you tell us a little bit about your pipeline?
[00:17:05] Ben: Sure. So what we’ve built is a web-based platform for deep learning image analysis. And again, we’re focusing on liquid and smearables and cells. At the core, the goal of the pipeline is to deliver fast, accurate results to patients. But to do that, you’ve got to start from all the way back to sample prep. So we work with sample prep vendors who then work with the scanner manufacturers to make sure we can get a good image. Then that goes into the pipeline. The pipeline starts at annotation. It then goes into what we call finder runs, where we try to find additional images. Then it goes into sort of the curation stage where we’ve developed tools.
[00:17:50] Actually, I wish I could show it to you. It’s a 3D modeling tool. We can be basically see sort of into the black box of what the AI is seeing. We take all the feature sets and put it into 3D space, which then becomes basically a 3D view of groups of the images and that allows us to very, very rapidly curate and clean data sets because if you get even just a few bad images in hundreds of thousands of images, you can start to take the algorithm in places where you don’t want it to go. So that whole process is really, if you look at the front end part of our deliverables to the clinics and the labs, what they see versus what we use internally, there’s probably as much complexity on both sides.
[00:18:44] Aleksandra: Do you have a video of that 3D visualization?
[00:18:49] Ben: I do. Yeah, I could get that to you later.
[00:18:52] Aleksandra: I can link it in the show notes.
[00:18:54] Ben: Okay, perfect. Yeah. I’m glad to show you that.
[00:18:56] Aleksandra: Aleksandra: Yeah. So what products do you offer?
[00:19:00] Ben: Okay. Today, we offer, as I mentioned, we go after veterinary, human and environmental clinics and labs. So we’ve got a mold test. It counts in classifies about 150 types of mold and particulates. So that’s for homes and buildings. The next one is the veterinary market, and we’ve partnered with a company called Zoetis. They’re a worldwide leader in pharma, diagnostics and reference labs in the veterinary space. We launched a product in September with them that does fecal, ova and parasite. It finds ova parasites in dog and cat feces. It does a really, really good job at that. The studies that we did with Zoetis showed that we’re as good, that the AI got to the point where it’s as good as a parasitologist at Oklahoma State University, and there’s been multiple journal articles written on that. Then we’ll go into the human side, which we’re doing pap smears, blood smears. So a white blood cell differential, bacteriology, gram stain, direct smear. Then there’s other things that’ll be coming.
[00:20:17] Aleksandra: I am a veterinary pathologist myself. So…
[00:20:20] Ben: No way.
[00:20:21] Aleksandra: Yes, I am. I am a veterinarian by training and then specialized in pathology. So just to emphasize how important this parasitology aspect of veterinary medicine is because in humans, usually you only worry about parasites when you’re immunocompromised or if you maybe traveled to some exotic locations.
[00:20:44] Ben: Correct.
[00:20:44] Aleksandra: Whereas in animals, this is a daily problem. You should deworm your animals every three months to control the parasites. So being able to detect it rapidly and reliably is very important. And like you said, it’s a huge market because there are many, many companion animals and this needs to be controlled.
[00:21:09] Ben: Yeah. And if you take that to production animals as well, there are many production animals, they just sort of deal with the parasites and they’ve got to control it. So they’re doing equine, bovine, poultry. All are doing fecal testing as well.
[00:21:27] Aleksandra: Obviously, not only companion. I actually did one year of bovine practice. So I worked with that as well.
[00:21:36] Ben: Wow. Yeah, that’s awesome. So we’ve taken that and had a lot of success there, but then we’ve also done the same thing on the human cytes. We’ve got a human ova and parasite test. It’s been in production since August of 2019. Probably over a hundred thousand samples have been put through there. It works really well. There’s a journal article in the Journal of Clinical Microbiology that came out on that. We did it jointly with ARUP, Dr. Couturier and Blaine Mathison , were just instrumental in helping us there. We’re techies. If you look at the core of our company, we’re techies. So we’ve really relied on and had to develop good relationships with expert on human, vet and air quality around the world.
[00:22:21] Aleksandra: So is this your model of working with experts? You have them externally or do you also have some experts in your company? How does that work?
