Name: What’s New, What’s Next at Proscia for Life Sciences
Uploaded: 2025-03-20T16:16:10.335Z
Duration: 1 h 8 s
Description: What’s New, What’s Next at Proscia for Life Sciences

Transcript for "What’s New, What’s Next at Proscia for Life Sciences": Today's webinar on what's new, what's next at Prosha for life sciences. Thank you so much for being here. I'm Ashley Faber, director of product marketing at Prosha, and it is my pleasure to welcome you and to moderate today's webinar. We have a great presentation and discussion for you today, and we're going to get started very shortly. But first of all, I'd like to introduce our speakers. We have none other than David West, Prosha's CEO and cofounder. Hello, David. Hi. We also have Adani Hamid, our product leader for life sciences. Hello, Adani. Everyone. Alright. So let's jump into an overview of what we'll cover today. So first, David will walk through Proche's vision for the future of digital pathology in life sciences. He'll explain how our strategic priorities align with industry trends and how we're evolving our solutions to help you capitalize on these opportunities. Second, Adani and I will dive into what's new in Concentric LS with both an overview and a live demonstration. This will give you a firsthand look at our latest features and capabilities designed specifically for r and d. Then David will provide an overview of our multimodal pathology real world data offerings, and Adani will share a quick sneak peek into immediate Concentric LS platform improvements that are right around the corner. And finally, we'll open the floor to questions. So please enter your questions in the q and a box, throughout the presentation, and we'll try to get to all of them at the end. So with that, I'll hand things over to David. Thank you, Ashley. Thanks for all taking the the time to be with us here today. One of the best parts of my job is getting to work with the incredible scientists and pathologists and technologists that, you know, give me and give us at Prosha a front row seat to the trends in technology and medicine. I think there's never been a more exciting time to be in life sciences. Advancements in both the biology and the technology are driving drug discovery and development and all coming together, at this moment. Next generation therapies, require a deep understanding of disease at a molecular, cellular, and tissue level, all the way from discovery, through development, through commercial deployment. Therapeutic pipelines are colored increasingly by sophisticated precision medicine strategies. At the same time, the world is undergoing this revolution in the computational domain. In pathology, these whole side images contains billions of pixels, gigabytes of information per image. At the same time, advances in AI just over the last twelve, eighteen months are opening up the aperture on the possibilities to generate insights from that data. And life sciences organizations have been compounding data for many years. You know, we've certainly seen this, in pathology for many of you who have been using Concentric for many years. And now digitization and routine clinical practice has triggered this explosion in new high value, this new high value data modality in the form of whole side images and then these rich multimodal datasets that they enable when combined with other omics. These four trends, I think are elevating pathology from just this downstream diagnostic tool to a central driver of therapeutic innovation. Applying AI and digital pathology to drug development programs, provides a tangible strategic opportunity across the entire discovery through development life cycle. Digital pathology and AI is revolutionizing biomarkers and companion diagnostics. We can now identify, response to patients with precision enabling indications for patient populations that weren't previously viable or targeting subgroups, that would otherwise be missed with conventional approaches. Diagnostic, prognostic, and predictive AI models are increasing the probability of success by enhancing decision making at every stage of of r and d. Quantitative insights, you know, help give life sciences companies greater insights to go no go decisions from a discovery to preclinical to beyond, potentially saving millions of dollars in development costs and and, of course, precious time. Digital pathology and AI, are transforming clinical trial design. With more precise patient stratification and quantitative endpoints, you can achieve greater or faster, enrollment, potentially with smaller cohorts or getting stronger, more reproducible evidence of of efficacy, ultimately accelerating your path to approval. And finally, we see the power of real world evidence and post market surveillance creating continuous learning loops from patient outcomes and helping you expand indications, identify new biomarkers, and demonstrate long term value. Let me share a couple of powerful examples how computational pathology is already transforming drug development in areas that may be relevant to, many of you listening in today. Advanced therapies like ADCs are you know, they selectively, attack cancer cells while sparing healthy tissue. I mean, the potential here is massive. The top three ADCs are expected to to generate $17,000,000,000 in global revenue by 2028, but they demand ultra precise quantification and target expression, to determine which patients will benefit. And this is where AI comes in. It, you know, shows the promise to measure expression with far greater precision and reproducibility than traditional, manual IHC IHC based, scoring methods identifying, potential responders more accurately than before. Digital pathology and AI is, also, you know, of course, not just for the latest and greatest therapies, but also for helping to unlock, the potential of those that have been around for some for some time. You know, despite transforming oncology, we know that response rates for immune checkpoint inhibitors remains disappointingly low in many of the indications often below thirty percent. AI powered patient selection tools show promise