100: AI in Pathology: DigiPath Digest #9 (From Retinal Assessment to Tumor Prognostics)

In this episode, we celebrate the 100th edition of the Digital Pathology Podcast!
Thank you so much for being part of this journey!
You are my Digital Pathology Trailblazers and I prepared a Digital Pathology Trailblazer manifesto for us!

This is the 9th edition of DigiPath Digest, and we are attracting more and more people to this series.

I am also working on a new YouTube digital pathology course and am offering the first 100 enrollments for free in exchange for feedback.

During today's episode, we cover several papers including research on AI for predicting post-operative liver metastasis, validation of AI-based breast cancer risk stratification models, AI applications in clinical microbiology, advances in parasitology diagnostics, AI for retinal assessment, and AI models for detecting microsatellite instability in colorectal cancer.

We also unveil a Digital Pathology Trailblazer manifesto emphasizing the ethos and dedication of the community.

Join us to stay current with literature, advancements, and insights from the fascinating world of digital pathology.

00:00 Introduction and Announcements
00:25 Live Podcast Proposal
01:40 Welcome and Audience Interaction
03:05 Updates and Apologies
06:11 YouTube Course Announcement
07:23 Technical Difficulties and Solutions
10:00 Digital Pathology Club and Vendor Sessions
11:28 First Research Paper Discussion
17:38 Second Research Paper Discussion
20:07 ER Positive and HER2 Negative Patient Subgroup Analysis
20:59 Independent Prognostic Value of StratiPath Breast Solution
21:59 Challenges and Benefits of Image-Based Stratification
22:58 Technical Difficulties and Live Stream Interaction
24:22 Introduction to Paper Number Three: AI in Clinical Microbiology
28:07 AI in Parasitology Screening and Diagnosis
29:30 Physics-Informed AI for Retinal Assessment
33:08 AI for Microsatellite Instability Detection in Colorectal Cancer
36:42 YouTube Course Announcement and Digital Pathology Trailblazer Manifesto
42:25 Celebrating the 100th Episode of the Digital Pathology Podcast

THE ABSTRACTS WE COVERED TODAY

📄 A novel model for predicting postoperative liver metastasis in R0 resected pancreatic neuroendocrine tumors: integrating computational pathology and deep learning-radiomics.
https://pubmed.ncbi.nlm.nih.gov/39143624/

📄 Validation of an AI-based solution for breast cancer risk stratification using routine digital histopathology images
https://pubmed.ncbi.nlm.nih.gov/39143539/

📄 Potential roles for artificial intelligence in clinical microbiology from improved diagnostic accuracy to solving the staffing crisis
https://pubmed.ncbi.nlm.nih.gov/39136261/

📄No longer stuck in the past: new advances in artificial intelligence and molecular assays for parasitology screening and diagnosis
https://pubmed.ncbi.nlm.nih.gov/39133581/

📄Physics-informed deep generative learning for quantitative assessment of the retina
https://pubmed.ncbi.nlm.nih.gov/39127778/

📄Artificial Intelligence Models for the Detection of Microsatellite Instability from Whole-Slide Imaging of Colorectal Cancer
https://pubmed.ncbi.nlm.nih.gov/39125481/

▶️ YouTube Version of this Episode
https://www.youtube.com/live/Uwca5rzAtEA?si=Rd8r4LVM1utEKWdt