[00:22:30] Ben: Yeah. It started off everyone, it was all external, but as we’ve matured, we do have hematopathologist on staff and we have multiple people like that on our advisory board, but now, with our relationship with Zoetis , they’ve got experts around the world that we work with and they’re really… Because we work so closely with them, it’s almost like we’re on the same team. Then we’ve developed a relationship that started off early on with ARUP. We made such an impact with that fecal test, they saw the vision of how this can do many other things. So they’re doing a pap smear study with us right now. We’ll be doing blood study on white blood cell, red blood cell differential. We’re about two thirds of the way through our FDA clinical study on that. So you’ll hopefully be seeing a cleared algorithm from us here in the near future.
[00:23:26] Aleksandra: Congratulations. What about your competition? In some areas, probably you don’t have too much competition because it’s very innovative, but in hematopathology, in cytology, there have been cytology like pap smear where the first applications of digital image analysis. How do you position Techcyte there now at the moment, and what differentiates you from the rest of the people providing these services?
[00:23:55] Ben: Perfect. That’s great. So you’re exactly right. There are some entrenched areas that we’re going after, like blood with CellaVision and pap smears with Hologic . And in those cases, they’re using technology that’s older technology that we can then use deep learning tech technology and take it further. We will use non-proprietary hardware. So in CellaVision example, it’s a very, very proprietary hardware where we could use any whole cyte scanner. So we could take advantage of a small scanner all the way up to a large Hamamatsu S360. The other thing when, for example, with Hologic, they’re going to want to support their prep system and we’ll be able to support Hologic’s prep system, BD’s prep system. There’s others around the world where we can use the power of deep learning to find common features across all of them and take the technology further.
[00:24:54] Aleksandra: Let’s say I want to be your customer and take the advantage of an animal, my dog. I want to use your test on my dog. Is this something I can do? Or does it go… What platform does it go through? Is it for individual animal owners or more for companies? How does that work?
[00:25:13] Ben: Yeah, so there’s two ways our technology will be applied specifically in the vet market, and it’s the same as human as well. You have reference labs and then you have clinics. The reference labs will use the technology on big, huge, large, fast, reliable whole cyte scanners, like the Hamamatsu S360. Then you’ll also have, for people who just want to take a fecal and send it off to the reference lab, we’ve got that solution. In that case, we’re helping the reference lab be more efficient and faster and more accurate.
[00:25:48] Where this is revolutionary is this will go out to the veterinary clinics. There’s about 70000 vet clinics worldwide. So I don’t know if you’ve… Yeah, you’ve interviewed Mika at Grundium. So we take their small eight inches by eight inches scanner. That goes into a vet clinic and we worked very, very closely with them. So you put the slide on the scanner, you don’t even go to the Grundium UI. You just go to our UI and hit scan and it automatically finds those dots, the fiducials on the cover slip. It sends it up to us. We analyze it and the results are displayed to the vet technician.
[00:26:27] Aleksandra: Oh, great.
[00:26:28] So it’s based basically throughout the whole, I’m just taking animal care, as an example, throughout the whole animal care process. You can basically get it at your vet’s clinic.
[00:26:42] Ben: That’s correct.
[00:26:42] Aleksandra: Or if they are sending samples to a reference lab, it can be implemented there as well. Great.
[00:26:49] Ben: That’s correct.
[00:26:49] Aleksandra: So you said you’re techies. How do you innovate on both sides, on the scientific medical veterinary and environmental side and on your technology side? Because AI and the deep learning, this is cutting edge. This is new, but it’s moving. It’s moving and advancing. So how do you follow the trends?
[00:27:13] Ben: Yeah, we do have to take an overall comprehensive look at the process. Again, we were talking a lot about fecal here, so let’s continue on that trend. We went to… Being a veterinary pathologist, you’ve probably done a fecal float. Have you done a fecal float before? How you have to… You first take the fecal material, you mix it all together in a cup. Then you pour in sugar, water, and then you have to put it down a test tube and it’s such a messy process. It actually overflows the top of the test tube when you put the cover slip on the top of the meniscus. As we looked at that, one, I think vets like to send their fecals off to reference lab because it’s such a messy process.
[00:28:00] So with our relationship with our customers and our employees in Luxembourg, we went over there and went to Medica, which it’s got to be one of the largest trade shows in the world. It was 18 halls of everything medical that you can imagine. On the third day of walking that, we met a company from the UK called Apacor. So Apacor, they baked this little device here and they’re the leader in the world on human fecal concentration. So you take the feces, it goes into this… You use this little scoop, you put it in here and it really makes it a clean process. It then goes into centrifuge and centrifuges down and you don’t have everything overflowing. So we worked with Apacor, modified this product, so it could be used in a veterinary in area and then made it so it’s clean, efficient, easy to do a fecal float.
