182: AI, Quality, and Standards: The Next Chapter of Digital Pathology

This session is a practical walkthrough of where digital pathology and AI truly stand in early 2026—based on five recent PubMed papers and real-world implementation experience.

In this episode, I review new clinical adoption guidelines, AI applications in liver cancer imaging and pathology, AI-ready metadata for whole slide images, non-destructive tissue quality control from H&E slides, and machine learning–assisted IHC scoring in precision oncology.

This conversation is not about hype. It’s about standards, validation, data integrity, and clinical translation—the factors that decide whether AI tools stay in research or reach patient care.

Episode Highlights

• 01:21 – Practical digital pathology adoption guidelines (Polish Society of Pathologists)
• 08:05 – AI in liver cancer imaging & pathology, and why framework alignment matters
• 18:10 – AI-generated tissue maps as metadata for WSI archives
• 23:01 – PathQC: predicting RNA integrity and autolysis from H&E slides
• 32:14 – ML-assisted IHC scoring in genitourinary cancers
• 29:42 – Digital Pathology 101 book + community updates

Key Takeaways

• Digital pathology adoption still requires clear standards and validation workflows
• AI performs best when aligned with existing diagnostic frameworks (e.g., LI-RADS)
• Metadata extraction is a low-effort, high-impact AI use case
• Slide-based quality control can support biobanking and biomarker research
• Automated IHC scoring improves consistency—but adoption remains uneven globally

Resources Mentioned 

• Digital Pathology 101 (free PDF & audiobook)

Publication Links:  a. https://pubmed.ncbi.nlm.nih.gov/41618426/                                                                 b. https://pubmed.ncbi.nlm.nih.gov/41616271/                                                                   c. https://pubmed.ncbi.nlm.nih.gov/41610818/                                                                 d. https://pubmed.ncbi.nlm.nih.gov/41595938/                                                                 e. https://pubmed.ncbi.nlm.nih.gov/41590351/ 

Transcript

00:00:02 - 00:01:15
Aleks: Welcome digital pathology trailblazers. Good morning. 6 am from Pennsylvania, saying hi in the chat. So if you are here, let me know that you're here and where you're tuning in from. And of course, what time is it? Did you actually wake up at 6:00 a.m. to do this with me? I know some of you do and I appreciate that very, very much. So, let me know where Okay, I see people. It's there's always a little bit of a lag before like when I go live and then I'm like, are they coming? Are they


00:00:40 - 00:01:59
coming? You are coming. So, let me know if you hear me well, if the mic is working. Let me know where you're tuning in from. And it's already February. Can you imagine? 2026, the second month. Coffee ready. Waiting for some of the reactions of you to know that everything is okay with the tech because regarding the tech, I have a new computer and different new things that hopefully will not give us any trouble today. And my trailblazers are here. Welcome, Monica. Amazing. uh if audio good,


00:01:21 - 00:02:51
visual good, then let's dive into the papers um into the abstracts that I have for you today. We have a few cool ones and I am so super happy to start with guidelines for the adoption of digital pathology and clinical pathology units recommended by the Polish Society of Pathologists. this. Oh, and we have guests from Estonia. Amazing. Eastern Europe. So, I'm super happy, of course, that Polish Society of Pathologists gives us these guidelines because I'm from Poland and uh I'm looking where the authors are


00:02:05 - 00:03:15
from and they're from Bitcoin. I went to high school in Pausnine. Um, so I know where that is and uh I'm of course super proud about this being published in diagnostic pathology and I checked. So let's go let's go high level first because uh these are guidelines for the adoption of digital pathology uh in clinical pathology units. This publication comes out of Poland but it's not specific to Poland. And actually I went in and I want to show you what's in there. um for you to actually maybe


00:02:43 - 00:04:03
we're going to review it in more detail. So, if you would like to go through these guidelines uh with me at some point, leave me a comment uh with the word guidelines because we can just go through them together. Uh but for the purpose of Digipath Digest, I just want to mention this to you and but I'm going to show you what's in there. Give me one second here. So, um they have advantages and applications of digital pathology. We're familiar with that. But we also have let me just double check


00:03:29 - 00:04:47
that I'm sharing. Yes, we also have key technical considerations. We have scanner image quality parameters, resolution and color depth and image compression and data formats in digital pathology. I'm going to go back to this um color depth. This this is something important in pathology. We had a podcast episode on that and Monica who's here in the live stream today listening. She was the guest. Um and also Tom Kimpa from Barco. they produce um screens, medical screens and they did extension collar. So that's important.