Transcript

0:00

Today was one of the highest attended the Japan digest. it's already the ninth edition. So next week we're going to be celebrating 10th edition. We have several updates. That of course, international audience tuning in life. Pretty cool promotion for the YouTube digital pathology. Course that I'm creating right now. It's all covered in the live stream that you're going to listen to in a second. There's one thing that I, wanted to ask you, but I didn't cover in the live stream is. I am entertaining the thought of hosting live podcasts. Now all the podcasts other than the Japan digest are prerecorded. But I would love to invite guests life and give you the opportunity to ask questions life. When they show up, let me know if this is something that you would be interested in. Let me know. And LinkedIn, the amps or wherever you communicate with me, it can be comments. It can be email. Email is a good way. I always read all my emails. If you would like live podcasts with guests where you will be able to ask them questions. If you join life, obviously, right. let me know what you think of this. And without further, do you let's dive into the ninth digit path digest. Learn about the newest digital pathology trends in science and industry. Meet the most interesting people in the niche, and gain insights relevant to your own projects. Here is where pathology meets computer science. You are listening to the Digital Pathology Podcast with your host, Dr. Aleksandr Zhurav. Good morning. Good morning, everyone. Welcome. Digital pathology trailblazers. We are starting on point. I already see people joining. Let me know where you're tuning in from. Say hi. I am saying hi from Poland again. It's 12 p. m. 6 a. m. EST. But I'm lucky because I'm in Poland and I don't have to wake up so early. So, Of course, big shout out to everyone who has to wake up at 6 a. m. or earlier or late in the night. So let me know where you are tuning in from and while I wait for those people who maybe need one or two minutes to join, I will give you a few updates. Also microphone working, audio, visual, everything okay, right? Let me know in the comments. I can, so I realized I can see all the comments, but you can only see my responses on Facebook, YouTube. And I think streamer, if you're directly through the streamer link, but not on LinkedIn. So on LinkedIn you will not see my responses. Like for example, at the beginning, I always give a comment on welcome and say hi you will not see it on LinkedIn. So if you are on LinkedIn and welcome Ika then just comment and I'll show you on the screen. And Devika, let me know where you're tuning in from. So here, the updates as you can see by the background, you're probably are already used to my Polish background at my cousin's office. And it's kind of a special place, not only because it's my cousin's office, but my cousin is a forester and my dad used to be a forester, like my whole family and my dad used to have an office like this so yeah, it's special and we're still in Poland. And I didn't post a podcast from last week's live stream shame on me. I apologize sincerely because I dunno, I think I was supposed to look up some stuff that I didn't note on my PDF, like explanation of some abbreviations, and I was supposed to put it in the podcast. And the life came in between reality came in between and I didn't post it as audio. So it's only available as video on platforms, right on YouTube is the easiest to get to. And Devika is joining from India. Welcome. And she's working at the biotech and interested in the digital pathology space. Welcome. Welcome. So great to have you here. So, oh, and I see likes on Facebook. I love it. Give me some likes. If you're already I already see several of you give me some likes or some other reactions so that we can show it to more people. Okay. So, no podcast last week, unfortunately, sorry for that. But today I'm going to sit here after the live stream till the podcast is out. So everybody who's more. The listening person, then visual they will have availability of the podcast and there was a, such a nice thing that happened on LinkedIn. The NHS, sorry, NHS national histo technology society. They shared the post about the live stream. I was, I was touched. Like that was so nice. And yeah, let me know where you're tuning in from. Say hi and let me know in the comments. But the N H S E shared our live stream, our DigiPath Digest. And do you know it's already the ninth? And we said that when I do 10 of them, I'm going to buy myself a better tablet. So here's one tablet that I'm having. And then there's another tablet that I'm having. And of course I have the computer and camera and even a phone anyway, but. This. So after the 10th, well, the 10th is still gonna be here in Poland, and I think the 11th is gonna be in the us. So around that time, I'm gonna be starting to look for a new tablet. Mm. And so NHS thank you so much for sharing this post, and whenever you see it and you're planning to join, feel free to share. That takes us to more people, and that's the point, right. And September is going to be here soon. And in September, I'm going to park city to the digital diagnostic summit. You might have seen posts on LinkedIn that I'm going to be chairing a session. And I had a live stream with Bianca Collins, the organizer. And I asked her, Hey, like, why did you even invite me? And she's like, well, you're