Aleks: People in Tasker are kind of waking up to the fact that data is the new currency, and they’re looking for data sources and what they can do with this data. Are you noticing this as well?
Peter: Yeah, we’re definitely hearing a lot of people talk about, just at some conferences recently, the CAP Data Summit and the up here, there’s a lot of value in the data that pathology generates. Not just not to report data, but all sorts of, you know, molecular and other data. And that, you know, it’s super, super useful in precision medicine. It’s very useful for like matching patients to trials. It’s very useful for research. And they’re starting to acknowledge that and also kind of want to use the value of that data.
To elevate the department a little bit too, which I think makes a lot of sense. So we want to help people unlock that. We honestly have customers who’ve used our products for years and don’t really know they’re sitting on the actual structured data aspect. So we like to have those conversations and having people come around and acknowledge, all this stuff to record. those actually, is there like a database we can access? And we’re like, absolutely, [00:01:00] let’s show you how to do that. So a lot of times people are sort of have it and they’re not using it yet, and that’s not hard to enable. And then, you know, if they’re not. Doing structural reporting at all, all they have to do is start. But a lot of times they’re already doing it and we just need to help them like start to make use of that data.
Aleks: AI has this promise that now you can write whatever as a medical report. You don’t have to be so structured and it’s going to find all the information anyway. I’m here at mTuitive. We’re at PathVision with Peter O’Toole. And there is a little bit of a different side to this story and mTuitive is actually an expert in structured data. Welcome, Peter. How are you today?
Peter: I’m good. I’m good. Thanks for having me.
Aleks: How’s the conference so far?
Peter: It’s great. I love Pathology Visions. It’s always a great place for me to learn a lot about what’s going on out there in AI and digital pathology. So it’s my favorite conference.
Aleks: I’m going to ask you some questions that will probably challenge that [00:01:00] common knowledge of what AI is capable of.
Peter: Sure.
Aleks: But I want to explore that. But before we dive into, let’s start with you. What’s your background? What’s your role at mTuitive and a little bit about mTuitive.
Peter: Sure. I’m the president and chief software architect at mTuitive. I actually started at the company when it was founded. It was founded by three people, John Murphy, Mark Law, and John Everson with a set of skills in the software business. I joined shortly after and my father got involved as a pathologist, a recently retired pathologist, and Colin Murphy, who’s our current CEO. We all kind of started a nucleus of people back in 2003 or 2004.
And we started implementing, we had some ideas about decision trees and structured reporting and things like that and pathology and got started when synoptic reporting became kind of a new idea or a more popular idea and [00:03:00] that led us to where we are today and evolved over time. But one thing we did from the very beginning was make sure that synoptic reports, which had a lot of benefits by themselves, also as a byproduct of what they do create structured data that can be used for measuring outcomes, for measuring quality, for measuring whatever it is that you want to improve. And whereas the idea of synoptic reporting always had a lot of benefit at the individual patient level, like let’s say the completeness of the report, the consistency and standardization of reports, making sure that oncologists and surgeons are getting back, pathology reports that are speaking the same language and putting things in the same places. And including all the right information, that next level of structuring the data behind the scenes and not just making it look synoptic is sort of something that was different about our approach from the beginning and still to this day is, yeah.
Aleks: So when you talk to people who don’t know anything [00:04:00] about this space, like at dinner, what do you tell them you do?
Peter: You know, yeah, I try not to put them to sleep. I mean, I tell people that what we do is really about changing the traditional process where medical reports were created with essentially a word processor or a dictation process. We sort of re-engineer the process to be structured data first, but also not just structured data, but essentially it’s about clear and consistent communication between specialties and handoffs between diagnostic procedure, diagnostic elements over to treatment modalities, but also coming out of surgery and passing data onto pathology and onto oncology, and kind of helping to define very clearly what are the really important things each specialist needs from the previous one and make sure that that’s in there.
Aleks: I think that’s a key I want to highlight that the important things that each specialist needs from an hour.
Peter: Right, right.
Aleks: Each medical specialist [00:05:00] because like especially in the US the specialties are very narrow and there are specific information that people want from each other and they want to get it from the report fast to be able to provide fast care.