to outperform current predictive methods, including, till density measurement and tumor burden tumor mutation burden. This reduces trial and error, treatment cycles, improves patient outcomes, and optimizes health care resources. We're also seeing remarkable advances in genomic biomarker prescreening prescreening directly from HNA images. I think this is one of the really exciting areas of AI in pathology. This capability is driving streamlining of clinical trials by allowing you to stratify study populations without running expensive genetic panels on every single patient. Not to mention what this could do outside of clinical trials as a prioritization tool to determine which samples in your vast repositories could be sent for whole genome sequencing. So as a result, digital pathology has shifted from, you know, being viewed as an innovative option focused on productivity gains to an indispensable cornerstone to our customers and ask them, what would you do if you couldn't use digital pathology tomorrow? They respond with some variation of, you know, I'd have some serious questions as to whether our breast or lung etcetera oncology program was gonna go away. Are we going out of business because digital pathology is always used to develop drugs for this disease? This is just how this works. And in the past, digital pathology, you know, really has always been a valuable tool to address operational challenges, accelerating image reviews, improving collaboration across global teams, and driving standardization and assessments. But today, the conversation is becoming so much more than that. Digital pathology has really become that essential cornerstone required to bring new drugs to market and get them to patients. And the organizations that invested early in the digital pathology capabilities, you know, perhaps initially just for those productivity gains, or to connect different silos across the organization, are now finding themselves, with a significant competitive advantage. They've been accumulating data over many years and now have tools to do so much more with that data. It's been a great privilege for us to work with the many customers who had been with us for a number of years and fall into that bucket. They have the infrastructure, the expertise, the growing repositories of digital pathology data that serve as that foundation for, you know, for the use of of AI and kind of next generation of, approaches. You know, I'm biased, of course. But when we talk to industry leaders across life sciences, they've made significant commitments in this area. We listen to market leaders and hear them more or less say the same thing. These aren't isolated, experiments or even department mental level initiatives. These are strategic enterprise level commitments, from organizations that are leading the industry. The question is no longer, you know, whether to invest in digital pathology and AI, but how quickly you can scale these capabilities, to remain competitive and increasingly data driven future of drug research and development. That's where we come in, and that's what we wanna share with you today. So for those of you who are not familiar with Prosha's solutions, let me share how we accelerate, you know, drug r d and precision medicine across the entire, discovery through development continuum. Our flagship product, our flagship platform, Concentric, serves as this central hub powering scientific AI driven workflows, unifying people, data, and applications. Built on the Concentric platform are two flagship products supporting pathology everywhere that it's practiced, from discovery through diagnostics, from biopharmas and CROs to hospitals and commercial laboratories. The first flagship product, Concentric LS, which many of you are familiar with, drives image based studies from discovery through early clinical development phases. And our FDA five ten ks cleared and CE IVDR certified Concentric APDx solution powers late stage clinical trials and accelerates the deployment of novel diagnostics. It approaches glowing growing, global clinical, lab network. Just to put that in perspective on the, you know, lab network side, twenty two thousand patients are diagnosed on Concentric every day as of today. It's amazing for us to have this front row seat to see the distance between, you know, life sciences and clinical get smaller as pathology is rewired for, the future of medicine. This includes the generation of new multimodal real world datasets, high quality research ready whole side images along with associated genomic and clinical data that can be delivered via Concentric, which many of you are already using to power your research studies, and augmenting an existing data foundation. So clinical data being delivered into a research environment. We have this open platform support. We've been big believers in openness with respect to image formats, with respect to AI applications. We now support over a 20 AI applications, as well as the developer tools, to drive your own proprietary AI development and deployment. Altogether, this centralized platform, the data and AI solutions allow life sciences organizations to maximize the potential of pathology and their data, and accelerate every milestone of of therapy development. So looking ahead, I wanna share some approaches strategic focus for life sciences. You know, this is a kinda high level view of some of the things we're thinking about to advance our vision for for digital pathology. We're constantly listening to customers, many of you who are on the call today, and we're excited about many of the things that you're telling us as we look to improve our products and expand our capabilities. As is usually the case when developing product roadmaps, changes can happen. We're constantly iterating through that. That said, when I look at this here and we're concentrating our efforts on three key imperatives that directly address these challenges and opportunities that we hear from our life sciences customer base. First and foremost, we're committed to enabling our customers to strengthen every stage of R and D with