a digital pathology influencer. I want to have the digital pathology influencer chair a session. And I thought that was that was so nice. So, that's going to be happening and we were talking about the YouTube course. YouTube course is it's coming along. So, I mentioned to you that, oh, I need to make a curriculum out of the YouTube videos. Let me know if you hear me well, I think my microphone is working. I think everything is working. But just. Let me know in the comments. Anyway, so, the YouTube course, right? And I thought it's going to be like kind of a mini course, but then I looked, we have like 370 videos and counting, and there is like a full curriculum to be made out of this. So I already have the outline outline. And what I'm going to do is. For the first hundred people. So if you want to be one of those first hundred people who get this course for free, I'm going to have it for free because I would love you. And I'm going to ask you for something in exchange. It's going to be a testimonial. If you guys can give me a testimonial for this upcoming course, then for the first hundred people, it's going to be for free. So leave me a comment YouTube course, and then me or my team are going to get back to you. Saying that oh, my microphone is gonna die. That is not good. How can we mitigate this? Hmm, maybe. Sorry guys. I don't know. I think I have an additional cable. It may be charging the microphone. Let's try it. Expected problem. Sorry for that. Live streaming is like that. Okay, here's the cable and don't worry. That's the last update. And then we're going to be jumping in into the. Papers. I hope it's not gonna, oh no, disconnected. Not great, not great guys. YouTube. Okay, I have people wanting YouTube course and you might, oh my goodness guys, now I disconnected my microphone. Okay, we can do it, we can do it. I'm gonna find something else. Can you still hear me? Let me check my, my settings. Microphone. Audio. Audio. Is it working still? Sorry for this hiccup. YouTube course. Oh, and we have people from Wroclaw again. Okay. I assume you're hearing me. My microphone is saying that you guys are hearing me. The audio is still good. Thank you so much. Loud and clear. Okay. Perfect. Thank you so much, James. Okay. Yeah. Crisis saved. Okay. So, a YouTube course. Let me know YouTube course for free for the first hundred people. Why am I giving it for free? Because I want your feedback. You know, if you love it, then of course the, I would love to hear it. If you don't like it, then also give me feedback because then I'm going to improve it. Then. So this is going to be for free for the first hundred people, then it's going to be for 97. I want it to be just one time fee and 97 and it's going to be updated every year. So what happens with with this? Space, I took a couple of courses, right? And then the moment you take this course, there are new developments and the stuff that you learned in the course, I mean, it's good to know, but then it's obsolete. So I'm going to be updating this every year. So everybody who buys it or, you know, gets it, those people who get it for free, we'll keep it for free forever. Even the updates. And then there's going to be a special deal that they thought of. People who join live stream on live. or watch the recordings, we'll be able get the special discount code. So you guys who are here are the eligible. Well, I don't have the code yet because the course is not made, but there's going to be a special incentive for those who joined live to get it for less money when it's out and when the hundred free ones are already given away. So that's it. And of course, if you're a Digital Pathology Club member, then you will have it. No, like, you will have everything. Digital Pathology Club members, which is a membership that is 97 a month, is like, gets all the courses. And it's going to be revealed in September. We're going to be we're going to start having sessions about like live sessions, Q& A sessions. And there's also another thing and that's the end of the announcements guys commercials are over in a second then there's gonna be another thing probably starting end of september either once a month or twice a month I'm going to be meeting with the digital pathology vendors and digital pathology businesses those who are my sponsors are going to be already digital pathology play sponsors are going to be like invited by default But that's going to be open to regardless whether you are a sponsor So if you are a sponsor in a place to become a sponsor or not, it's going to be a session for the vendors to leverage what's happening in social media in this digital pathology space so that they can get in front of more people. And sometimes they're going to be guests. Sometimes I'm going to be analyzing the performance of stuff that I'm doing. And that's it for the commercials. Thank you so much for staying here through the commercials. And let's do the papers because that is the point of DigiPath Digest. And I don't know, nobody's saying where they're tuning in from, no, other than Wrocław. Joanna Joanna is from Wrocław. Fantastic. Okay, and now. Okay, my friends, we are starting with paper number one. A novel model for predicting post operative liver metastasis in R0 resected pancreatic neuroendocrine tumors, integrating computational pathology and deep learning radiomics. We have a lot of integrating pathology in radiomics, but