Aleks: On average tech startups last from 2 to 5 years. I’m proud to say that in the digital pathology space, they last longer. And today my guess from Indica Labs has been in the space for over a decade, so let’s welcome her.
Welcome digital pathology trailblazers. Today my guest is Kate Lillard. She is the chief scientific officer at Indica Labs.
And Kate, this podcast is so overdue. What we were supposed to do it when Indica was ten and now Indica is already 12? But here we are meeting and I want to learn what happened in those 12 years of Indica Labs. But before we dive into that, tell me about you. Let’s start with you.
Kate: Okay. Thank you. Thanks. It’s a pleasure to be here today and to be able to talk to you. I mean, as you said, we’ve been trying to make this happen for some time now. So I’m I’m like super excited to be here. So a bit about me, my actual background is really in cancer research, though.
I did a doctorate in molecular genetics and biochemistry, and that was at the University of Cincinnati Medical Center in the early 2000. Went on to do a postdoc at UT Southwestern in Dallas. And I think my plan was 100% that I was going to stay on sort of academic path at that point, demand Yeah. So about, I guess, 40.
Aleks: Year employee number three. So you’ve been there forever for all.
Kate: Yeah, Yeah, I was, I was about four years into my postdoc and my advisor unexpectedly passed away, unfortunately. So ultimately I decided rather than starting over again on a new postdoc, it was a good time to change track. And I joined a Perio at that time. That was in 2008, and that’s really how pioneers. Yeah, I landed in digital pathology at that point.
Kate: I didn’t have any experience in digital pathology before that and lots of experience in microscopy and imaging. So I was excited about the space, but not digital pathology. So it was relatively, I think, early days in the field and quite an exciting time really to be in it. And as you said, yeah, I think it was a leader in the space.
So and altogether a great company too.
Aleks: To work for. So you were there and then you moved to India. How did that happen? What was your career path to the place where you are now? The Chief Scientific Officer.
Kate: Yeah, sure. So actually I think it was sort of a natural transition actually. So the CEO of India lab that’s in and still is Steven Harrigan and me, we were both working at a period at really the same time overlapping settings on the development side. And I was more on the applications and product management side, but we were both remote employees.
I was based in the UK and he was based in New Mexico. So we would regularly come into the office and Vista, California at the same time. So we spent quite a bit.
Aleks: From the UK and New Mexico.
Kate: Yeah, Yeah, exactly. So he comes.
Aleks: From a large commute.
Kate: Yes definitely. Once a month. Once a quarter quite often. But actually one day in late 2010, maybe early 2011, he told me about his plans to sort of spin off this company into the labs. And it was the start of a period because he was going to take the Mario SDK and develop research-focused algorithms that speak employed.
Yeah, the SDK software development kit to develop rhythms for you that could be deployed in the image scope viewer. So I was excited about his, plans, and because of the sort form of customer research customers were, my own bread and butter customer at Furio. So again, more potentially more products to sell in the near term.
But I wanted to work with him on this sort of new venture because I knew there was a lot of need out there. So I told him at that point that I was going to join him as soon as he had the company established and needed to commercially launch the product. So that happened in 2012.
So was just about a year after him establishing in the collapses the company and when I.
Aleks: Joined that’s fast first startup.
Kate: Yeah it was it was.
Aleks: Years as it was able to like to have people join him.
Kate: Yeah and I am single.
Aleks: For longer than the year.
Kate: I was employee number three. So Blood Photograph was actually employee number two and he was the first developer and is still in a dark lab as well. But yeah, so I joined as chief scientific officer and obviously as CSO. I was then and still am involved in a lot of sort of strategic sort of decision-making about what we’re doing in the company scientifically.
Kate: But I’m also actually in charge of a sales and marketing team management air dining office thing. Yeah, so I do a lot of different things and I think a lot of people find that strange that a CSO would sort of manage the sales and marketing side, but, everybody that we employ in sales to do the sales or to write our marketing continent, they’re all scientists themselves.
We don’t have sort of a traditional sales team, at least not on the research side of the business. We try to employ bench-level scientists, a lot of our previous sort of customers, and we use sort of a consultative approach to try to figure out what the customer you’re trying to do scientifically, what data they need to generate.