Peter: Yeah.
Aleks: So this is how you’re going to structure.
Peter: Like if you look at the CAP protocols as one example of structured and synoptic reporting, there are 110 of those for different types of cancers, right? So there’s very different prognostic indicators that every different oncologic type specialist is going to want to get from the pathologist, right? So staying on top of that and making sure it’s always up to date with the latest standards is a big benefit of structuring things and I think enables…
If you remember, like years ago, there was this book, The Checklist Manifesto.
Aleks: Yes. The Philo and I have read it.
Peter: And you know, it’s a great book. And still to this day, that checklist mentality is very helpful, like in any specialty, in ensuring completeness and consistency. Right. So. I mean, when you combine that, [00:06:00] goes hand in hand with, since we’re being so consistent with structured and standardized, why don’t we actually compile this data and do something with it after the fact, right? And when you, like in any industry outside of healthcare, if you have like a really great, clean database of standardized data, there’s all sorts of great things you can do with it and it’s no different.
Aleks: That’s like the greatest bottlenecks or at least like that was the greatest bottlenecks cited before the large language models. Let’s unpack that actually because you know a couple of years ago the bottleneck was like it’s not structured, it’s text, it’s like free flow text of the pathologist, every pathologist, even though there are checklists, there are templates but now we have large language models and they can sort through that.
Can they actually? How well can they do it? And how does this tie into what they can do when the data is structured in a specific way?
Peter: Sure, I think they can do [00:07:00] a lot, right? There are a lot of papers being published. You can use, you know, previously NLP and now like augmented with LLM techniques, right? You can do pretty well retrospectively, you know, pulling data out of a report and structuring it after the fact. And that can be really useful for like speeding up, you know, abstraction, you know, cancer registry, use cases like that. We have always really been focused on…
clinically actionable scenarios. So, like if you sign out a pathology report and you want that data to immediately populate a staging form for an oncologist, it needs to be 100% ready to go at that moment and not really go through a process like a retrospective curation and sort of sign-off process on the data quality, right? So if you need the data to be ready to go, like immediately, without any sort of quality checks beyond the pathologist or the know, whichever doctor signed this case out, essentially generated and curated their own data in real time, signed it out [00:08:00] and it can be used right away. You know, it’s considered high quality and like what we call clinically actionable data. I think, you know, certainly there’s some scenarios where LLMs can help with that, and we’re not saying they can’t. And downstream, like in a retrospective scenario, when all you have is text, that’s great. You know, you can get some data out of that. And we’re really doing like, we’re doing a lot of research in…spending some time and money on how LLMs can help even with the input of the synoptic report. So you may be synthesizing information from other sources besides just what you’re seeing under the microscope. So can we get something from the chart? Can we go pull that information about whether the patient had neoadjuvant therapy and bring it forward to you? Because that’s not on the slide, right? So I think it’s a hand-in-hand approach, right? But we still think the ultimate goal of having a highly structured database is, is what people are trying to get to. If we can spend maybe more of our, you know, all the brain power going into using LLMs to extract data, if the data was already there, [00:09:00] they’d be putting that brain power toward like using the data.
Aleks: What I wanted to use this analogy and tell me what you think about it because when you have digital pathology, right? A scanner is going to scan any slide. But if you have a not so perfect slide, the image that you’re going to generate is not that perfect. You have to read through artifacts, ask for a rescan, ask for rework. If you are working with digital in mind, then you focus and it doesn’t cost you any more work. It’s not that the technicians in the lab need to like spend half an hour more to prepare various slides. No, they just know it’s going to be scanning, it has to have specific quality. And it’s kind of an analogy to, okay, we now have this AI that can be leveraged and the choice is, okay, are you going to leverage it to clean up stuff that was not clean at the beginning? [00:10:00] Or are you going to leverage it to extract more to be more efficient. Like I know like my experience from the consumer AI, right? Even when you I thought that retrieval augmented generation, this process of AI just going to the sources is going to solve all the problems on the planet, until I started using it. And it still goes off track. So let me know what you think about it.