[00:29:02] Now, why didn’t someone do that? Someone in the veterinary industry, I have no idea, right? But I think it’s because we took an overall look at the whole process. We knew that tech and AI was going to play a very critical role, but you’ve got to create a simple process that goes from the very beginning. So once we did that and then we found out that when you stick a fecal float in a microscope, it returns a blurry image. Then we had to innovate there, and then that’s where we came up… Actually, I’ve got a picture of it here. That’s where we came up with the cover slip idea. So here is an example of one of those cover slips. Hopefully, you’ll be able to see it.
[00:29:45] Aleksandra: Yes.
[00:29:46] Ben: Yeah, you can see the dots. So that there’s printed fiducials And we did lots of studies to find out where the eggs are floating in that mixture. So if you think of a cover slip, the mixture of fecal material and then the slide, the eggs actually have a specific gravity and float. So we did a bunch of studies to find out how many microns beneath those fiducial dots the eggs are at, and that was the breakthrough that we patented and figured out how to get good images. I’m partial now. They’re beautiful. The images that we get of these fecal parasites, they’re beautiful and the algorithm just does unbelievably well.
[00:30:25] Aleksandra: I’m laughing because now I start understanding why you want to have, “fecal innovator,” on your tombstone.
[00:30:31] Ben: Exactly.
[00:30:32] Aleksandra: A lot of thought into this process. That’s for sure.
[00:30:37] Ben: Yeah, it is.
[00:30:39] Aleksandra: So is there anything else that you would like to share with the listeners? So we talked more about parasitology, but okay. Let’s touch a little more on the pathology, on the pap smear and the hematopathology. What are your products there?
[00:30:54] Ben: Yeah. So let’s start off with blood because that’s where everything really started for us. We do a white blood cell and red blood cell differential. That’ll apply to humans and animals. Basically, we count and classify white blood cells and red blood cells. We view that as a core test that will lead to other things in the future. Bone marrow aspirates, fine needle aspirates all will stem from that. Inclusion bodies such as malaria, babesia, bartonella, other things like that, even leading to sepsis. So if you look to the future, we see the first step phase one is to automate an existing test. So people are looking and doing parasitology, they’re doing hematology today through a microscope. So we’re going to automate and improve that existing test. But then in phase two, we’re going to really make it so…
[00:31:56] In most of these tests, like a fecal test, 95% are negative. So we’re going to reduce the need to look at the negatives by doing a secondary classifier. So we see that as phase one. Phase one is automate. Phase two is do a secondary classifier, which will eliminate negatives. Phase three is then we will present a diagnosis and an expert is just going to confirm the diagnosis. It’ll be very, very quick. After that, phase four, we see is using inexpensive tests to eliminate inexpensive tests to eliminate expensive tests. So you could use a blood smear in hematopathology to eliminate a flow cytometry.
[00:32:36] For every flow stain that they use, that’s a $400 test. So if we could do more with microscopy and morphology by using features that humans don’t even see today, that gets us to that phase. Our ultimate goal in phase five would be to use AI to perform tests where microscopy is not the gold standard. For example, sepsis. Sepsis today is a PCR test, which is slow and expensive. And we think we could do all of that with morphology and looking for features that humans don’t readily recognize such as the monocyte distribution with, what percentages of white to red blood cells. There there’s a whole bunch of stuff that the AI is just going to be able to figure out.
[00:33:26] Aleksandra: This is interesting. I have never heard or I have never read about this being done in cytology, that you want to derive non visual features, or let’s say features not perceivable for humans to let AI do the work and predict something. This is however, now being done in histopathology where you predict mutational burden from the images. This is something a pathologist is not able to detect visually, and sequencing is done for that. Exactly the same trend, or parallel trend is happening in histopathology with the same rationale to eliminate the expensive tests and replace them by same quality inexpensive tests.
[00:34:22] Ben: That’s correct. Yeah. We have that exact same vision and that will occur from starting off in blood and other things, but then also with pap smears. So pap smears, we’ve made a lot of progress on a secondary classifier because again, with pap smears, you’re dealing with a lot of negatives and you really want to be able to eliminate the negatives. The current technology with Hologic and BD, I think their FDA clearances, I think for 25% that you can just eliminate. With deep learning, we think we can go much way, way further than that where you could eliminate 50%, 60%, 70% of the negatives.