00:04:08 - 00:05:12
Um then we have digital archiving in digital pathology and they point out in the abstracts I'm going to tell you in a second that um there is there are still requirements uh for archiving of glass slides and that's not going to go away. cloud solutions in digital pathology, image quality control, a big thing that many people are talking about and we're going to have an abstract on that as well. Um, monitoring quality for digital sliced assessment and end user computer requirements in digital pathology.


00:04:40 - 00:05:50
Technical specification like everything right maintenance and calibration of scanning devices, system integration and data management in digital pathology. a correct assignment of microscope slide specimen to virtual images like we all know that uh okay a glass slide can be mislabeled but hey what about mislabeling the pair of glass slide and digital image I did not read about that consideration before approval of hardware and software and digital pathology systems validations recommended validating procedures


00:05:15 - 00:06:42
um so long story short super proud of this publication um not only because it's from Poland and h I think there is value in like new countries embarking on this journey describing h how they're implementing because they already have a lot of resources so not only uh are they describing what are they going to do in Poland uh but also they have access to all the resources of everybody who um came before them and they add additional um consideration, add additional angles from somebody who's newly implementing.


00:06:00 - 00:07:32
I love that. So, highly recommend to go to this one if you are starting your digital pathology journey, if you're starting to implement um and just like the newest publications are going to have the most upto-date basically are going to reflect the most up-to-date reality. Let's see if I did not miss anything from the abstract. from the abstract itself. Um yeah, they present a clear and practical framework um for the effective implementation of digital pathology in routine an anatomic pathology practice. Um and they define


00:06:54 - 00:07:58
digital pathology as digitization of microscope slides into high resolution host images. uh we know that definition and digital systems are becoming increasingly integrated in pathology laboratories. Um but they mention here hey the physical archiving of microscope slides remains a legal and procedural requirement in many countries. uh particularly for hisystological and psychological materials. And um they have this publication with recommendation addressing key technical, organizational and uh legal aspects of


00:07:26 - 00:08:46
digital pathology implementation with an emphasis on ensuring consistent quality and minimizing variability in diagnostic outcomes. um very high value in this type of review publications which are pretty difficult to uh to assemble because every time you write one of these there is more and more to consider. Uh but we sure need them because then somebody else can take them and um use them. And we have somebody from let me show this on the screen. Silver Spring uh Silver Spring, Maryland. That's 6:08 a.m. So kudos. Thank you so


00:08:05 - 00:09:29
much, Sean, for joining us today. Okay, let's go to the next one. Incorporating AI into imaging for surveillance and diagnosis of liver cancer. Um, I'm thinking like this live stream would not be itself if I didn't have some technical difficulties. today and I have a new PDF U annotator and I have this pen that of course is not working on the screen that I'm supposed to use it. So, uh, apologies for me using the mouse and not making straight lines, but I'm going to use it anyway. Uh, this publication


00:08:48 - 00:10:03
comes from France. Oh my goodness, it looks ugly, but it's real, guys. At least you know I'm not AI and I actually like cannot make a straight line with my mouse. I hope it's a little bit entertaining for you. Uh so we have primary liver cancer, hepaticellar carcinoma or interhypatic colangio carcinoma. Uh they remain a leading cause of cancer mortality worldwide. And I think that for every type of cancer there is this statement that that it is the leading um leading cause of mortality. And one time


00:09:29 - 00:10:42
I had like stats which cancer is the most prevalent. Um but obviously every type of cancer is a cancer diagnosis that has to be dealt with. Um but the current surveillance and diagnostic strategies um they rely on so here we're talking radiology at the moment uh this is contrast enhanced CT and MRI and they suffer from operator dependence limited sensitivity and interpretive variability we know that every time a person looks at something and a different person looks at this you have interobserver


00:10:06 - 00:11:24
variability whether it's pathology images radiology ology images or whatever you're doing I guess that you're trying to assess. So that is a limitation of non-computerized ways of diagnosing which were the ways of we were diagnosing forever and we still are but AI offers uh this potential across the liver continuum continuum to um with deep learning and um different modalities now foundation models and whatever comes out and the newest thing that uh keeps outperforming the old thing. Um the deep