I want to point out something super important here. Remember, when you see this circle here, Or like here, I created this color cause it's kind of H and E color. So I decided I need to highlight my stuff with H and E cause I'm a pathologist. Okay. Oh, we have Columbus, Ohio. Welcome Kimmy. Fantastic. Okay. So what happened here? This is a group. This is journal of translation. Sorry, my H and E color didn't work. Journal of translational medicine group from China, Shanghai. And who do we have? Like we have pathology, oncology, we have diagnostic and interventional radiology, radiology, and like neuro endocrine tumors, like a lot of groups together talking about this particular thing that they did. So let's start. What is R0? R0 is when you resect a tumor and the margins are clean. So there is no tumor cells at the edge of the resection when you take the tumor out. Okay. But there is also still, there is an option or like a risk or it happens past operative liver metastasis. And this significantly impacts the prognosis of pancreatic neuro endocrine tumors, pan nets even after R0 resection, right? So what do we do here? We combine computational pathology and deep learning radiomics and then to enhance the detection of post operative liver metastasis, liver of this pancreatic tumor, right? So we're talking about different organ metastasis. And what happened here? They used clinical data, pathology slides, radiomics images from 163 patients. And they had the R0 resection. And digital image analysis and deep learning identified liver metastasis related feature in CHI 67 stained hall slide images and then enhanced CT scans to create a nomogram. And I'm like, what is a nomogram? Guys, maybe, you know, I didn't know. So I checked it for you. So, and I did a better, a better job in making notes. So nomogram is a mathematical device or model that shows relationships between things. That's a pretty high level explanation. But basically nomogram is like not a sophisticated model, but like a bunch of. of numbers like that indicate risk or relationship between things, right? So like age maybe smoking status or like points, data points, and then you like connect them together and by connecting them, you know, okay, something's gonna, there is a relationship between something and something else. So here there was a multivariate logistic regression identified and What was regression? This regression identified na nerve infiltration as an independent risk factor for liver metastasis, and then they calculated a omics score. This, this word omics is coming up more and more often where you basically. Combine pathology images with other stuff like radiology and proteins and like some genomics, something that ends with onyx, but basically pathology is part of it. So histopathology images are part of it and. We then do paths and then, then they invented this pathoma score, which was based on a hotspot and the heterogeneous distribution of I 67 staining, and this showed improved predictive accuracy for liver metastasis. So, the Deep Learning Radiomics score, so here we have the Pathomics score, and here we have the DLR, Deep Learning Radiomics score, and it achieved AUC of 0. 5. 87, 87. 5. And then the integrated nomogram, which we talked about that combined. So here, here are the data that this nomogram combined the clinical clinical pathological and imaging features. And this demonstrated an outstanding performance with AUC of 0. 985. Like super great. Almost one. And in the training cohort, let's, let's be clear, training, but 0. 961 in validation. So great. And, and the conclusion of this is so basically like a bunch of data modalities and I get those questions and I was Like, even two months ago, I was not really sure, like, how to answer what is being done since we started the DigiPath Digest. I'm more on top of stuff, which was the point of DigiPath Digest, right? So, the, the conclusion, the conclusion is a new predictive model that integrates computational pathologics scores and deep learning radionics. This can better predict post operative liver metastasis in PANET patients, and it will aid clinicians in developing personalized treatments. So a lot of data is better than less data. Okay. Our second paper is, and if you have any questions let me know in the comments, even if I, so, the ones that I can address immediately and they're quick to answer, I answer immediately. The ones that will require like some digging up or, you know, later later answer or more, more thorough answer that I will not fit into this live stream. I'm going to answer In the comments, wherever you are. I know that most of you are on LinkedIn, but I know people join on YouTube and some on Facebook and other platforms. We are on Instagram. We're not on TikTok. I unfortunately abandoned TikTok for Poland because of the first debacle where like I was getting kicked out of my live stream. So paper number two, validation. of AI based solution for breast cancer risk stratification using routine digital histopathology images. So like normal agents and the paper, the journal breast cancer research. The group is from Sweden. It's Pina Sharma at Al Ska Institute at Stockholm, Sweden. And I like this one because, you know that I like when commercial solutions are are evaluated in scientific literature. So this was actually a commercial solution, a commercial test called Stratt Path Breast. This is a CE