And then we try to design a solution that works for them. So it’s a little bit different, I think, than a lot of other companies in the space. And maybe if you look at it from that perspective that everybody sees and this and so the marketing team is a scientist, it might not be as strange that they’re managed by the chief scientific officer, but that is my job in a nutshell.
Aleks: So the three original employees are still there at the Indica?
Kate: Yeah.
Aleks: Congratulations.
Kate: Steven Rutter and myself, we are all still there.
Aleks: That says a lot about the company. And yeah, just to comment on this, this is an interesting and I think very a fit for purpose approach for the scientific environment, the scientific digital pathology environment, image analysis environment to have people that the sales have an in-depth knowledge of what this product is being used for, that’s kind.
Kate: Of not so it’s somebody in between or on the sales side, but it’s yeah.
Aleks: And you can have exactly you can have direct communications. I don’t know how to say there are no unnecessary links in this chain and I don’t want to say that sales are unnecessary, but often there needs to be a translator. There is no need for an interpreter.
Kate: Let’s go. Right. Exactly. The person who’s setting the quote together knows the customer’s need end of the day.
Aleks: So about the company about in the collapse. Tell me like the nuts and bolts, where are you located and why there and where are you located and why not where the main company offices and all of this backstory about that.
Kate: Yeah, sure. So actually as I mentioned before, I live in the UK. I have lived in the UK before joining in the collapse. I moved here when I was working for a period. So that’s sort of historical I guess, but in the lab. It’s headquartered in Albuquerque, New Mexico. So initially we were a little bit further north and for Alice, New Mexico.
Kate: And by the way, Steven Pash, again, the CEO, is from New Mexico. So that’s the main reason we were based here, to begin with. But we moved from Horowitz, New Mexico, into sort of Albuquerque city proper and to sort of shiny new headquarters in 2019, which was interesting. I mean, oh yeah, as we were definitely sort of bursting at the seams at our old office and we moved into this amazing new headquarters.
And then, you know, we’re almost immediately bursting at the seams again, but then hope it happens. So that kind of change changed everything in terms of who was up.
Aleks: In office space.
Kate: It freed up some office space. But I think Albuquerque again, and Steven’s sort of hometown, but it’s also a little bit underappreciated. This is a tech hub that is a great place for a software company like ours to be based because there are a few local universities that have solid computer programming courses. There are also government labs.
The Sandia and Los Alamos are just in the immediate area, so they have a lot of focus in charter programming and AI as well. And not everybody wants to work in the defense sector, so we get quite a few people moving over to end with us on the development side for that reason.
Aleks: Oh great. So how many are you now? How many people work for the company at the moment?
Kate: Yeah. So as of this morning, we were 111 people. I have to keep looking. I think that’s changed quite frequently. Oh, yes. 111 people. And I think more than half of those, about 60% or so, are actually based in the headquarters, and the rest whatsoever, the engineering folks and product development. And then we have quite a big deal staff team that are customer facing like the application scientists and also technical support I.T.
Kate: And we’re based everywhere like we did a team in the UK, as you probably imagine, meeting based here, but also in Europe. We got direct sales support in China and Japan and also of course many in the US and Canada as well.
Aleks: In Print International. So are you guys still growing? Are you expecting Indica to be 211 people next year or is this the perfect size? It depends the what the business world will bring.
Kate: It depends on what the business world will bring. But we’re definitely still recruiting 200 employees next year. Probably not, but it’ll definitely grow.
Aleks: So do you have open positions at the moment?
Kate: We do have open positions at the moment, mainly in the sort of development side, obviously always, always looking for new engineers within the company, people to come in, particularly on the AI side as well.
Aleks: I can link to your open positions in the show notes, so maybe you’ll find some good people for Indica after this podcast. But let’s talk about your product, right? What? So I know Halo, but Halo is not just one halo. It’s a lot of different parts. Okay, let’s talk about that. Is it? How is your product built and also how can it be used and how is it sold?
Kate: Sure. So we don’t just have sort of software products, we have services. I would want to mention that everything we do is focused on the digital pathology space. As you mentioned, Halo is the the software platform most people are familiar with from Indica Labs as well as Halo API, which is essentially the AI component of Halo.
So Halo is a modular platform. So you said it’s many different parts and so the modules are algorithms that are integrated within the Halo interface and the modules give you sort of all the parameters to do a specific type of analysis. So it could be a pretty generalized module like Multiplex HCI, which is for looking at quantification of biomarker expression in Brightfield slides, or high plex AFL for looking at multiplex fluorescence biomarker expression, or is your fish quantification.