Peter: Well, I think it’s a good analogy, the whole slide imaging analogy you just made, with essentially an AI can’t sort of structure data that’s not there. So first of all, I think that process of synoptic reporting has been really beneficial to ensure completeness. So again, if the data is not in the report, if it was just left out, you can’t structure So that’s one. Right, so that’s one reason to sort of start with a structured process, which oftentimes is an easier way to enter data, checking a few boxes [00:11:00] and so forth. So, what was the second part of your question? I don’t remember now.
Aleks: So the first was how does and you already mentioned that like, okay, what do you want the brain power the power of the Ai to be used for right cleaning up or averaging it for, for something else for like more impact, right? Not only speed right because and I already said that about the consumer, I like it takes time to interact with these systems…
Peter: Yeah,
Aleks: …you don’t have this time in the clinic.
Peter: Yeah,
Aleks: then you mentioned this concept of for our, for clinical readiness of data, right? You don’t have the time to, I mean, you can have a pre-written report that’s going to be faster than a non-pre-written, but still, you don’t have time to tweak it. You have to tweak it as little as possible.
Peter: So I mean, we like to [00:12:00] pull information from as many sources to save the pathologist’s time as possible. It’s a lot easier if those start out structured. So for example, like genomics reports coming back from the lab. It’s tough to see something that’s so high tech when you think about what they’re doing to sequence a whole genome or whatever, or next gen sequencing, and then send it out as a PDF that’s 50 pages long.
So a pathologist may be reading through that and pulling out relevant bits and pasting it into their report. We’ve worked with companies like Keras Life Sciences to get structured data from them directly into the synoptic report. So I think just by upfront structuring, you can avoid a lot of that and save time. And again, we probably have companies parsing those PDFs and trying to do that. When if we just started structured instead, we could be using that. You know, energy going into AI to maybe figure out, know, if we have all this, if we’re sitting in all this structured pathology data and it’s correlated with images, let’s use that as a training data set. [00:13:00]
Aleks: So I want to take a step back because in general, what is this reporting workflow? And then to compare and contrast the workflow before AI and the workflow with AI, either already implemented or potential partnerships that you are thinking of or where it could give additional leverage.
Peter: Yeah, yeah, sure. I guess it depends on which type of AI we’re talking about. But at this conference, with all the cool, I love being here and seeing all the cool things people are doing with computer vision, image analysis type stuff. There’s no reason if you just used a computer to figure out someone’s ER, PR status, that you should then have to resort to pencil and paper to get that information into your report. So we’re talking, we’re working with a lot of these companies directly. So that their data can immediately feed the reports, right? So that’s one way of just, the workflow for the pathologist is changing a lot, and we don’t want their reporting to go backwards and be more difficult, because there are all these additional [00:14:00] systems that aren’t in their LIS, or they may not be plugged into their LIS or their IMS, but we can still work with those companies to directly pull that data into the report, right? And also working with companies like Path Presenter, who hopefully try to provide a middle layer there to make lives a little easier, but.
Even if there isn’t like an IMS that’s capable of pulling in AI results, we can work with those vendors directly. So that is changing their workflow already today. And in a simple way, we have people asking us all the time, can’t we just take the information coming out of this model and plop it in here? And we’re like, sure, that’s a great way to do it. so, I mean, in a sense, that’s going to make it a lot easier for them having data coming straight off the digital pathology viewing screen into their report without having to do any manual transcription, copy paste, stuff like that. But in addition, you know, again, like this is a little bit more future for us, but we’re doing a lot of research and prototyping right now [00:15:00] into how we can take, you know, the other interesting piece of AI, the LLMs to help the reporting process as well. And again, a lot of that I think is processing data, but we’re also doing a lot with, you know, how do we, how do we analyze it after the fact? You know, what’s a natural language querying mechanism?
Even when we’re sitting on the structured data, we created an MCP tool that we can use in an LLM-based chat to query our database. Like model context protocol. it’s sort of what allows like ChatGPT or other tools to have other APIs they can reach out to to solve specific tasks. So if you want to look up some pathology reports based on, let’s say like, CAP data, you can connect to our tool to make that query into mTuitive from an LLM, which could be like a user interface we’ve provided or even that a different company has provided.