intelligent enterprise grade software. That means strengthening our solutions, including our Concentric LS and Concentric APDx solutions, with purpose built workflows across the discovery through development continuum, enhancing preclinical and early stage clinical development and concentric LS. It also means ensuring that insights and data can seamlessly transfer across teams and organizations, particularly enabling more seamless collaboration between pharma and their CRO partners. And we're trying to make that even more seamless for our customers in 2025 and beyond. We're also working to make Concentric LS as well as our APDx solution and clinical trial use intelligent with next generation native AI capabilities that drive productivity, drive better science in everyday workflows. We know there's many parts of the drug discovery and development process that can benefit from AI optimization, whether that be LLMs, vision models, vision language models, or some things that we're not even thinking of. While we intend to continue to evolve our integrations with best in class image analysis providers like PhysioPharm and and more, we're also, you know, focused on bringing these AI native capabilities directly to the workflows, in Concentric to reimagine how science gets done. We're also going to grow interoperability, standardization, and compliance capabilities to ensure that the Concentric LS product support growth across users, datas, multiple teams, making sure that's a robust, scalable solution that can grow as organizations are creating more data and using our products in larger, more scalable ways. Scalability, compliance, huge focuses of ours. That's bucket number one. Second, we're focused on empowering R and D organizations with real world data assets. We're expanding our multimodal real world database to provide a broader coverage across indications. We're also going to be enriching that data with deeper AI generated insights to deliver pre analyzed information as part of those real world data cohorts and ensuring that can be delivered directly to Concentric where you're doing your science. Third, we're committed to advancing our customers' novel AI strategies to provide them with a competitive advantage in the market. We're making incredibly seamless to deploy novel AI models on the Concentrix platform. We're also expanding our concentric embeddings capabilities to include additional foundation models as those come onto the market. Finally, we're strengthening our end to end companion diagnostic offering to ensure that new biomarkers can transition into clinical practice. We see this full vision here where our customers are using versus real world data and our Concentric LS platform to drive insights. So we think these priorities align, you know, directly with the industry transformation that we've discussed today and will help R and D teams harness pathology in this moment when next generation therapies, precision medicine, and AI are converging to make pathology more impactful and more valuable than before. I'm going to pass it over to Adani to talk through the latest and greatest in Concentric LS, which I think you'll all be really excited about. Thank you, David. I'm looking forward to seeing these solutions come to fruition for our customers. Now to take a step back from the future and into current day, I wanted to take some time to go over the recent enhancements we made to Concentrix LS platform that are now live and available with the latest version. Shameless plug, upgrade your system. So I'm going to go through, streamlined fluorescence workflow we've done for those especially that using Hyplex fluorescence imaging, faster and more structured annotations, and I will hand it over then to Ashley to walk us through, smarter automated quality control or automated QCS, we call it, and as well as new foundation models added to our concentric embeddings collection. Alright. First, one of the things we've added that that I am quite excited about is the ability for to save a set of active fluorescence channels as a group. That can be applied easily across multiple images within a single repository where you house your studies, but also reuse it in other studies if it's relevant. We're excited about this because it allows our customers to leverage Concentric LS, for discovery and translational medicine R and D, especially if you're focused on spatial biology. One of, the main gain you have is it increases the review efficiency or it also allows you the ability to actually configure phenotypes if there are phenotypes relevant in terms of your multiplex images. With these predefined multichannel configuration, that can easily be applied across multiple images, it just makes it quite easier to review one image after the other or compare images across in one repository compared to without this solution, you're in a position where you're continuously deselecting and selecting fluorescent channels as you go through images. Plus, the other part is the channel groups are searchable, allowing you to easily find repositories containing specific channel groups. This is really maintaining what Concentric does today, which is where if you have these metadata or any items that are relevant to the image already, associated to images or studies, they are searchable and you are able then to refine them. This also goes into the ability to do data reuse if that is needed. One of our, continued goal with Concentric LS is to allow users to standardize and reduce variability by maintaining consistent configuration across the organization. Users can add channel groups to shared libraries, thereby allowing them this standardization. In terms of this workflow, when I do the demo, you will see this. It's more you do also have the option to do a pilot study such that you're not, adding this to the library, but you're creating one just unique for the study you're working on. And if it's relevant, you can make it into a library item as well so that is reusable in future. As, most of the other solutions in Concentric, channel group is governed by permission and access in