We also have really specific modules for different tissue or cell types. So we’re looking at microbial activation, exons, or accounting or measuring muscle fibers. So the idea is really that labs can pick and choose what algorithms make sense for them, what modules make sense for them based on the types of image analysis they’re doing, and then they can scale according to those needs that also according to their budget.
So if they don’t need something right now, they can always come back and add additional modules later. So that’s generally how it’s licensed. So we can add a certain number of modules. You can also have multiple sets of Halo as Halo as well, but we try not to sort of force people to have everything that they if they don’t need it.
Aleks: So okay, so those modules are algorithms that you can kind of buy out of the box and deploy them. What do you do if you need to tweak them?
Kate: Right? So all of the algorithms are tweetable so they can have each of the different parameters, let’s say like the intensity threshold, they’re all modifiable. You can change how the nuclei are segmented, how the membranes are segmented, and how the cells are phenotype. There are lots of different options and those are all sort of built into the modules, the ability to tweak and optimize the individual settings.
Aleks: So you’re like have the algorithm that does most of the job already there. You don’t have to build it from scratch, but you can adjust it to your datasets.
Kate: That’s right. Exactly. Exactly. So we also have image management. So that’s like for the research space that’s Halo linked in. On the anatomic pathology side, we have a case and image management platform called Callaway Peak. That’s our newest product at Indica. In terms of an actual platform. We also have EMR, Halo, Prostate, the AI algorithms with prostate cancer detection and grading, and then our Halo app platform also has API for integrating other third-party algorithms like the X is the most recent one that we have integrated with.
So that’s from B file. I think that yeah, that’s, that’s sort of it on the, on the product side and in terms of what we offer currently.
Aleks: Is a lot of components. How did you come up with this? What was the path to the creation of Halo like? How did it grow? What was the core thing and why did you go in the direction that you went?
Kate: Right? So actually, as I mentioned earlier, Stephen founded in the Gap, was building initially algorithms for the period image scope, VR. So this is exactly how we got started in terms of the lab visit business.
Aleks: By the way, Image scope viewer at this video, I made a video once on YouTube on how to annotate the image scope. This is one of my top videos after like, I don’t know, three years into my YouTube channel. So yeah, it’s still something in use.
Kate: Yeah, it’s still out there. Anything is so yeah. Then we were integrating algorithms into images though, and at the same time actually, ironically we had integrated with like a slide fast viewer. So when a video was sort of acquired by like that was quite an easy transition in terms of what we learned when we were developing at Indica though.
But anyway. Image So as you said, great for viewing, and great, for annotations, but it didn’t have all the functionality that our research customers needed for image analysis. Things like like being able to store cell-by-cell data, individual cell information, individual objects, information, the spatial information about all the objects that that wasn’t supported.
There was also only one option for tissue classification with the image scope platform and I think we wanted to be able to explore other options, more advanced algorithms for classification, random for deep learning, things like that. So I think all of this led us to create Halo as a new interface for running the algorithms that we’ve previously developed for the image scope and slide path viewer.
So Halo became our main gooey. That was around sort of 2014 and 2013 with Halo was released after that took a few years and we developed Halo Week. I think it was a pretty natural step for us to develop image management. We had a lot of customers who needed a way to sort of store and view and share the data they were collecting in Halo.
These days Halo Link is more of a be run completely separately from Halo and it integrates other third-party image analysis and AI. So it’s really more of a standalone AI in that, but it was a very natural progression for us to add image management to our image analysis and capabilities and then the most recent addition, as I said, was Halo TI, and that was back in 2016.
We were approached by a customer who was using Halo and with Build, really asking us to build a clinical view that they could use to run their companion and prognostic algorithms. And ultimately this was the impetus for us being able to create Halo. I think that’s probably pretty unique in the space that the image management system was created really with image analysis and AI at the heart of it.:
And the question for us at the time was how do we make this easy for our former customers? Our diagnostic lab customers should be able to build algorithms and Halo, and we can then be able to transition those into more of a clinical lab. So Halo IP has all this cool AI technology, the ability to integrate all this technology, but it can also do sort of the routine workflows on diagnosis, secondary console, tumor boards, synoptic reports, all of that kind of stuff.