[00:35:02] Aleksandra: Indeed. But I think it’s not because of… So at least the 25%, I’m not sure if that’s not the regulators being cautious because I think the technology is okay to eliminate more. Did you already engage with regulators on that, and do you have any-
[00:35:22] Ben: yeah.
[00:35:22] Aleksandra:… feeling about that? Do you think…
[00:35:25] Ben: So we’re starting off just with the base. We’re starting off getting our phase one cleared, which is automate an existing test. Once we get that done and prove that you can use different scanners and you’ve probably heard of softwares and medical device with the FDA. So we’re working through that. Once we’ve proven that and they gained some confidence in the technology and in the process, then I think the next step will be to look at how much we can eliminate from a negative perspective. The existing technology really does use a shallow classifier and it doesn’t have the probably even a capability to do more. With deep learning, we’ll be able to go past that for sure.
[00:36:11] Aleksandra: So gaining confidence in the whole process and also interoperability would be the two main things that you have to demonstrate to the regulators to hopefully have it cleared for broader use than what’s already there.
[00:36:29] Ben: Ben: Yes.
[00:36:31] Aleksandra: In the whole process, what is your main roadblock? What is preventing you at the moment from scaling? Let’s say you would like everybody to use this. You have everything cleared by the FDA or whatever regulatory agencies. Is there anything that is outside of your control that still would need to improve for you to be able to go broader?
[00:36:56] Ben: Yeah. We run into scanner issues with every single sample type that we deal with. So we have a love/hate relationship with the scanner manufacturers because we absolutely rely on them and we’ve got good relationships with them. I think we’re one of the only companies that I’m aware of that has worked with key scanner manufacturers to get access to their API, which then allows us to put AI and computer vision in front of their scanner to then make the region of interest and the focusing better than it would be if they’re just up to their own focus algorithms.
[00:37:49] Our relationship with the scanners and the scanner manufacturers, and I hope the technology, they need to be fast, reliable, inexpensive because all the AI companies and all the… People aren’t just going to digitize just to digitize. There’s got to be a reason and the reason is generally accuracy and lower cost. So AI gives you the ability to get both of those things. We can be more accurate and we can do it with less cost and that will then drive more digitization. So our relationship with the scanner manufacturers is good and I think that will be a competitive differentiator as we go forward because we can make their scanners better for non-tissue sample types.
[00:38:38] Aleksandra: When we look at the whole technological innovations, it’s funny that somebody has to play catch. AI ideas and deep learning is not new, but in the seventies, there were no hardware capabilities to move those ideas forward. Now, we have the hardware capabilities we can move forward. There’s plenty of people who want to innovate in healthcare and innovate in digital pathology. Now, the scanning is the bottleneck
[00:39:11] Ben: yeah, we’ll keep pushing each other.
[00:39:11] Aleksandra: Yes. Bottleneck. So it’s like that, I guess every area of technology. But yeah. If there is enough incentive like you say, then the people who have to… Or the area in this pipeline that has to catch up will have to catch up and they will because that’s what we do. But digitizing just for digitizing, you can do that as well. It has value, plenty of value in education, but beyond that, there’s need to be… If you don’t know what’s about, it’s usually about money. So there has to be good return on investment for people to pursue this. I hope there will be and in many areas, there already is. That’s why people are pursuing this.
[00:40:01] Ben: Yeah. Health care is interesting because it really has to do both. You’ve got to save money, but then you’ve got to be more accurate. Just one last example. We have found instances because the AI is so persistent at looking at every single pixel, we’ve found instances where we’ve found, technologists and using our AI have found parasites that they were then not able to actually find on the glass slide. They know it’s there and they still can’t find it and the AI finds it. So I think that’s the future. We will be able to help humans become better and really become more accurate and lower the costs.
[00:40:44] Aleksandra: For the publications and the video that you mentioned, I’m going to link this in the show notes and perfect. Thank you so much for taking your time and walking us through the Techcyte digital diagnostic process.
[00:40:59] Ben: Thank you. It was a pleasure to talk with you and we appreciate the opportunity.
[00:41:02] Aleksandra: Have a great day.
[00:41:03] Ben: You too. Thank you.