IVD, so in vitro diagnostic approved or cleared, or, you know, allowed to use in Europe. And this Stratipath is a company, and they have this test, right? This is an AI based solution for prognostic risk stratification of breast cancer patients. And they stratified them into high and low risk. Low risk groups and they just use h and e stained histopathology images and it is validation study. So basically a group, currently a group took it and made a validation study to check the prognostic performance of Strati Path breast into independent breast cancer cohorts, which is fantastic that they did it and published it because. I don't know. I think we need more of that because then you can refer to this literature and check, okay, is this up to my standards? Is this up to my regulators? Like, can I have trust in this solution? So, what happened? The methods, the retrospective multi site validation study included, 2719 patients, that's a lot of patients, with primary breast cancer from two Swedish hospitals. And then the StratiPath breast tool was applied to stratified patients based on digitized whole slide images. From those surgically resected tumors, and then the prognostic performance was evaluating using time to event analysis by multi variable Cox proportional hazards. So time to event was, do I have? Spelled out. No, but time to event was like, okay, what is the risk of something happening in a certain time? And their, their end point was progression pre survival. And the results were that the ER positive and HER2 negative patient subgroup the hazard ratio associated with progression free survival between the low and high risk group was two thousand, two point seven six. So the high risk was two point seven six times higher risk than the low risk group of something happening to them. Like, well, what the something is. That there was no progression free survival for a certain time, right? So that, that they, they had progression and they had the different subgroup as well. There is er mi plus her two minus Nottingham histological grade subgroup here in this subgroup. That hazard ratio was 2.2 low versus high. I know I had a note for this, but that's okay. It's fine. So there was doubt indicate an independent prognostic value of this strati path breast solution among all breast cancer, cancer patients. And this is it also shows improved risks, stratification of intermediate risk, ER plus HER two minus breast cancer. And this provides information relevant for one treatment decisions. Then to treatment decision of adjuvant chemotherapy, and has the potential to reduce both under and over treatment, which we all know that chemotherapy is not a benign treatment. And if you know, anybody who went through it, or if you went through it. It's, it's, it's terrible anyway. So I think it's valuable, super valuable to, to reduce under and over treatment and then image based stratification provides an added benefit of short lead times and substantially lower costs compared to molecular diagnostics. Okay. So here is my you know, star or caveat. Like, is this, I mean, it's the CEIDD, so maybe that is all okay in Europe, but basically this trend of, okay, you have some stratification tool or some like image based tool I don't think, at least in the U. S., any of them can replace it. And molecular tests, what I have seen so far in the literature is okay, you have like a this, this image based prediction or image based vision of something, for example, a mutation or whatever. And then you have to confirm and the confirmation can be faster because you already have narrowed down the stuff that you need. You need to confirm. So, but here they say provides added benefit of shortly. Thanks. It's substantially lower cost compared to molecular diagnostics and therefore has the potential to reach for their patient groups. And here, you know, we would need to go to the full paper, but the question is, okay, like, how many of. Oh. My space is, my, my stuff got disconnected, my papers got disconnected, that's okay guys, we're gonna figure it out, we're gonna figure it out. Why is everything failing me today on my tech? You know what happened before, like 15 minutes before the live stream the electricity went out and then I was reconnecting for like 5 minutes to to the internet and Hello Fermilag, great to have you. Okay guys, we're trying to reconnect with my tablet so I can draw on my I probably can't, so we're gonna do it that way. We can still do it, we can still do it. How do I do it? We're gonna start again. I'm just gonna do it straight from the computer and I probably will not be able to draw on it. See, that is a sign that I need a new tablet. Why did it fail me today? Okay, we're sharing again. Give me one second and in the meantime, let me know where you're tuning in from. I only have India, Wroclaw. I know some people are from the U. S., but where from? Which place in the U. S.? Shift screen okay. Now, you will need to tell me if you see this screen. Okay, I see myself on the screen. You should be able to see the screen. I don't have the screen, so, that's okay. We'll do it, we'll do it, we can do it. Let's do paper number three. And can I draw, I'll figure something out. You see my mouse, right? Let me see if I see my mouse. Yeah, I see my mouse. I'm going to be showing with the mouse. We can do it. We are digital pathology trailblazers. By the way, I have a digital pathology