Aleks: So we have like a full ecosystem, right?
Kate: Yeah, full ecosystem. I think that again, really unique in the space that we have, the sort of full bench to bedside and the discovery phase using Halo and Halo API to developing companion biomarker algorithms and then being able to sort of share that information using Halo link during a research phase and then transition that into more of a clinical environment down the road.
So pretty unique.
Aleks: One, very important question I need to ask about this brother. I said Halo Hello.
Kate: I saw that.
Aleks: One for myself. I saw a pin recently when I was at the STP conference and I met your team at the booth and you guys have those pins with like funny little sayings and one of them, like you say, Halo or Hello Ice and Goodbye or something like that.
Kate: Yeah, yeah.
Aleks: We’ll come up with this name.
Kate: Lots of funny little things. And so Halo Steven came up with that name and I think it was really, as he described it because in the image analysis markets for the cells, there was like a ring around the cell and it looked like a halo around the cell. And that’s initially where it came from. Not to be confused with the video game.
He wasn’t a gamer or anything.
Aleks: I don’t even know about the video games, so I’m hopefully not confusing. So you mentioned you guys also do services. What kind of services do they offer?
Kate: Yeah, and I’m glad you circle back to that because it’s a growing focus of the company and we have a few different services teams like the oldest one, the one that’s been around the longest is our former services team and they’re providing outsourced image analysis services to Army customers, crows that might not have their image analysis capabilities, that kind of thing.
So this team is probably our biggest internal. It’s like having a really big internal customer for all of our products that are using Halo and Halo API for all their image analysis, sharing our results for healing. And so we know if like the pharma services team isn’t very happy about something in the product, then our customers are probably feeling that they.
Aleks: Have like immediately tested at scale in the House. That’s right.
Kate: Exactly. Exactly. They’re our beta customers on things that are sort of coming out. You know, we’re so intimately familiar with the product as well, which is great. So that’s one service. We also have our Eye Diagnostics team. I mentioned Halo Prostate II, which is one of our clinical algorithms that was developed by the A.I. Diagnostics team, along with collaborators.
Kate: They’re also working on one breast, seeing different cancer detection algorithms, and they were initially called collaborations because they’re collaborating with a lot of academics and pharma to really around the world to collect data and trying to build algorithms that are as generalizable as possible on different scanners and through different labs. As we all know, things look very different from one lab to the next, even if they’re using the same reagents and scanners.
Kate: Even so, they’re trying to make everything very generalizable, doing a lot of validation and ultimately regulatory clearances. So and they also work with the pharma to develop ad-based companion diagnostic algorithms to pair up with new therapeutics, for example. So that’s a major focus for us as well. And then last but not least, before I forget to mention our professional services team, they are providing managed hosting services to customers that are using our Halo platforms on Amazon Web services so they help businesses.
Aleks: Separate from pharma services.
Kate: Is separate from pharma services because.
Aleks: US Pharma services do image analysis and they do more like technical support.
Kate: Or I.T. Yeah, they’re all about it. Yeah. Setting up the AWP as infrastructure architecture that is, you know, really most optimized and performant for running all of our Halo platforms and they offer a sort of advantage service for looking after that deployment. So I think this is great for customers who don’t have IT internally or maybe they just want to make sure they have the most performance set up to run our platforms.
Kate: But that’s one of the newest parts of our services. But again, going rapidly as well.
Aleks: When did you start this on the server?
Kate: Only a few months ago. Officially launched a few months ago and we have a few customers, so nice to see it grow.
Aleks: So let’s talk about your customers. Who are your customers? Who is using Halo?
Kate: Right. So I mentioned earlier that our originals imaged the modules and then subsequently the Halo and Halo. I mean, designed for research-focused life science customers. So historically that means that pharma is made up of probably the majority of our customers, followed by other life sciences type customers like academic medical centers, government labs, and Xero.
So those are those are our sort of major life science customers. But I think with the addition of Halo IP and ultimately our air diagnostics products, we’ve rapidly expanded actually into the clinical space, the security, some really large clinical pet tumors in the UK, in the US, some in mainland Europe, and we’re expanding now to other parts of the world.
So it’s a phase is a very exciting time for us as we sort of expand our capabilities on the clinical side and regulatory clearances. And it’s a little like being a startup again 12 years later than this time. So with more focus on the clinical, clinical space.
