Digital Pathology Podcast

117: Tertiary Lymphoid Structures in Colorectal Cancer Prognosis | Dr. Aleks + AI

• Aleksandra Zuraw • Episode 117

Share episode

Send us a text

Leveraging AI for Deep Insights into Tertiary Lymphoid Structures in Colorectal Cancer

In this episode of the Digital Pathology Podcast, I introduce 'Aleks + AI,' a new experimental series leveraging Google's Notebook LM to delve deeper into scientific literature.

Today's focus is on tertiary lymphoid structures (TLS) and their potential to predict colorectal cancer prognosis. We discuss a study published in the October 2024 issue of Precision Clinical Medicine, exploring different methods of quantifying TLS using digital pathology and AI. 

The paper title is: "Comparative analysis of tertiary lymphoid structures for predicting survival of colorectal cancer: a whole-slide images-based study"

The findings highlight TLS density as a reliable predictor of survival and its correlation with immune responses and microsatellite instability. We also touch upon the potential for AI to streamline TLS analysis in clinical settings and the broader implications for personalized medicine. Join us as we dive into the intersection of digital pathology and computer science, featuring insights and commentary from my AI co-hosts, Hema and Toxy.

00:00 Welcome and Introduction
00:45 Introducing the New AI Tool: Notebook LM by Google
01:11 Experimental Series: "Aleks + AI"
02:06 Deep Dive into Tertiary Lymphoid Structures (TLS)
03:18 Understanding TLS and Their Role in Colorectal Cancer
04:20 Quantification Methods and Key Findings
05:02 Implications for Personalized Medicine
09:02 AI in TLS Analysis and Future Prospects
11:00 CMS Classification and TLS Density
12:08 Study Limitations and Future Directions
15:40 Final Thoughts and Wrap-Up
16:28 Feedback and Future Plans

THIS EPISODE'S RESOURCES

PUBLICATION DISCUSSED TODAY

📝 Comparative analysis of tertiary lymphoid structures for predicting survival of colorectal cancer: a whole-slide images-based study
🔗https://academic.oup.com/pcm/article/7/4/pbae030/7826772


📱 Send us a text

Support the show

Become a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!

Aleksandra:

Welcome my digital pathology trailblazers. if you've listened to the last DigiPath Digest, which is the previous episode, just go one back if you haven't. in the DigiPath Digest, we have it weekly and we already did it for 18 weeks I'm super proud of us because every Friday morning at 6 a. m. in Pennsylvania and different strange hours in the world. You show up and we discuss abstracts. But the thing is, and that was the feedback I got at the beginning. In the beginning, I said, no, I cannot go deeper. It's more for skimming the trends, checking what's out there. I cannot go deeper into every single paper because I just don't have the bandwidth. I would love to, but I can't. But now there is a new AI tool that lets me do it. And the tool is called Notebook LM by Google that can generate podcast episodes from PDFs. So the idea in the last DigiPath Digest live stream was, Oh, what if I could generate those deeper podcasts with AI. Listen to them and provide a intro, outro and commentary. So I decided to do this experimental series. It's going to be called"Aleks AI" and I will use notebook LM by Google to generate podcast episodes about the latest in scientific literature. So we're going to still have DigiPath digest. be on the lookout for our live streams. if we identify some good papers that we cannot go deep into during the live stream. That's what's going to come out in the following week before the next live stream. And I always provide interos, outros, commentary to keep things clear and engaging and transparency is key here. And transparency is key here. Whenever, new tools come out, I'll show exactly how the content is generated and rest assured, I listen to everything to stay fully in the loop. So whatever my AI hosts are speaking about, it's going to be vetted by me. So let's explore the science and innovation together. Today's episode is about tertiary lymphoid structures in colorectal cancer prognosis and how AI can be leveraged for that. So let's dive into it.

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. Aleksandra Zuraw.

Toxy:

Thanks for the intro, Dr. Alex, and welcome everyone to today's deep dive. We're going to be looking at a fascinating paper. All about tertiary lymphoid structures.

Hema:

You can call them TLS for short.

Toxy:

Okay, TLS. And their potential in predicting colorectal cancer prognosis.