00:10:45 - 00:12:18
learning based models have improved ultrasound detection of small liver tumors. Uh they are enabling automated triage and uh reducing worklo workload on CT and MRI. We uh in in this particular application they uh achieved expert level performance for lesion detection, segmentation and characterization. And um the models support standardized interpretation through frameworks such as um lirads and this is liver imaging reporting and data systems. So um sorry now you see it uh and it's so so something to highlight here right


00:11:31 - 00:12:54
more and more we see this oh we are building a model but we are referencing this framework we are operating within these diagnostic criteria so it's not freestyle AI uh anymore it's actually AI aligned with uh the framework that we are using in the diagnostic space which is a step towards integrating this in the diagnostic workflow whenever the lab or or the group or the institution is ready for integrating you don't have to figure out okay how does the AI um model decision AI model um decision support


00:12:13 - 00:13:22
maybe or diagnosis align with what we have been doing visually so far it already does because that's how it was designed And here um they are referencing this liver image reporting and data system and you know there are different systems different diagnostic criteria for different cancers. Um but the highlight here is yeah we are aligning with what we already know is working and um represents current framework. So um interesting things the AI algorithms can distinguish between habitat cellular carcinoma and colangia


00:12:47 - 00:14:03
carcinoma. uh they can classify this plastic nodules and even predict future cancer development from biopsy slides. If I could, I would put a little heart here because now we're entering pathology space. Um and so the advances in foundation models and multimodel AI AI promise to unify radiology, pathology and molecular data. And we are seeing abstracts about that right when there is multimodality even like little multimodality just radiology and pathology or uh just hisystologology hisystopathology and


00:13:25 - 00:14:51
molecular data. So like just pairing um as like labeling one data modality with another data modality that is not text or annotations. H but now more modalities will come into the picture and the more data points we have the more um the better the models are going to be. So for uh me in the pre-clinical drug development space whenever uh I am a pathologist so toxicologic pathologist so I evaluate pathology but um there is this approach called uh weight of evidence approach so we take all the different um available data and we and


00:14:08 - 00:15:27
that's not different than than clinical practice right I'm just relating to my own experience the so the weight of evidence whatever other information we have be it um clinical pathology lab data IHC uh wherever right we take it and this ways into our interpretation um so here that's kind of a parallel to the multimodality now the AI models uh can try to and I'm doing like quotation mark with my fingers h they can interpret the weight of evidence as well. And if you're just joining, let


00:14:47 - 00:15:59
me know. Um what did we say at the beginning? Ah who just joined if you want to go together for the guidelines for adoption of digital pathology from the recent publication um by a Polish group that we discussed at the beginning. Let me know in the chat. Put the word guidelines. Um and if you don't, that's okay as well. So um but of course we have challenges as well. there is no uh digital pathology without challenges. There's no new technology without challenges. So um even though we have the more


00:15:23 - 00:16:53
capabilities with more modalities, we can uh implement the weight of evidence into AI models uh while clinical integration faces major challenges including data privacy, regulatory approval, cost sustainability and algorithmic bias. um and several of these we already covered in previous um previous digipath digest for the algorithmic bias. I'm going to refer you to the seven parts uh AI um in pathology seven parts article series and there is a course that I prepared on this it's called uh pathology AI makeover this you


00:16:08 - 00:17:33
will see a QR code in for me it's the left upper corner uh definitely upper check if left or right for you but in this pathology a makeover if you're new to this uh if you're just starting pathology um and AI exploration. This is a course I prepared that includes um all these all our live streams edited in version audio and uh video and more about AI. So if you're new, I highly recommend you can scan the QR code and see what's in this course. But we have a few more papers yet. So


00:16:51 - 00:18:12
yeah, what are the challenges? Uh we said the challenges and what do we need to overcome these challenges? Well, we need scientific proof. So obviously prospective multic-enter validation studies are essential to confirm clinical benefit and safety. Um and they highlight also okay careful implementation, trustworthy and explainable AI tools. If we have that, we can enable earlier detection and greater diagnostic precision h and more equitable liver cancer care. Uhoh. I was saying uh oh because I hear my


00:17:32 - 00:18:49
kids walking upstairs. I hope they keep sleeping. They are not supposed to wake up until 700 a.m. And I told them specifically, if you wake up earlier, don't come in because I have live stream. Um, I may get off camera if they come in here. So, just stay with me if I disappear for a second. Stay with me. Um, okay. This one is super cool. This the next publication next abstract from slides to AI ready maps standardized standardized multi-layer tissue maps as metadata for artificial intelligence and digital pathology.