Aleks: So 12 years in operations and a lot has happened in that imaginal space and your 12 year 12, right, Because your first year, 2012, this is the year when deep learning came to the image analysis scene. Alex Nat, the CNM Convolutional neural network, outperforms other image analysis solutions to classify dogs and cats in natural images.
Aleks: Where are you aware of that? At that time we.
KaOnlye has been aware of it within Indica, but I can be honest that, I wasn’t aware of that point. And I think my main experience.
Probably was.
Kate: What I wanted then. Yes, exactly what was it? However, my main experience with artificial intelligence in digital pathology was GENI, which was the sort of machining that did deep learning tissue classifier that was being offered by a period at the time. And we integrated the random forest algorithm for tissue classification into Halo. Both of these were machine learning or not deep learning.
So this is, you know, pretty early days. Our random forest algorithm was integrated in 2013 when we first released Halo, but then it was a big improvement. But still, there were a lot of limitations as to what these things do. I think that if you asked me, I would have classified myself as an AI skeptic then, at least in terms of what AI or deep learning to do or not deep learning, but what I could do in digital pathology, I hadn’t experienced deep learning at that point.
Aleks: So when did you incorporate AI or deep learning specifically into your software?
Kate: Yeah, right. So this came about around 2017. I think we were pretty early out of the gate in terms of integrating these things into a user-trained image analysis system. But I mean, this came down to Lado, who latched on to AI as an important trend in 2016. So I think you would have known about that certainly in 2012.
But you participated in the Chameleon 17 Challenge, which was basically to design a classifier to identify breast cancer, mets, and lymph nodes and to degrade them and I think if natively we were all really surprised when his injury came in in the sort of top three at the time that the challenge closed is different now because they open the challenge again.
And of course, A.I. develops at exponential speed. So there’s a lot of other options now. But at the time we were the top three and he made a presentation and participated in the initial publication. But I think this was a major if you start.
Aleks: Believing in A.I. then or not.
Kate: I did. I started to believe in an API at that point. We had in the lab, I think, a particular tissue classification challenge that I least personally used to judge, you know, the value of new techniques that we were bringing in and that was bone marrow, lie detection in the kidney. And I think when it came to that, our sort of random forest tissue classifier was truly random in terms of performance.
Aleks: And I’ve experienced that as well in other different software that with uses random forests, I was around them sometimes.
Kate: Yeah, if you do random. But what I showed Ming was how he trained the big network that he used in the Million 17 challenge, but he used this to identify one Marriott lie and not only in which you know, I would, I would have been really happy about just seeing that being trained. But he could also train the same classifier to pick out Glomeruli and all different types of stain.
So, yes, if I see silver, you know, you name it and the same classifier. So I was blown away by that. And I think that’s when I understood how transformative this was going to be, that I was going to be in the pathology space and I was no longer at that point.
Aleks: So what can your module do? What are abilities? I know you guys have the special nucleus detection model that runs there with other algorithms. Heather But what else can you like just do annotations and training for anything?
Kate: Yeah, you can. You can add your annotations and train like a tissue classifier or actually, we have several different tissue classification networks that are integrated into Halo II. Bigger isn’t one of them anymore because it’s now So data is even though it blew me away early on, we’ve got this just again, exponential speed of development really.
But as you mentioned, we also have networks that segment cells. So to help identify nuclei and membranes, membranes are the newest and I’m super excited about that you know, constantly playing with with the membrane segmenter as well as algorithms for actually phenotyping the cells that are being segmented. So all of these things actually can be standalone. So to classify different tissue classes and to quantify the amount of different tissue classes, two segments, nuclei cells, or membranes, they can also be used in conjunction with all of our more traditional hardcoded algorithms.
So for example, Multiplex. I see. Yeah, that’s right. Yeah.
Aleks: Unique I think, or something that is not that yet emphasized or leveraged because you guys have the classical computer vision approaches and computer vision algorithms that you have in those models and how do they interrelate, how do they work together? And can you do everything with Halo API or is that not the best way? Because to me, like from other software that I know I software, you don’t have the other parts.
You train everything from scratch. Maybe you have something pre-trained and here we have a combination. How does that?