Aleks: So question, what about spoken words? So [00:16:00] there are AI rollouts for different medical specialties and dictation was part of pathology workflow already first time. Now you have automatic transcription and AI working to power that. How does this part of the workflow fit into what you’ve used?
Peter: At mTuitive, our products have always worked in concert with voice recognition when that’s what people want to use. So, for example, maybe you’re doing a structured report, but there’s a pretty large part of that that you want to maybe do free text. That’s totally understandable. And so you might just use Dragon or any other commercially available tool to dictate into that free text box. also,
We even have like a speech-enabled version of our user interface that allows, yeah, that allows, and to be honest, it’s not used that much, right? Because I think at end of the day people say, oh sorry, I can just sort of, [00:17:00] you know, I’m okay with checking these boxes for this part. In fact, that’s probably faster than saying all this stuff. But when we’re getting into an unstructured part of the report, and we do a lot also with surgeons’ reports, so operative reports for cancer surgery. Started and they do a lot of narration, maybe more than pathology. And so again, there, it’s a bit of a hybrid approach. But I mean, the two things work well together. luckily too, most of the voice enabling tools have always been pretty easy to work with and to expose your app to those in a way that you can call out, click a button or whatever. So it’s definitely not mutually exclusive.
Aleks: But you say it’s not used that much.
Peter: No.
Aleks: I would expect it to be like the ultimate hack to input tag.
Yeah. I think it’s, you know, we’ve been doing a lot of user research lately among like all of our users in different geographies, different specialties in the UK, US, Canada, surgery, pathology. And I mean, it’s definitely a personal [00:18:00] preference issue. It’s not like 100 % of people like one thing or another. There’s definitely a mix of approaches. It’s interesting to see that a lot of people using dictation are using it to say like, use my standard appendectomy template or something like that.
Okay, so, and then essentially you’re put, so I mean you’re using it but you’re not dictating that whole thing. But yeah, mean all those things….
Aleks: That look like with like an agentic workflow with different types of data that are there.
Peter: Yeah, yeah, I mean, and that’s really interesting too, and I know, I mean, our team’s been doing a lot of research with the semantic kernel from Microsoft and like how we can orchestrate different AI services. So I mean, all of that plays a role. For us, just the heart of it is still how do we get a very structured report under the hood and a very consistent and like readable and standardized report outwardly, because that is still what’s better for the treating physician to process, right? Is sort of that synoptic style thing and that data underneath to still be useful. How we get there, [00:19:00] yeah, it’s going to be a combination of all of these techniques, for sure.
Aleks: It’s definitely, exciting…
Peter: I think another thing is just design in general, like user interface design, mean, and design of products to make them as useful as possible was kind of a core principle for us at the beginning, and one we’re really getting back to lately and investing a lot more in. I think you might hear this a lot lately in general, not just in healthcare, but, you know, that sure, AI is really smart and it’s doing a lot of great stuff for us. What will differentiate companies in the future, especially software companies, a lot of it is design. You know, that human element that is making the product easier to use but also pleasing to use and thinking through the different ways to make that workflow as smooth as possible. That’s still something that we focus a lot on and I think that’s what the, as we’re blending in different AI capabilities, keeping that human element is really important too.
You know?
Aleks: What’s going to differentiate is going to be, is it user friendly? [00:20:00] So there’s always like in new tech, doesn’t matter if it’s digital pathology, Ai, whatever, there’s always resistance to change. And I think you encountered this resistance even more in specialty professionals. Pathologists being one of the specialties, right? Surgeons being another specialty. have like…
Okay. Very highly trained individuals and then you’re trying to tell them, hey, you can do what you already trained so long for, actually better with this tool. How do they respond? And I’m interested in like, okay, they had nothing like that, if that’s an option even, and now you’re entering and trying to help them do their job better. How do they react?
What are the questions that you get? What are the pushbacks? And like, why would you still recommend?
Peter: Yeah, mean, think when we got started with snoptic reporting, [00:21:00] was a huge challenge.