Hema:

Yeah, it's a really interesting area of research.

Toxy:

The study is titled Comparative Analysis of Tertiary Lymphoid Structures for Predicting Survival of Colorectal Cancer. A whole slide, images based study. And it was published in the October 2024 issue of Precision Clinical Medicine.

Hema:

That's a pretty solid journal.

Toxy:

It is. It has an impact factor of 8. 2. So before we jump to the specifics, could you just give us a quick overview of what TLS are?

Hema:

Sure. Imagine like, um, organized clusters of immune cells. Okay. That form within tissues outside of those traditional lymph nodes.

Toxy:

So not the typical spots.

Hema:

Right. And they act kind of like these mini command centers. It's coordinating the immune response.

Toxy:

Against threats.

Hema:

Yeah. Threats like cancer cells.

Toxy:

So they're like these little pockets of immune activity.

Hema:

Exactly. And the researchers in the study wanted to investigate if we can use those TLS.

Toxy:

To predict how a patient with colorectal cancer is going to do.

Hema:

Exactly. It's all about personalized medicine.

Toxy:

They analyze a huge amount of data for this,

Hema:

1, 600 patients from four different hospitals, plus they use data from the Cancer Genome Atlas.

Toxy:

Wow. And they didn't just look at the presence or absence of the TLS.

Hema:

Right. They tested three different ways of quantifying them.

Toxy:

Using digital pathology.

Hema:

Exactly. They were working with HE stained whole slide images.

Toxy:

That's really comprehensive. It

Hema:

is.

Toxy:

So, so what were those three quantification methods?

Hema:

So first, they just counted the number of TLS within the tumor's invasive front. The second way was measuring the diameter of the biggest TLS.

Toxy:

And the third? The

Hema:

third one, they calculated the density of TLS per millimeter of tumor edge.

Toxy:

That third one, density, was the most interesting finding, right? It was. What was that?

Hema:

The research found that TLS density was the most consistent and reliable predictor of survival.

Toxy:

Across all of the data sets?

Hema:

Across all of them. Even after controlling for other things like age and tumor stage.

Toxy:

It's really interesting. So it's not just about having them.

Hema:

Right.

Toxy:

It's about how densely packed they are.

Hema:

Yeah, a higher density, you can almost think of it like a bigger army, a more organized army of immune cells fighting the tumor.

Toxy:

And that could really help guide treatment decisions, right? If we know the density.

Hema:

Definitely. If we identify patients with a high TLS density early on,

Toxy:

we could tailor their treatment plans. Yeah,

Hema:

right. Maybe choose therapies that boost that immune response.

Toxy:

Or monitor them more closely.

Hema:

Exactly. This is

Toxy:

really exciting to see how digital pathology is being used. for these insights.

Hema:

It really is. It's like we have this window now into the tumor and the immune system and how they interact.

Toxy:

I know the study also looked at specific types of immune cells.

Hema:

They did. Higher TLS density was correlated with more CD3 plus and CD8 plus T cells within the tumors.

Toxy:

Could you give us like a quick breakdown of what those T cells do?

Hema:

Of course. So CD3 plus T cells, they're a broad category. Involved in a bunch of immune responses, CD8 plus T cells are more specialized. They're like the soldiers.

Toxy:

The ones that go in and attack.

Hema:

Exactly. They target and kill those cancerous cells.

Toxy:

So more of those T cells, especially the CD8 plus ones, the killers, in tumors with higher TLS density, kind of adds to the idea that TLS density shows us a strong anti tumor response.

Hema:

Exactly. It's not just about having those TLS. It's about how many immune cells are there and how active they are.

Aleksandra:

Quick comment here, the density. So in this case, it's TLSs, but the density, being relevant reminds me of immuno score. And in general, the research I did for immuno oncology, the density of particular lymphocytes or particular T cell types was always a feature that we were quantifying with image analysis.

Toxy:

This is all really interesting and has some big implications. But I know there's even more to unpack here.

Hema:

There is. Especially when we look at MSI status and how AI could be used for TLS analysis.

Toxy:

Those are definitely topics that deserve their own conversation. We'll dive into those after a short break.