terms of who can create and modify them. It is under roles and permission as well as inter the access people have for different studies. It ensures secured that you have secure and customizable use user user, review user workflow in this case. Additionally, if the work you're doing is something that needs to be to meet some regulatory compliance, it is, traceable in the in our detailed audit log, and that's an option also if a study you're doing requires reason for change to be met, that is something you can enable and channel groups would be governed under that as well. And, if the data is if your study is archived, your channel groups also won't be modifiable, allowing you to maintain data integrity. In addition, some of the other things we've done as, you know, as we're talking about what is currently available is that we've improved, intensity scaling for fluorescence images. This is something you'll see a new image ingestion. We auto adjust in terms of what kind of brightness it will be so that it's quite easy for novice users to automatically open an image and it's, the kind of fluorescence image that they would be expecting. You still have the ability, of course, to adjust this as you look at fluorescence images. And, additionally, we've also added cam gamma correction, which would improve visualization, especially for images that are nonlinear and staining. All of this, we will go through as we as I go through the demo shortly. Second and next up is one of the things we've done is more structured annotation, also faster. This is, this is we I'll go at around a later point as well that we've gone a bit even more faster. But even in what we have today, we've added some structuring as well as the in terms of what we can handle has gone up quite a bit. In terms of in research and development, precision and efficiency in pathology image annotation are essential for accelerating studies and generating robust training data for AI development. Our latest version delivers some more intuitive annotation workflow complete with annotation classes, empowering pathologists and scientists to enhance your approach skills while streaming streamlining reviews. In terms of highlights, you have, customizable annotation classes, which allows you to categorize your annotation. In terms of examples could be, right, tumors, trauma, necrosis, lymphocytes, or it could also be something more specialized that is very specific to your own r and d workflow, which would be where you would actually create them and then reuse them in one study or across studies. To this allows you to improve right consistency and enhance data curation and ground truth labeling for AI training as well. As we as David just mentioned, right, we are we also anticipate this, annotations to progressively be less manual and actually be, more done by AI. However, you still need some amount of data curation, and this is where in terms of things like classes allows you to, let's say, annotate some of your images to see if they have a specific category that's relevant so it can be used from that data curation perspective or from a training dataset. The other parts that we've added is annotation access, annotations tools instantly in a viewer with a reimagined toolbar, where basically, you know, in a research and development workflow, annotations are a sort of analysis. Right? You do have a set of data that you're getting. So we've enhanced that such that you can see it in a table, which makes it quite easy for you to compare your different kinds of measurement, as well as, Concentric already had metadata associated with annotations. And if users are desire to use that feature in this table format, it's a bit easier to see those kinds of things as well. And, additionally, one of the things we've added is also the ability to create, rapid succession of annotations. We've enabled a keep annotating tool mode where such that if you've enabled it, you are able to create multiple annotations. Again, it goes into the whole if you're tasked with, let's say, annotating quite a bit for an AI training dataset, we were eliminating the click such that if you've enabled this, you can go ahead annotate quite a bit and you'd be done. However, we also did not wanna forego the workflow such that for pathologists, for example, they are typical. They typically are annotating maybe one item and an image or maybe two, and this would not be as helpful for them. So you still have that capability as well. Alright. With that, let me jump into our demo. Give me one second. Alright. I'll start off with the order that I we did a presentation. What you have here is a curation of a repository. It is not a study. While I am a neuroscientist, this is not a research project. So what you're seeing is a curation of images that fulfills the two workflows that I just discard discussed. So here, what you have is a set of fluorescent images, and then, what you also have here is actually real world data, which David will talk a little bit about from our process real world data in terms of what we have here. If we go into the fluorescent images, I will start off with some of the things that we talked about, which is that we now have gamma enhanced, you know, an ability to adjust the gamma value. For that, it's not something you would need particularly in this when you see this overall image, for example, in this case. However, if we pick one of the channels here, this one, you can see it's relatively dim. You probably can't see anything. Since fluorescence also does better in dark mode, I'll go ahead and switch to dark mode. And you are in a position where you could just increase the intensity and see how that goes. However, it's, if it's not ideal, what you would we can do is enhance the gamma correction. And this is actually a little helpful, In a typical use case, when you're looking at fluorescent images, you would be in a little bit of a darker room, and I do have a light right here, so I could barely see things myself. So it's actually helping me. So you can adjust that and this, like everything else is covered by, permission and access. In