Aleks: What was it like before? Just whatever they wanted to write? Dictation, free film, I’m seeing these cells and…
Peter: Yeah, yeah, just an essay. Scott Campbell, who’s a big influence in this world and with SnowMed and stuff. Like, I don’t want to say this without attributing it to him, but he calls it pathology poetry. So…
Aleks: I am very much familiar with the veterinary side of things.
Peter: Right, right.
Aleks: Because we do not have the requirements to structure it in a way I knew it’s useful.
Peter: Yeah, and there’s a lot of times where that expression is needed. like our, what’s that? I like, he said that at a couple of conferences lately and I’m like, I’m gonna use that. But we’ll give him credit.
Aleks: And you know who can, the only people who can read this poetry are other pathologists, right? And the report is not for other pathologists; the report is for other treating physicians, [00:22:00] maybe for the patients, and I think we are guilty of writing for ourselves.
And I think there was already an undercard of people who thought, let’s have some standardization for specific things like breast cancer and other areas. And I know MSK had protocols out there, maybe on paper, maybe on the early web before CAP did, and some other groups as well. And people may have had their own internal standards, like, hey, for these cases, let’s make sure we include this information. I think when it became more formalized through CAP and other groups, that helped, but also having companies like us go out there and say, actually, this is also a good idea because once we follow these formats, we can actually do a lot. Like, we’ll stage the tumor for you. We’ll calculate the grading. We’ll do a lot of automation and sort of decision support, and try to make life easier. So I think I didn’t really get into that earlier, but a lot of the…
Aleks: Let’s unpack that. Once you have the data in a specific format [00:23:00] in a specific way, you can perform different… Without like going into the AI black box and having to explain it.
Like normal stuff.
Peter: I was just listening to this podcast. I love the NEJM AI podcast. That’s probably a lot of your listeners listen to it. And they had Karen Deep Singh, I think is his name at UCSD, talking about their implementations of different types of AI. And he said, know, LLMs are great at certain things, and not great at other things. So he said, using it to help him process a bunch of literature on maybe new treatments and how they’re applicable for different types of patients, yes, great use case. He said, people in his institution were using it for a risk calculator. He said, it’s not good at that, we shouldn’t do that.
So things like those more mathematical formulas or logical formulas, we’ve been doing that for a long time because doing reporting [00:24:00] in a structured way enables you to make computations in real time that are safe and very deterministic. And they help save the pathologist a lot of time, right? Or other users too, not just pathologists. But there are things that, if we can tell you it’s a PT2A based on the size of the tumor, we’re in the depth of invasion and all those things, by the time you get to that part of the report, that’s already filled out.
And so there’s a lot of that sort of thing that honestly has been a part of the process really forever. And I think that helped ease that resistance to change. Because it’s like, whoa, I don’t want to do a new thing. But they start doing it and say, there’s sort of some stuff in here for me that actually does make my life easier. And even to this day, mean, we talk to pathologists all the time who say, like the guidance that these protocols provide, especially if I’m a little outside of my subspecialty or I’m a generalist. This helps me keep up to date on the very latest, most important indicators to include. You know, include for a different type of cancer that I don’t see very often. So I think there is, there was initially a lot of resistance to change as with anything new. [00:25:00] I think we’ve been beyond that pathology for quite some time. I say that, but in the US and Canada, I think we’re sort of beyond it. We do a lot in the UK, still a lot of more resistance there. So it’s a little cultural.
Aleks: It’s so funny because they have like this national initiative to push everything to digital. Yeah, but for more for the imaging for the slides. Yeah, and then there’s this whole other part of work which is…
Well that initiative like NPIC and specifically which I think they might even be here at the conference they did use us for structure reporting on that project and so not always resistance just I mean that it’s just a little bit slower going in terms of adoption Yeah at a national level, but in those pockets where there’s like a lot of innovation Yeah, they’re they’re doing it to get out with a different set of standards.
Aleks: What’s the next step?
Peter: Yeah, I mean for us it’s really …you see its…
Aleks: Like any technologies that like light you up.