trailblazer manifesto that I'm gonna read after we go through the papers. Oh, and hello from Peru. Fantastic. Sabina, great to have you here. So I'm gonna, I'm gonna read this manifesto. For you and for now, I'm going to find the papers. Okay, we are here. And paper number three is potential roles for artificial intelligence in clinical microbiology. So, I know that last week people were happy that it's not all tumors. It's not all oncology. So here we have microbiology. And this is from Mayo Clinic, Arizona, Aaron Graf et al., also Columbus, Ohio, and South Florida Morsani College of Medicine in Tampa, and we are in American Journal of Clinical Pathology, and this is a review, and the review search was conducted on PubMed using key terms clinical microbiology and AI and they checked what, what has been done in the literature. By the way, if you want PubMed alerts for whatever topic, including digital pathology and ai there is a video on YouTube. You can go and find how to set it up. And so the results that numerous studies highlight potential labor as well as diagnostic accuracy benefits to the implementation of AI for slide based and macroscopic digital imaging. So slide based which is going to be the smears and like staining bacterial staining. So they also say these range from grand stains, interpretation to categorization and quantification of culture growth. So I don't know if any of you ever had to count bacterial count cultures, but you have this like round. plate and there are dots where you like inoculate those bacteria and then you like go and count. No, you have the, it's a plastic Petri dish and you like make a dot on the top of the of the dish and And underneath you have the colony and like, if you have a lot, it's a lot of like dots and then you have a clicker in the other hand to like count them. I'm like, how about we just take a picture and count them with image analysis? Like this is so easy. It's just dots. So, I'm very happy to learn that that's happening because I did that. I don't know some, vandy, hello. So I did that when I was doing my PhD and like, why do I need to do it? Start studying for my PhD. So now people don't need to do it. There are image analysis image analysis solutions for that or microbiology both for smears and then, you know, recognitions on skin. Same slides and macroscopic. And what's the, what's the thing here is that the conclusion is that AI applications in clinical microbiology significantly enhanced diagnostic accuracy. Yes, it definitely would increase my accuracy and efficiency in counting dots. Like really computers are more efficient in that. Okay. I need to make myself smaller again so that you can see the paper. Okay. Now And yeah, but the, the problem is, or the problem, the, the, the thing is always, okay, will the regulators prove we need to get clearances. So us food and drug administration and needs to clear devices that are medical devices. And that's, that has to happen in microbiology. So, our paper number four this is also a kind of review. So those reviews we can go quickly through, but it's no longer stuck in the past. I like the title. New advances in artificial intelligence and molecular essays for parasitology, stream screening and diagnosis. Why is this important? Because Parasitology, we have like blood parasites and you have to screen for them if they're in any case of blood donation. So, the findings here were that artificial intelligence has emerged as the promising tool for blood and stool parasite review. And, and there are already like commercial vendors doing this for veterinary medicine, but also for human medicine. for your time. Pexite is one of the companies that is pretty prolific in, like, on different fronts for for everything, all the parasites for veterinary medicine, blood parasites, also mold and microbiology counting. So they're leading, leading the way. But here they also were talking about so not only stuff that you can use AI and count or calculate or detect on images but also additional technologies like MALDI TOF, metagenomics, flow cytometry, and CRISPR Cas but these are under investigation for the, for their diagnostic utility. Definitely. AI on swift smear screening is something that could be high leverage and a cool thing. Next paper nature in nature communications physics informed deep generative learning for quantitative assessment of the retina. This is cool. Okay. So, this was done in London and Cambridge, basically in the UK brown at all. And What, why is this important? Why is this AI for, and this physics informed, like, what the heck? Why are we here doing physics? Well, physics is part of medicine, but so the disruption of retinal vasculature is linked to various diseases. So, one of the very famous one or like the common ones is diabetic retinopathy and macular degeneration, and you can basically lose your sight when you have these. So there is a novel algorithmic approach that generates highly realistic. Connect, sorry, highly realistic digital models of human retinal blood vessels. I'm like, like, why would you need a model for this? Like, why is it generating a model? But basically it's doing, it's, it's segmenting the vessels. So you'll have the image and then. You have a program that segments the vessels and extracts insight from this image. And this approach is using physics informed generative adversarial networks, GANs. And it enables