00:18:10 - 00:19:35
This is so cool, so simple, so innovative. Um, and this is European uh publication uh group from Gratz, Austria, Czech Republic, Bernal, uh, we have Germany, Shite, and we have Lizaban, Portugal, which I may be visiting pretty soon. Um, and from the authors, I know Norman Zerba. I'm looking if I know anybody else. No, but so uh whole slide image uh images we know what they are. They are high resolution digital images created by scanning the entire glass slide and they are created at the multiple


00:18:53 - 00:20:20
magnification and they are the cornerstone of digital pathology. Right? H but they are also being uh used in neurology, veterary medicine. Um, shout out to all veterary pathologists, hematology, microbiology, dermatology, pharmarmacology, and toxicology. Shout out to toxicologic pathologists. Um, okay. And I see people wanting the guidelines, uh, guideline, um, specific guideline, uh, live stream. We're going to do that then. Okay. Maybe next time we can go through them. um immunology and forensic science. So


00:19:41 - 00:20:47
when we are assembling cohorts for AI training or validation uh we would like to know what's inside these cohorts right um what is the content of uh those whole slide images. So so far it was like opening the image checking what's there and then uh somehow assigning it to your cohorts. There are no standards that currently exist for this metadata. Um so metadata wishful thinking right we would like to have this uh information what's in this tissue in the metadata sometimes we have most of the time we


00:20:14 - 00:21:38
don't have so we rely on manual inspection which is not suitable for large collections with millions of objects. Come on. Okay. My highlighting is imperfect. Um so what they propose what this group proposed a general framework to generate 2D index maps tissue maps that describe the morphological content of whole slide images using common syntax and semantic semantics to archive um excuse me to achieve interoperability between cataloges. What does that mean? they have an AI model that goes into the


00:20:58 - 00:22:11
slide, checks what's on the slide and then includes this in the meta data so that when you look for a specific something in the slide and we're going to say in a second what something you don't have to open the slide because the AI model already uh did that and assign the metadata. So they have three layers. They have source, tissue type and pathological alterations and each layer assigns whole light image segments to specific class providing AI ready metadata. And this AI based metadata extraction


00:21:34 - 00:23:07
from whole light images. Um they used it to generate tissue maps and integrating them into a whole slide image archive. So it enhances search capabilities within the archives and facilitating the accelerated assembly of high quality balanced and more targeted data sets for a AI training validation and cancer research. Um so they used TCGA data set and they um checked it uh basically validated with expert opinions but the AI goes in and says what it is what tissue type and pathological alteration. So it doesn't like go and


00:22:21 - 00:23:39
diagnose uh or gives you a specific diagnosis but it's going to give you the information about neoplastic non-neoplastic and um whatever the the training there was and then it extracts it and puts it into metadata. I think it's so efficient. I just like such a lowhanging fruit for uh the use of AI and empowering everybody who is doing AI research on pathology. Right? So let me know your thoughts. Was it going to be helpful for you if you're a computer scientist working in the space?


00:23:01 - 00:24:28
For sure. Okay. Have a few more. We have two more. Okay, path QC. This is also an interesting one that kind of builds up on uh the the concept that we had in the previous uh publication and the extraction of information from the slide for metadata. Um this is uh for quality control and determining um molecular and structural integrity of tissue from histopathological slides. This is San Diego uh California and we are um talking about application in biioanking. Um quantifying tissue molecular and


00:23:53 - 00:25:19
structural integrity is essential for bio banking and uh current assessment method methods especially for molecular there rely on destructive testing that depletes valuable specimens and rely on manual evaluations. So again not scalable right whenever uh we have to do it manually we cannot scale and pathq uh is a deep learning frame framework that directly predicts the tissue RNA integrity number the ren and the extent of autotosysis from hematoclean and eosin stain hallide images of normal tissue biopsies. This is um


00:24:36 - 00:26:03
in general some something that I see in the biioanking biomarker um development biomarker work space because you often rely on purchase specimen and me personally with my own eyes h I have seen specimen that were paid for and didn't contain any of the tissue uh that was supposed to be there or it was or oral all autoytic, all necrotic or like it was useless basically and uh you had no way of basically checking it before purchasing like block. You can look at the block very rarely. Well, even if they were