Peter: Yeah, yeah, mean it’s really continuing to invest in like how to make the process for [00:26:00] the physician as easy as possible, but also what else can we bring to the table for them, right? So like if that’s decision support, if that’s trying to sift through the massive amounts of precision treatments that may be applicable to certain diagnoses or figuring out what are the most important data points to include for this type of case because of those emerging…
treatments that are all based on very precise profiles, right? For us, like, I think it’s, and how can we use LLMs in a way that leverages all this data? So, one thing is, we’ve built, you know, sort of natural language query UI into the data collected by all these protocols and being able to ask you questions, you know, how many triple negative cases do I have for breast cancer right now? How many, you know, and seeing how well we can tweak the technology to help people sort of…
We build a lot of analytical outward-facing reports in our products [00:27:00] based on the data that’s collected that are really helpful, where we know.
You know, labs or cancer centers need this report. It’s not a question. But what about all the stuff we don’t know? We’re thinking like, let’s give them a box to type into and see what questions they ask. See how well the LLM can do to give them an answer. That’s something we’re experimenting with right now, showing it to our users, testing that a little bit. And we think that is showing a lot of promise. So, I mean, we’re excited about kind of all the same stuff as everybody else, but trying to really focus it on making that experience really worth while.
It’s not just let’s do stop to reporting because we have to. Let’s do it because it’s making our lives better, helping us make a better decision, helping us collect data for our own benefit, but also downstream, actually automate some processes into oncology and other areas. And also, I guess I should bring up too though, outside of cancer, we talk a lot about cancer care with this, but a lot of pathologists are saying, but we want to do this [00:28:00] for other types of cases too. For liver disease or other skin benign stuffs, and we’ve always done a fair amount of customization outside of the cancer box, but I think there’s a lot more of that coming. So we’re seeing people voluntarily start reporting outside of that requirement.
Aleks: I think this plays into a bigger, because now with AI, the leverage is everywhere. And even though it’s a little bit more permissive, the error combines, the air compounds. And it’s not like, okay, you can still extract stuff out of garbage kind of data, a lot more than before, but if you want to take it to the next level, like in different modalities, you mentioned sequencing reports, surgical reports, pathology obviously, and then all the way up to information that gets to patients. Like a patient [00:29:00] has trouble, understanding of pathology reports, they probably are going to paste it intoChatGPT and ask what that is. Most of the time they will get an answer, but why do they need to do it? Like, why? That is the information about their health. They should not have to decipher. So I think like this whole, I don’t know, I don’t want to call it data hygiene, but it is kind of like data hygiene to have good structured data at every level. And another thing I wanted to tie it into is the multi-modal diagnostics where you have data coming from different sources. to have data that is good is going to prevent us from clouding the material you’re working with. And AI is fantastic, but it has its limitations as well. Anybody who has worked even with the commercial version, knows that you have to kind of police it, and you have to police [00:30:00] it less if you have a template, if you have a structure, if you can send it like when this, dictating, give me this and that template, right? So I think the combination of what you guys are doing with the new technology is going to improve patient care and improve the life of a healthcare provider.
Peter: That’s the goal. It’s definitely not either or. How can we augment what we’re already doing with all this new technology and vice versa? You know what I mean? It’s great to show off just how much you can do on free text with this stuff, but what can it actually do when it’s given really good data is even better, right? And also that’s why we’re working in concert with a lot of the companies here, right? Because we all have different areas of expertise, but together, like stitching these things together.
Stitching the workflows together but also the data and the images and everything else, you know, it’s very powerful. So it’s important we like all kind of work together to create an ecosystem that, you know, benefits [00:31:00] the pathologists or, you know, in the patients really in the end, but benefits the practice of pathology in a way that just one company can’t do when they’re or one technology can’t do when it’s on.
Aleks: Definitely, people in Talscar are discovering or kind of waking up to the fact that data is the new currency and they’re looking for data sources and what they can do with this data. Are you noticing this as well?