terms of all users can make this adjustment, however, in terms of whether this will be saved as a setting to be, such that if you come back to review it yourself or if other users are reviewing these images for that to be saved, you do have to have the appropriate permissions. But you now have this, and you can also see in the graph that the gamma correction application, so you could see that. Alright? Now going into the item that we spend majority of the time, what, some of the ways you can use channel group is if you can see, and I'll actually present it in terms of from the side of how it can be a little painful without channel group. So if I'm interested in, let's say, looking at this nuclei and a CD one sixty three label, or let's add a few more, you can see that you're in a position where you can see this item, which is great. But if I want to, let's say, review this in the next image, you're in a position where when you go to the next one, you have to do this again, basically, to select these items again. And even within that, say say, this image, you have a specific phenotype you're looking at and you wanna look at those combinations. You're in a position where you'll turn these two, you'll look at the this value. Okay. This is interesting, but now I wanna see these other combination. You're turning those off, and then you're going in and turning those. So you can kinda see how it could be cumbersome. And this isn't, this is an image that has, you know, not as many channels in terms of we are getting to the point where customers are labeling up to 30 or even close to 40 channels. And so we knew we needed to do we have we need we knew we needed a solution where this level of clicking needed to be reduced. And so with that, what you have is with channel group, and I'll show you how it's saved. But if we skip that for now, I have things that are saved. I am not a cancer biologist, so I'm not gonna claim these are appropriate phenotypes. I think they are, but I'm not gonna make that claim. So if we go with these appropriate with these if this is a phenotype, let's say, we're interested in, you can see how once I select that, the appropriate channels are selected. If If I go to the next image, for example, here, you can see those same channels are selected. So what this allows you is basically if you go from one image to the next, you're no longer clicking over and over again. Rather, you're in a position where these items are selected. Again, I've curated a few, so I can just easily go to the next. You can see now, the items that I desire are on. And if I wanna go to another one, I can just go to the next. And so you're in this position where this review, as you can see, that multiple click I was doing to illustrate what it can be is no longer something you have to do. But you also still have that if you need it. In terms of saving the channel group, it's quite simple. All you have to do is select the channels you're interested in. You say save channel group, and you can go ahead and save it. And if you want it to be added to the library, you just enable this, and you would get that. Alright. And so with that, I'll close the channel group section, and let's go into fluorescence. So I'm gonna go to bright field light mode and go to a bright field image. And this, like I said, is a set of images from our real real world data. And in terms of and you can see in this case, it's an HME labeling from, female, sixty five to 59, each group, and the anatomical side is also there. And you have the diagnostic, which is, adenocarcinoma. So in this case, for example, if we go ahead and zoom in a little bit, and we're talking about annotations, right? So in terms of annotations, we talked about especially classes. If you can review this image, how do we I'm going to do a little enhancement. This is just a me thing. I would prefer this to be a little enhanced, so I'm just going do some correction there. So if, you have the option to just go ahead and annotate as you see fit. Now in my case, let's see. In terms of the art classes already created, there's a tumor. And I can go ahead and say this is I'll go ahead and, annotate a tumor. But I'm also like you know, I admitted that I am a neuroscientist, not a cancer biologist, so I'm just gonna create something that's a little more vague, and I'm just gonna call it anomaly and, apparently, can't type and go with that. If I go ahead and create that, I will go ahead and turn this on and I'm going to keep annotating also on. And if we go from there, you're and this you can see at least I can see the nuclei is off in this region, and now I've annotated that. And if I have other regions I wanna look at, you can see that I'm gonna go to a lower zoom to make it a little bit easier. And if I go to this region, this looks a bit interesting. You can see this area also has some areas that are a little bit off, maybe, again, maybe tumor region, but also could be a necrosis and so on. So I'm gonna go ahead and also annotate that area as well. Alright. So with that, if I say this is good, this is good. And you can see if I open the table, you are able to see actually, I didn't do my second annotation. Let me just I didn't connect the two polygons. If I can go ahead and this area broadly annotated, you can see this is annotated. So you have these items in this case. And, one of the things I didn't do was actually, was I needed to select the class first, which is okay. I can illustrate then the point that I wanted to make, which is I can select these items and assign them to an annotation to a class. If I go ahead and add that, you can confirm and you have that. So in this case, again, I can hand it over to a pathologist who actually knows whether what to pull as a tumor or a necrosis, and they would be in a position basically, assuming they have the appropriate permission and access again. If they have that, they would be able to reassign this to the appropriate places. Alright. With that, I will, stop sharing and send it over to you, Ashley. Thank you so much. You could bring the slides back up. Yes. There we