segmentation and reconstruction of blood vessels network. And they checked it on. Drive and star. What the heck is drive and star? So these are data sets with a lot of retinal images. So like in pathology, we have the TCGA, the tumor cancer genome Atlas, where like you can validate or train your models. You have a lot of images available there. And here there is a drive. digital retinal images for vessel extraction and star, star structured analysis of retina data sets. And it was validated there. And it was a state of art vessel segmentation without a human having to help, so without annotations, without interpretation. And the physics stuff that was relevant here were two physical laws. And the Mary law, which states that vessel diameters, branching distance, and branching angles are optimized. to form a balance between pumping power and blood volume and minimize resistance to flow, like very much physics, right? And there's fluid dynamics. So basically they incorporated these laws, these physics laws into the, how this gun operates the, the generative adversarial networks. And that's why it's better because it kind of has rules, which is. Something that I like a lot in the digital pathology space, which I had to learn by doing is like, whenever a new method in image analysis was coming up, I was like, Oh, this is the next thing that's going to solve everything. And then I learned the method and I saw that, oh, it also has limitations, like every method, right? So, here it's a combination of the deep learning. Even generative generative ai. So that there is a, one of those of this pair of networks is a generative network. And we have rules, we have like physics rules, so it kind of narrows down what the AI is doing and then the AI is not going crazy. It's basically operating within the laws of nature that were taught to it, which is fantastic. And this is you know, nature portfolio. So. I impact factor and let's do one more, right? I have time. Do you have time? Artificial intelligence. Let me check. Yes. Oh artificial intelligence models. For the detection of microsatellite instability from whole slide imaging of colorectal cancer. So that's not new, but it's relevant. Here we are again in the space of predicting molecular from image without doing molecular tests. So this is a group, this is Gavinofa et al published in Diagnostics Basel. And this is a group from where Italy, Italy, Italy, Belgium, and Italy, mostly Italy. And I'm checking if we have any, anything else. I thought we would not have time for this, but let's just go through this one. So basically I love the, I love the introductions. Which I always think, Oh, like, we all know this because we're in digital topology trailblazers, but then I take like a radiology or ophthalmology thing, and they have the same basic type of introduction for people from other disciplines. And then there I appreciate it. So let's not get annoyed with abstract with introductions that are obvious to us, but basically whole slide imaging can scan digitally whole slide images at high resolutions. And this is our digital revolution, although my opinion, the revolution is going to come when we have imaging direct to digital from tissue direct to digital without making glass left. That's what I'm waiting for. That's going to be like, we're going to be, we're going to be unstoppable when this is the case, because now everybody is telling us you'll still have to like, Add on top of the analog workflow and you just like scans as if you were printing images, printing photos and then scanning them instead of taking digital images, which is the case but the direct to digital is coming. So anyway, here the potential predictive role of the most relevant artificial intelligence driven models is predicting microsatellite instability directly from histology alone is discussed focusing on colorectal cancer. And where's the result? Yeah, basically the use of digitized. H and E stained sections and trained algorithm allow the extraction of relevant molecular information, such as, such as micro satellite instability status in a short time and at the low cost. So here again, Low cost. I would put the star here because the computers that compute this are not cheap. So, you know, I mean, if somebody can send the image there and they do it for them, and it doesn't cost as much as a molecular test. That's one thing. If you need to get all those computers, because currently these things are pretty computationally heavy then it's not as low cost as as advertised So we keep that in mind because we are digital pathology trailblazers So let me finish this one and i'm gonna read you the manifesto and the possible advantages, okay The possible advantages related to introducing deep learning methods in routine surgical pathology are underlined here And the acceleration of the digital transformation of pathology departments and services. Is it recommended? Yes, I recommend the same I recommend that we go digital I recommend that so Oh, yeah Vendy is asking what's this youtube course that everybody is commenting for those who who change? Who who came later? So there was a commercial part at the beginning so i'm working on a youtube course You That is gonna be basically a structured curriculum based on my YouTube videos from my channel. We have over 370 videos and I thought of like making a little mini course with a few videos and then I started Working