Kate: Work? Right? So hey, the way I component, obviously you can train. We have sort of out-of-the-box solutions for nuclear segmentation using AI, and membrane segmentation. So the algorithms we’ve trained ourselves, but you can use Halo AI to train if it’s a different type of cell that just looks a little bit different, different stains that weren’t used in our training set, things like that.
So there’s an option for that. But you have a nuclear segment, your membrane segment where you can use those within the traditional algorithms as well. So again, if you go to the multiplex, I see analysis so you can use a specially trained AI nuclear segmenter to identify the nuclei and the in any membranes do just a much better job at identifying the cell compartments compared to some traditional image analysis.
So that’s kind of where they intersect. You can also use tissue classifiers you built in Halo AI to analyze the different tissue classes, so analyze tumors, separate stroma, or whatever. So that’s sort of where they come together. I think. So the question, the second part of it, do you need what can you do everything in AI, You can do a lot of things in AI.
You can, you know, do tissue classification. What about a different tissue class in segment cells? You can phenotype the cells according to, you know, again, whether they’re tumor cells or lymphocytes, things like that. But, from our perspective, at least, traditional image analysis modules are still needed, not just technically because, you know, just to be honest, not every problem needs AI.
If you’re trying to quantify the intensity of a biomarker. Traditional image analysis is, you know, maybe works pretty well for that job. So yeah, decent job. So at times when you don’t need it, there are also, to be frank in the world, GPUs that sometimes be hard to get a hold of. There are some labs that just because of supply chain issues or even affordability don’t necessarily have or want to be able to run AI.
So I think as a company, at least we try to be perfectly honest with our customers about when A.I. is going to be an asset for them and when it won’t be. And we try to give them options at the end of the day. So most of our algorithms can be run without any AI component currently. So we’re also sort of actively researching ways that we can get our customers to be able to benefit from an AI with, you know, with a more moderate sum of or less expensive hardware requirements.
And, you know, ultimately I think that would be the ideal solution all around.
Aleks: Yeah, I see this trend of modular work, so it can be like you can structure either all an AI-based AI model is deployed within the results of other models or like you guys are doing this AI model deployed within the results of classical computer vision or the other way around. If you define your image analysis problem well and structure your image analysis solution logically, you get better results than trying to throw A.I. at a huge problem and expect it to solve everything.
And like you, I’ll show you so correctly, there is still some design work to have a good image analysis solution, which was the case, you know when I was starting in this space in 2016 and is still the case. There’s no magic pill, unfortunately, we have not invented that yet but maybe in the next ten years. So what I think is going to happen in the next ten years in digital pathology, was a question I wanted to ask you when indicatore in ten.
Then it was like magic, let’s put it. What about the next 12 years?
Kate: Right? Yeah. So what I’m excited about is the coming decade because I think they’re my first decade in this industry and more than a decade actually, probably almost every presentation and or poster that I would see pathology visions or at the digital pathology and AI Congresses were really about the adoption of digital pathology and how you adopt digital pathology.
And maybe because of COVID and or maybe because of all of this sort of innovation that’s happening in the field, I feel like people are now starting to talk about what they’re doing with digital pathology rather than just talking about how they adapt. And I think.
Aleks: That’s a good start.
Kate: And yeah, it’s it’s reassuring to me that, you know, we were sort of moving to this new phase in in the industry. I think it still feels like it’s early days for deep learning. There’s a lot of excitement in the space that I also feel a lot of skepticism and maybe even some resistance to using AI at least, you know, when it comes to sort of the clinic and in patient care.
So I think over the next decade, we’re going to continue to see all of the sort of major hospitals and health systems going completely digital. We’re seeing that in the UK and now in parts of the US as well. But the UK has had a lot of, you know, major initiatives to get these NHS digitized and I think that’s going to spill over into other countries, other health systems, and in hospitals.
So I think that sort of digitization ultimately of pathology is going to, well, obviously exponentially increase just the amount of digital data that we have. And I think that will fuel that sort of AI revolution, the sort of next step in pathology. But I guess all myself, I wouldn’t sort of hedge my bet that that’s going to happen in the next five years.
We might be realistically saying 5 to 10 years down the road when that becomes the sort of next phase of digital pathology.
Aleks: It’s funny like you cannot predict because some things happen so much in technology. Like recently I was in New York at the DPI Air Conference, which you guys in Europe have in London in December. And like several people said, Oh, in 2010 I was talking to at the stage and saying that everything everywhere is going to be digital pathology in five years, right?