00:25:19 - 00:26:33
glass slides assigned to the block, then um there was not an option to actually view them before purchasing. And now in the bio banking space and the viewing of whole slide images of these specimen is kind of a standard procedure. So uh if you use a biioank that doesn't do that just find one that does because then you can at least look at it and see what you're buying. But here they are going one step further because uh they take the hi sorry uh h& images and predict other properties. So


00:25:57 - 00:27:07
um provides pathq provides sample quality control through the direct quantification of the molecular integrity ren and structural degradation autolysis. uh it extracts morphological features from the slide using a recently developed digital pathology foundation model UNI or UNI which we talked about in one of our digipath digests. Uh I should like have prepared the number of the episode so that you can immediately go to the episode. Maybe it's going to be on the screen later. uh but we did talk about


00:26:32 - 00:28:11
this one when we talked about uh foundation models and then we have a supervised model that learns to predict the RNA integrity number and autois score from these morphological features and pathic is trained um and applied to the genotype tissue expression cohort GTX and that's a lot of postmortem samples 25 over 25,000 non-deseased postmortem samples across 29 tissues from 970 donors and then uh they were paired with ground truth rein and ottolesis scores were available. So excuse me then so um they predicted an average


00:27:24 - 00:29:12
Pearson correlation of 47 uh 0.47 47 and uh for the ren and auto score 0.45 and I look at it and I'm like is that good or not? Uh so well it's not that bad because the Pearson correlation goes from minus1 to one. So this is very much on the uh on the close to one side uh of of this whole spectrum. Right? So hey, I take this every day of the week on TWW and twice on Sunday. Uh if this is available versus nothing available, right? Uh so I take this and they um have this available on GitHub.


00:28:17 - 00:29:41
Um so if you are looking huh is it something like you could run independently if a bio bank you're using is not using it maybe you can suggest um but yeah if you're buying specimen from bio banks let's put it in a different way if I was in charge of buying specimen from bio banks I would want to see every single one of them uh or like have some kind of uh computerized AI system uh to let me know okay these are good and flag them somehow so that I can look at the ones that are suspicious um


00:29:02 - 00:30:16
because there's different types of tissue going into bio banks and biomarker this biomarker work relies on this type of tissue and the last publication for today. And before we go into the last one, let me just give you the QR code for the book. If you're just beginning the digital pathology journey, you need this book. And by this book, I mean digital pathology 101. All you need to know to start and continue your digital pathology journey. That's the QR code at the bottom of the screen right now. I'm


00:29:42 - 00:30:57
just going to hide the course. Uh and I am super excited again to tell you that I did advance on the uh new version and we are planning a launch of the new version but everybody who has the old version and the QR code is going to take you to a free download page for the PDF. Um there's an option to have an audio book if you don't want to read it. So you can listen to me reading this to you and there is a new version coming out with updated information with uh stories from my podcast guest. So if you have


00:30:20 - 00:31:29
this book, you're going to be on my mailing list. You're going to have access to everything I put out together. And the next conference that I'm going to is going to be US Cap. So if you're going, let me know in the chat as well. And then we can plan some kind of maybe some kind of meeting space or something where we can just talk because uh in the last year when I was going to conferences I like go with my camera then go work with sponsors. And by the way hopefully next week at the pre


00:30:54 - 00:31:56
Valentine uh episode I'm going to reveal who's going to be the main sponsor for US COP. So I'm definitely going to be at their booth but it's still a secret. We're still working out the details. Um, but in addition to that, or maybe I can work with them and figure out a time where we can meet, maybe you record something together like digital pathology trailblazer conversation and I'm going to be letting you know what that's going to be. But if you want to be part of it, let me know that


00:31:26 - 00:32:43
you're going to Yuskap and I will make sure that you know where I'm there and and when we could meet. So, the last paper, my trailblazers, get the book if you don't have the book. And if you have a friend that doesn't have the book, send them the link once you scan the QR code. Okay, I'm going to hide the code for a second. This is a review machine learning and biomarkerdriven precision oncology. Um, which basically is how we do oncology. Now it's biomarkerdriven precision