go. Alright. Awesome. Thank you, Adani. That demo was awesome and great overview. So, yeah, before I I jump into the automated QC improvements, I want to take a step back for those who are not familiar with this solution. So automated QC is an AI powered product that identifies common artifacts across H and E, IHC, and other specialized stain types. And these artifacts are automatically segmented and labeled directly in Concentric LS, which can have a really big impact in three key areas. So first, let's talk about speed. Automated QC reviews a thousand slides in the time it would take to manually review just a 60 slides. And this is achieved even when running at maximum sensitivity, so you're not sacrificing quality for speed. And speaking of quality, automated QC delivers consistent, reproducible results with 96% accuracy detecting 99% of artifacts. And perhaps most importantly for your AI initiatives, this ensures that you're training your models on high quality data. So, you know, there's been published research showing that even seemingly minor artifacts and training data can have can significantly degrade your AI model performance, and automated QC helps prevent that issue entirely. So another unique advantage that we like to talk about with our solution is that it's scanner agnostic. So we've trained and optimized it across multiple leading scanner vendors so you can really maintain consistent quality standards regardless of your hardware infrastructure. So let's talk about the improvements with the latest version of automated QC. This product has now been used in over 2,000,000 slides processed in production environments. This amount of use has given our AI r and d team a lot of intel into where we could drive impactful improvements for our customers to make it really smarter than ever. So as a result, automated QC now has higher accuracy and artifact detection with improved sensitivity and specificity at the pixel level for detecting tissue folds, out of focus areas, and air bubbles. And for those same artifacts, it has more precise region mapping. So now more frequently, it's detecting artifacts, regions that more closely match those actual artifacts, and we've also added new resolution settings. So standard resolution and high resolution, similar to our existing sensitivity settings, it allows your organization to be in control, make sure the software meets your desired balance of accuracy and performance. And we've also just been able to optimize parallel processing, which reduces bottlenecks when running automated QC across multiple large scale image repositories. So that is it for automated QC. Now I'd like to transition to one of our, you know, more exciting developments or new foundation model integrations. First, I'd like to provide some context around the current state of foundation models in pathology. So as many of you probably know, there's an increasing number of foundation models available every day. Actually, a recent paper that we came across documented 40 different foundation models specific for pathology. So what we hear from our customers is not is that, you know, there's not so much a challenge in finding a foundation model existing that meets their needs. There is a challenge in selecting the right foundation models to use for a particular product for a particular project and breaking down barriers that prevent one from actually, like, using those selected models in practice. So really two important things to keep in mind here. There's no one size fits all foundation model. Some are better than others in specific tasks, use cases, working with specific datasets. So you have to be really deliberate about the foundation models you select depending on what you're trying to do. And there's also research that shows combining features from multiple foundation models can outperform a single foundation model in many tasks. So for that reasons, we found that data scientists and AI developers really want to have more than one foundation model at their disposal so they can address different use cases. They also want to improve performance, not only by doing rapid prototyping that allows them to select that best foundation model for what they're building, but also to try these ensemble techniques to combine features for multiple models to get higher performance. And they know new models are coming out all the time. They wanna be able to use the latest and greatest. So this understanding has really guided our approach to foundation model integration and concentric, which we call our product concentric embeddings. We've divide designed a flexible architecture that addresses these specific needs, allowing your teams to leverage the best of multiple foundation models while maintaining streamlined development workflow. So with concentric embeddings, your team can generate embeddings for multiple foundation models right in Concentric LS where your data already lives. And this is why we're so excited to continue growing the collection of foundation models available through Concentric Embeddles. So Concentric Embedding now includes two of the largest and most advanced pathology foundation models, h Optimus zero from Bioptimist and Birkow from Paige. These models have consistently been top performers in independent benchmarking studies for critical tasks, including disease detection, cancer subtyping, gene expression detection, and more. And just the massive scale of these foundation models helps ensure really, really strong performance for downstream models across a wide range of tissue types, disease areas, pathology tasks. So you can just see how concentric embeddings allows you to easily experiment with different models, combine their strengths, and ultimately build diverse AI solutions that advance your initiatives. And just quickly, I wanted to really talk about how these initiatives span from discovery to development. So with one of the benefits of having several foundation models at your fingertips for data scientists and developers on Concentric LS is that they can build a wide range, supervised and unsupervised models with impact across the