on the course outline and it's like gonna be a full blown course from YouTube videos. So if somebody wants to pay for it you're going to be paying for the curation, putting this in like having this in a structured way. If you don't want to pay, it's going to be on YouTube, but the comments are for those who would like to test it. I'm going to give it away for free for the first hundred people to get feedback. testimonials and feedback. If you love it, fantastic. If you don't love it, you let me know so that I can update it in a way that it's more useful. So whoever wants this course for free, a hundred people are going to get it for free when it's ready. Comment YouTube course, YouTube course, whether you're watching this live or you're watching this as A recording replay or whatever, and then comment YouTube course wherever you are. My team is going to go into LinkedIn into YouTube into wherever it is being streamed. And we're going to collect those comments and put you on the list. Of those who want to get it for free and are willing to give us a feedback Give me feedback if it's good or not. And now before we finish this fantastic live stream so yeah who wants that let me know for free 100 people 100 people only 100 because then I can only handle 100 I've got people giving me feedback, but yeah, okay, perfect. I'm glad you're going to get the YouTube course. Thing on YouTube, I think if you are watching on YouTube and commenting on YouTube, you might need to send me a DM on LinkedIn because on YouTube, I don't know exactly who's sending me the messages. Hello, Mohammed. Fantastic to have you here. So guys now. manifesto. I created this manifesto because we are digital pathology trailblazers. So let me find it. I have it on mine. Okay. And you tell me I should put it as text. Let's see if I can, so that you can see it. I cannot do it on the fly because I have, don't have my usual setup, but that's okay. I'm going to read it to you. So, Digital Pathology Trailblazers. We are the pioneers of a new era in healthcare. We embrace collaboration and open communication. Our vision, well, a little bit Okay, let me just read it to you and I will be updating this, because it sounds a little bit Like, anyway, it doesn't matter. Let me not comment it. So, digital pathology trailblazers. We are the pioneers of a new era in healthcare. We embrace collaboration and open communication. Our vision is unparalleled. Our resolve unshakable. We face challenges head on. Every obstacle is an opportunity for innovation. We don't complain. We create solutions. Our knowledge is our strength. We understand the intricacies of digital pathology. We make informed decisions based on expertise and evidence. Interoperability is our goal. We champion standards and seamless integration. Our systems speak the universal language of progress. We are lifelong learners. Staying current with literature, DigiPath Digest, and advancements is our passion. We seek knowledge relentlessly to push boundaries. Teamwork defines us. We recognize the power of diverse expertise. Our strength lies in our collective wisdom. Clarity is our commitment. We translate complexity into understanding. Our words bridge gaps and unite minds. We are the catalyst of change. From vision to reality, we make digital pathology happen. Our impact resonates through laboratories and beyond. Meritocracy drives us. Excellence is our standard. Innovation, our currency. We value contributions that propel us forward. The future is now. We acquire cutting edge technology and knowledge. With these tools, We shape the landscape of healthcare. Together, we are the architects of a digital pathology revolution. We don't just adapt to the future. We create it. What do you think? It's like very manifesto like manifesto. But what I wanted to include, I'm probably going to be like shaving stuff off of it. But what I wanted to say is what I said there, but like, in simple words, is that we have expertise and communicate with other experts. We make things happen. We figure out workarounds. We don't complain. We find solutions and basically like, you know, you are here. A lot of you is waking up at 6 a. m. or at some other strange hours and, and, and you guys are my tribe. So that's why I created this manifesto. I'm going to post it somewhere, maybe in the show notes, maybe in the comments of the live stream so that you can give me your impression of what that is. And again, if you want the YouTube course for free, come on YouTube course, and I wish you a fantastic day. As promised, I'm going to be sitting here until it's in the audio form as well for those who just listen to this and don't watch it. And I talk to you in the next episode. One more thing before you leave. Did you know that this was the hundredth episode? It was the episode number 100 of the digital pathology podcast. If you've listened till this moment, then you are a true digital pathology trailblazer. I would love to ask you a favor. Would you be able to give it Five star review. Of course only if you love it. If you don't love it, you probably are not listening to this. That far into the episode, but if you love it, could you help out to get it to more people and give it a five star review on the podcast app that you're using apple podcast, Spotify, whatever other podcast app. I would very much appreciate, and I can't wait. To have you in the episode, number 1 0 1 next week.