00:32:14 - 00:33:44
oncology and imuninohistochemistry uh plays a crucial role in this precision oncology precision medicine um parading. So here automated imunostic chemistry scoring and emerging direction in genito urinary cancer. So we are now talking specifically about uh GU cancers and this is a group from Canada. Where is it? Um immuninohistochemistry is essential for diagnostic prognostic and predictive biomarker assessment in oncology. Manual interpretation. We've heard that like in three or four out of these out


00:32:59 - 00:34:19
of these abstracts already is limited by subjectivity and interobserver variability but we have our tools that can help machine learning and AI and this uh allows algorithms to recognize pattern and learn from annotated data sets or basically paired data sets. Now we have different uh ways of um giving AI the ground truth. We have different modalities and um this automated quantification of biomarker expression on whole slide images is like the first application of digital pathology first application of


00:33:39 - 00:35:08
image analysis that actually made it into the diagnostic space. H it used to be based on um on hard hardcoded features or handcrafted features. Now it's AI based. uh but it has been there for uh quite a while and this review evaluates the role of machine learning assisted IHC scoring h in the transition from validated biomarker to discovery and emerging prognostic and predictive IHC biomarker for GU tumors. Erh and um these routinely used uh that you probably have seen and see all the time is ERPR her 2 and mismatch pair MMR


00:34:23 - 00:35:56
proteins uh mismatch repair proteins MMR uh PDL1 K67 um but they are also um including androgen receptors P10 cytoins europlain 2 nectin for an immune checkpoint point proteins and then um they also examine early evidence indicating associations between the machine learning derived metrics and clinical outcomes which is what we want to have with these tests. Okay. Are these tests uh related in any way to clinical outcomes and can give us outcome information? And we also have limitations. Well, one of them is limited


00:35:10 - 00:36:22
availability of training data sets. H variability and staining protocols and regulatory challenges. I don't know this like blanket statement of regulatory challenges. I think there are like well doumented regulatory pathways that you have to go through and they are challenging because regulators challenge you. So maybe uh that's what they mean regulatory challenges. Yeah, if you want to go into the regulated space regardless where you are, they're going to make it difficult because uh they want to make sure it's uh


00:35:48 - 00:37:07
working well and you can safely deploy it uh in patient care context. Um but overall machine learning associated IHC scoring is a reproducible and evolving approach that may support biomarker discovery and enhance precision in geooncology. I remember so that was for a research setting that I was doing it because I'm a veterary pathologist. So um I couldn't do it u in a diagnostic setting. I I could for animals but PDL1 is not necessarily a marker that is being used in animals. But I was doing this PDL1


00:36:26 - 00:37:54
scoring in a research setting and then I was also developing an algorithm a deep learning based algorithm uh to to to score it right and you've probably heard me like I sound like a broken record but maybe somebody is here for the first time. So uh I was like okay so me visually was guesstimating the percentage of uh cells that were uh positive for PDL1 in the tumor and then um me well in with the team developing a tool that is quantification tool that was not based on like single cell but was based on


00:37:12 - 00:38:29
area of tumor and then we were comparing this uh to each other and me gueststimating was the ground truth. So uh now we have better sources of ground truth and we should should look for better sources than um pathologist guesstimation and this is like a trigger like when you but but that's how we were evaluating IHC and how you we still are and in many instances because digitization as much as we talk about it and if you're here at 6 am you basically probably love it as much as I do the um


00:37:50 - 00:39:11
it's not present everywhere, right? Uh so um lost my train of thought. Um okay, so guesstimation. Yeah. So so this is because the digitization in the lab is only maybe like 10 depending on the country 15% of labs. Some countries even more like in the Middle East I think it's like 50% in some places. um but it differs very much and it's definitely not 100%. So these uh things that we're discussing are um rare events in the diagnostic space right even though I am a big proponent of them the majority of IHC evaluation


00:38:31 - 00:39:45
is going to be visual um which let's always put it in context like better to have a visually interpreted IHC than not interpreted IHC better to have a computationally uh supported IHC interpretation than visual whatever whatever you have, whatever is the best thing you have access to, do that and um like provide the best care you can. Thank you so much for joining me. Thank you for scanning the code to the book. I'm going to also include it in the show notes. If you're interested in going through


00:39:09 - 00:39:53
the guidelines, h I already heard uh some some voices of yours to go through the guidelines together through the new publication from Poland, which I'm going to be super excited to be going through a publication from Poland. Let me know in the chat, even if you're watching the recording. Thank you so much for joining me and I talk to you in the next episode.