And you know, like whatever, 20 years, ten years later, and it did not happen. But some things happened so fast, I don’t know, like I’m mind blown with the large language models.
Kate: And so, yeah, anyway.
Aleks: You cannot predict some things are going to be slower, some are going to be faster, but exciting times. But before we wrap you up, I need to ask you what didn’t go so well. Tell me about some failures that you had on the way to getting where you are right now.
Kate: Right. Okay. So I think this is this is a more difficult one. Everybody likes to talk about theirs.
Aleks: Synopsis of this, but.
Kate: Yeah, and I have a lot of personal failures. Then I’ll take that one specifically at the right. So a few years ago, this is actually before Hayley, I was probably about 20 teen years, something like that. We embarked on developing a platform called Strata. Strata was designed for our pharma customers who were running their image analysis algorithms that had all these sorts of outputs from the image analysis.
They also all this patient data, and maybe they had outfits from other types of analysis they were performing. But anyway, they were trying to combine all of this data and informatics, right? to use it to sort of predict treatment, responses, or prognosis. So we’re trying to pick out what morphological features of the tissue in combination with sort of biomarker or expression and maybe these features of the patient, which once you put together that you could ultimately maybe stratify us for the thing from customers that might respond versus not non-responders for a particular treatment.
So we knew this is what our sort of former customers were ultimately trying to achieve with their biomarker programs out there. And I think it felt like a natural thing for us to develop since we were already generating all of this data ourselves. So, we were I was super pumped to go out and talk to my customers about Strata to sort of give a demo of the platform.
And I think one of our customers was also really excited about it. But ultimately what we learned pretty quickly is that the customers that we were talking to of Halo were the ones that were doing the informatics. So the bias that they were, you know, analyzing the tissues and then throwing the data with sense to the informatics teams who were doing all the number crunching.
And we tried to get ourselves in front of those teams to show strata. But I think they were sort of much less interested in an easy-to-use platform.
Aleks: I can come to anything they want and they don’t need a tool for that. And we did hear about the images like the image analysis, people, they don’t care. They want this to be good. Then the data is extracted. I don’t think the data people cared that much about where it came from.
Kate: Yeah, it’s.
Aleks: I don’t care that it’s that it’s accurate. Right. But once it’s accurate they they think can work with it.
Kate: So yeah I’m just thinking exactly. So, it was it was definitely, you know the product we didn’t design for the customer target base we thought we were that we weren’t. And I think probably the takeaway lesson for us there was to get in front of the customers very early and often to show them what we were developing and make sure it was whatever they would be using in the lab.
We did that with Halo and all of our other platforms that some reason we waited for a while to sort of get in front of customers and found out that we were in the wrong sort of target market there, but easy to accidentally be living in an echo chamber where you’re excited about your product without actually beyond showing it.
But a colleague of mine at indicate that was around at the same time. Australia, she sent me a little screenshot of a strata PowerPoint slide that she found on SharePoint and sent it to me with this caption Rip Strata. So I think it was a heavy heart for all of us when we had to retire strata early in terms of development.
But I guess I still have a super hope that it’s going to be dragged out of retirement someday. So, you know.
Aleks: I’m interested because the concept seems so logical because everybody’s talking about let’s let’s change the silos and drag development. Let’s make people collaborate more. And you need tools for that. I do not know how to develop those tools.
Kate: But what’s the solution to help, right?
Aleks: Yeah. Okay. Thank you so much. Anything else that you would like to share? Anything that I forgot to ask you, I.
Kate: Can’t think of anything like. So I think we’ve covered a lot of ground today in terms of customers and products and history being a lot of nostalgia and or so it’s been a pleasure anyway, talking to you today about it.
I’m so happy that we met again. We were supposed to do it when Indica turned ten. I will link to your open positions and if at the time of publishing this podcast, it’s a different one, that’s okay. People can go and check and you have a wonderful day.
Thank you, Alex.
Aleks: Thank you so much for staying busy. It means you are a real digital pathology trailblazer and you might have some need for more in-depth courses about digital pathology, and different aspects of digital pathology. So I have a gift for you. We recently launched a membership site Digital Pathology Club at the Digital Pathology Place, and I would love to offer you a one-week free trial with access to all of our courses.
So go ahead, check the link below and grab your free trial and I talk to you in the next episode.