r and d life cycle, including model models to identify novel targets, accelerate clinical trials, stratify patients, and more. So we're really excited to continue to see what our users build. And now, I'll hand it over to David to discuss the fuel behind these AI models pathology data. Thanks, Ashley. So we kinda talked about two pillars of Prosha's product offering. First, our platform, which you saw from Adani. And, Ashley talked about some of our AI applications and foundation model technology. The third leg of this tool is data. And many of our customers are have been using Concentric to, you know, generate, manage, and analyze lots of their data for, many years. And I wanna spend a few minutes talking about one approach's newer offerings, our real world data offering. We haven't had a chance to do it in a forum like this yet. Let me frame why we think this matters so much for precision medicine. We think pathology data represents the missing piece of the precision medicine puzzle. And while the world has made tremendous advances in leveraging other data types, genomic sequencing, clinical records, imaging modalities like radiology have been used, in a digital form, standardized, and available at scale for for many years. Pathology has really lagged behind. It's lagged behind because in a clinic, unlike these other modalities, it's still largely an analog discipline. Physical glass slides, physical microscopes, really define the standard of care have defined the standard of care in pathology. And so as a result, all of the data ends up sitting on a piece of glass on a shelf, you know, collecting dust, doing nothing for for the world, locked away. And we don't we it doesn't have to be that way. The good news is that pathology is going digital in clinical settings finally after after many years. We're really at the beginning of this data revolution. And as all of that data is being created, it's fundamentally changing drug discovery and development. So it's catching up to all these other modalities. The growth trajectory really has been remarkable at the very early stages of this, but it's moving very quickly and becoming mainstream clinical practice, to use digital pathology over microscopes. You can see this on the chart on on the right where we're witnessing rapid clinical adoption of digital pathology. Analysts expect something like a billion slides cumulatively to be scanned by 2028. A lot of data that otherwise would be on glass. Earlier, I mentioned twenty two thousand patients are diagnosed on Concentric per day. That is 22,000 slides more, that can be de identified and used in research settings, Helps us gain a holistic characterization of tumors and disease dates with precision, accuracy, and reproducibility. We can use AI to unpack complex patterns within this data, develop AI based biomarkers, and companion diagnostics with real clinical utility, and the diversity from the real world, that's needed to have an impact in sort of modern, you know, modern therapies. So this is happening now. It's happening quickly. This is not some theoretical future. We've been able to, over the last year, work with many of our incredible customers to help them leverage this data in new ways. So today, our life sciences customers can access cohorts from Prosha's lab network. Over 2,000,000 unique patient records and 10,000,000 images, growing by 10,000 images per day and accelerating. All records within our network are multimodal, are targeted, and they're structured. We curate and enrich datasets using AI like LLMs or tools like Ashley just talked about automated QC to make sure this is structured high quality data. I think what really makes this, you know, approach to real world data, what makes it special, is that all that data can be delivered right into Concentric. So this is the platform that many of you are using to manage your own data generated in, you know, research settings, whether that's for discovery studies, preclinical studies, clinical development studies. You can now access data from the outside world, and leverage that in that same platform. It means that you can run AI models, perhaps that you've developed within Concentric to further enrich those records to generate novel insights that might be using off the shelf AI applications from Prosha, or some of our other integration partners, might be homegrown AI applications, including some of the new foundation models that Ashley talked about with our embeddings that are accessible via our embeddings API, which is, you know, especially popular with the data science teams. So that means you're not just acquiring data, you're sort of acquiring this ecosystem that makes data valuable and actionable. Then I could bring this to life and kinda, you know, talk shop here, with a couple of examples of tailored cohorts. Notice here, these span very diverse use cases from multi biomarker AI model development to enhanced patient stratification in one disease area to building a large foundation model across multiple disease areas to fuel drug discovery and development. Each of these was tailored to specific research needs and objectives. And you'll notice they're very different in size too, from four hundred and seventy five patients in a non small cell lung cancer study to over a million images in this large foundation model project. We have a team of data experts, data scientists, and frankly people who came from pharma that work with our life sciences partners to map those requirements onto our real world data network to understand those needs and then to curate those datasets to meet those needs. It's been a privilege to collaborate with our healthcare and life sciences partners to deliver these and many other projects in addition to the ones that you see here. Very much looking forward to doing more of this next year as we make this more accessible to our life sciences community and as our network continues to grow. With that, back to Ashley. Yeah. Ashley, over to you, Adani, for the sneak peek on upcoming improvements. Got it. Alright. Yeah. I'm excited to, I won't announce the date. I'll but, we