Peter: Yeah, we’re definitely hearing a lot of people talk about, just had some conferences recently, the CAP Data Summit, and here, there’s a lot of value in the data that pathology generates, not just synoptic report data, but all sorts of molecular and other data, and that it’s super, super useful in precision medicine, it’s very useful for matching patients to trials, it’s very useful for research, and they’re starting to acknowledge that and also kind of want to use the value of that data. To elevate the department a little bit too, which I think makes a lot of sense. So we want to help people [00:32:00] unlock that. We honestly have customers who’ve used our products for years and don’t really know they’re sitting on the actual structured data aspect. So we like to have those conversations and having people come around and acknowledge, all this stuff to record. Those actually, is there like a database we can access? And we’re like, absolutely, let’s show you how to do that. So a lot of times people are sort of have it and they’re not using it yet and that’s not hard to enable.
And then if they’re not doing structural reporting at all, all they have to do is start. But a lot of times they’re already doing it and we just need to help them start to make use of that data.
Aleks: I think there was this, focus on that pathology data is the image and there were so many hurdles in accessing, there still are. You have search models that can search by embeddings, but that’s a lot heavier lift than leveraging the text data. And now imagine matching the text data to the image data in [00:33:00] a structured way that also have a structured way of analyzing the images. So I think that’s a power that people may be were not aware of.
Peter: The structure reports are super high quality labels for images if they’re associated with images.
Aleks: Yes, another bottleneck, it’s always like, we don’t have labels, we don’t have annotations, the label quality is like a three-page report, three texts, like how are you going to convert this label to another label? That’s another super cool advantage of this in the time of trying to leverage data as a currency.
Peter: We can point you to the specific fields in the structured report that describe the aspects of the image you’re interested in training, right? So yeah, it’s pretty cool.
Aleks: You see, definitely something like a hidden gem. A hidden gem in a hidden file and an incentive to actually do it.
Peter: Yeah, [00:34:00] for sure. And if you’re looking for like specific cases, specific tissues, specific patients that match some very precise intricate profile, like that’s instantaneous when all the data’s been structured like that up front. So yeah, and there’s a lot of obvious ways that people can use that to build value for this, for their institution, for themselves, for the patient. And yeah, it enables a lot.
Aleks: Amazing. Let’s see what we’re working with.
Peter: So let me go ahead and just create a new, you know, invasive carcinoma resection protocol. As you can see, it’s a pretty simple process, you know, to check the boxes as you go through. There are certain things like grading that we can do a simple calculation for them. So getting back to what we talked about with structured data being an enabler as you go through, I could continue filling this out that as I’m filling this out, down below it’s calculated the PT [00:35:00] based on the information we have so far.
If I add more information that may change.
Aleks: I also said you had a bunch of different templates.
Peter: Absolutely, yeah. Things, you know, for other specialties. We actually have structured reports for mental health, which is a really big area at one of our customers, which doesn’t sound like it would lend itself well to structured data, but it really does. And here’s like, you know, compliance with an accreditation program from the American College of Surgeons for breast cancer, right? So that’s in addition to like what you might see in the CAP protocol. So this is sort of completing the protocol, you know, in the workflow of the pathologist they normally would.
Aleks: So how do they look at the slide and they go and fill all the…
Peter: Some do it after and some do it during. And we’ve been doing a lot of interviews with users to find out which one they like better. I mean, or to find out what they do currently. So some do it simultaneously with the image, either on the microscope or up on the screen. And some do it sort of literally after they’ve looked at the slides. And so once [00:36:00] the reports have been done in this way, you’ve generated your synoptic report, you’ve collected data, then it would land into the mTuitive insight database and so for example here’s a report every lab has to run every year for their accreditation and they they pull cases and manually compile a lot of this so we created a report where you just click a button and you can get all this right away.
Aleks: You mentioned matching patients to clinical trials, so definitely this information could trigger. And I’m seeing this as a trend. Because clinical trials are notoriously challenging for being run in general. So any technology that can help with that is fantastic. Thank you so much for showing.
So we’re at Path Visions right now. So if you are here, the booth is 406. This video maybe out when the conference is already over but if anybody wants to learn more about mTuitive where do they find you?
Peter: Mtuitive.com, that’s the easiest thing to do. Yeah, look me up on LinkedIn or whatever.
Aleks: Fantastic, I’m gonna link that in the show notes. Thank you so much for joining me and I hope you have a great rest of your day.
Peter: You too, thank you Dr. Aleks
Aleks: Enjoy the conference.
Peter: You too.