one of us to give you a sneak peek, something that we've been working on over the past few months is, an area that really quite I'm excited that we've prioritized here is as our enterprise demand grows, right, like, performance is something that we need to meet. And it's it's an exciting time because we're seeing usage growing, our user base is growing, as well as the data that lives in concentric is growing. So with that, area that we've focused on is performance, and, we've we're seeing two performance that are two times faster than what it was, like, for the areas that I've just talked about. Especially some of the areas that we we wanna highlight is annotation rendering. This is especially, quite useful in terms of annotations coming in that have let's say been annotated by a machine on another in an analysis software. And when you bring them in, these are in the 10,000 and so on. And that rendering is something that we're seeing at a very, like, real really, you know, exciting performance. You could still do your multi view on all the other things that you like doing on concentric. And if you need to edit as well, those are some of the things. The other areas also that we've done quite a bit of work is in terms of what as you do, for those of you who have used Concentric and for those of you who haven't, like, probably asked for a demo, you can easily, add metadata values with doing a CSV import in the sense if you wanted to add metadata values for thousands of images, you can do that quite easily. If you have a in an Excel, you add the values, and then you could do that. And that's an area that is also, like, where we've done some enhancement both in terms of the export of that data if you wanna share it with people outside of your organization, for example, or when you bring back, to add value to your metadata. And other areas are also in terms of loading of, images loading. In terms of the, demo that I did, I actually didn't use the current version that we're working finalizing. Rather, I've used the version that's a bit, one version below that. And what you will see is, for example, with fluorescence images, the loading will be faster. That's that's something we know we need we knew we wanna continue invest in. Again, as people get into the 40 channels, this the size of the images grow, and that's kind of pretty nice one to see in in our four dot three version that we're currently working on. Other aspects that are always nice and concentric is that you can copy the images if you wanna, let's say, share it with your data scientist friends and so on. And so those are some of the places that we've seen some enhancement. There are additional areas that we've also done some work, but I'm not mentioning them here. But these ones, worth, like, you know, highlighting. Additional piece that, like, is exciting to see actually on the pathology space is there's a general consensus towards standardizing, file formats. Currently, Concentric does support a whole range of, scanners and their file formats, but an area that is starting to that we start to we are starting to see in pathology is standardization across scanners. And not yet fully standardized, but, however, this is the beginning of it and, something that our users should be excited about as we were expecting to have, Hamamatsu DICOM as well as, Pramana DICOM images supported on our upcoming release. Thank you. And and thanks, Adani. I'm so excited about all the things that we've done in in the latest release and, you know, excited to invest in the things that Adani just talked about that are coming soon. One of the things we really believe in deeply is shipping continuous value to our customers or constantly listening to other feedback that we're getting from you and and iterating on, on our products. So so I wanna conclude though by by thanking all of you here for joining the webinar, and also the for the work that you're doing for for patients. I think the world really has been making incredible progress in fighting diseases, from cancer to beyond. I think the progress in medicine is is really outstanding. That said, for many patients, these unmet needs remain acute. And behind all the technology and products that we've talked about today, there's real people who are waiting for better treatments, who are waiting for better diagnostics. And at Prosha, you know, we sit here playing this humble role in that mission. We think that pathologists and scientists who are working, to serve those patients deserve great technology. That's why we exist as a company. Very committed to being your partner in this mission. To that end, thank you for all the work that you do. Please don't hesitate to reach out to us with your ideas, your feedback, your challenges. We're all ears and so appreciate all the work that you do and how you help us become a better company and a better partner for all of you. So thank you. Back over to Ashley. Thanks, David, and thank you, David and Danny, for great presentation and discussion. Unfortunately, we only have a minute here, so we'll need to follow-up with everyone who submitted questions via email. So I'm very sorry about that, but, we will respond to you. So don't worry. Your question will be answered. And you can also submit questions to support@prosha.com, and we'll be sure to get back to you there as well. But yeah. So I hope you enjoyed this presentation. We will be sending a recording to your inbox as well as the inbox of anyone who registered for this session. We encourage you to tell anyone who you think might be interested in this content about the session. They can sign up, to view it on demand using the same link where you registered. And if you are a Prosha customer, I always recommend heading to our help center, support.prosha.com for a full list of what's new in Prosha as well as more detailed documentation on the functionality that we went through today. We also welcome your feedback, your feature requests. Again, please reach out to us at support@prosha.com. And really appreciate your time today, and thank you again, and have a great day. Bye, everyone.