Oh my goodness, my AI host gave me a specific place in the podcast to put a break. So let me just invite you to download the"Digital Pathology 101 book". It's free on my website, digitalpathologyplace. com. So if you have not grabbed it yet, please do. In January, I'm going to be updating the AI chapter. So if you grab it now, you will get the updated version automatically.

Toxy:

Welcome back, everyone. Before the break, we were talking about TLS density and how it could predict survival in colorectal cancer.

Hema:

Yeah, and this research goes even further. They found a connection between TLS density and something else we use a lot in colorectal cancer

Toxy:

microsatellite instability

Hema:

right or MSI status

Toxy:

And we know that MSI high tumors tend to respond really well to immunotherapy

Hema:

They do and what's really interesting is this study found higher TLS density in those patients with MSI high tumors

Toxy:

So there could be a link between TLS density and how well someone might respond to immunotherapy. Exactly. That's amazing. It could really change how we approach treatment.

Hema:

It could. Imagine being able to look at a patient's tumor tissue and know right away if they're likely to benefit from immunotherapy. That's the kind of personalized care we want to be able to provide.

Toxy:

Definitely. Now going back to those three methods the researchers used to quantify TLS Why was it important to try a few different approaches?

Hema:

Well, it shows just how thorough they were, right? They wanted to find the most reliable method. Because just counting the TLS might not be enough. You could have a small tumor with a few big TLS, or a large tumor with lots of tiny TLS.

Toxy:

The density gives us a more, uh, standardized measurement.

Hema:

Exactly. It takes both the number and the size into account.

Toxy:

Giving a more accurate picture.

Hema:

Right, of that immune landscape within the tumor.

Toxy:

And the fact that density turned out to be the strongest predictor of survival across all those data sets.

Hema:

Really reinforces that idea.

Toxy:

And digital pathology is really making this kind of in depth analysis possible.

Hema:

It is. It's like we have this high resolution map of what's happening in the tumor.

Toxy:

You mentioned AI earlier. I did. How could AI be used in TLS analysis?

Hema:

Well, right now, pathologists have to identify and quantify those TLS manually.

Toxy:

Which I imagine takes quite a bit of time.

Hema:

It does, analyzing all those slides, but what if we could train an AI algorithm to do it for us?

Toxy:

That would be much faster.

Hema:

So much faster and potentially even more accurate.

Toxy:

Taking out that element of human error.

Hema:

Exactly, and it would make this kind of analysis more practical for everyday clinical use. We could look at even bigger data sets, potentially leading to even more discoveries.

Toxy:

It would be amazing to be able to bring these findings to patients more quickly.

Hema:

It would. And there was another interesting finding in this study that I wanted to mention. It has to do with the spatial distribution of the T cells, specifically the CD3 plus T cells.

Toxy:

Oh, that's interesting. What did they find?

Hema:

They found that in the tumors with a higher TLS density, there were more CD3 plus T cells, not just in the core of the tumor.

Toxy:

But also in that invasive margin.

Hema:

Exactly. Where the tumor is growing and pushing into the surrounding tissue.

Toxy:

So it seems like those TLS are influencing immune cell activity throughout the tumor.

Hema:

It seems that way.

Toxy:

Throughout.

Hema:

But here's where it gets a little unexpected. They didn't see a significant difference in the density of the CD8 plus cells, those killer T cells, between the high and low TLS density groups.

Toxy:

Oh, that's surprising. Do you know why that might be?

Hema:

We don't for sure. Maybe the CD8 plus cells don't rely on TLS as much for their activity. Or maybe TLS have a different role in how they regulate the CD8 plus T cells compared to the CD3 plus T cells.

Toxy:

It just shows how complex the immune system is.

Hema:

It really does. We still have so much to learn. This study gives us a glimpse into this back and forth between the tumor and the immune system. But

Toxy:

it also shows us how much more there is to explore.

Hema:

Definitely. And digital pathology is giving us the tools to do that. We can visualize and analyze that tumor microenvironment in incredible detail now.

Toxy:

I know the study also looked at how TLS density relates to the consensus molecular subtype, or CMS, classification for colorectal cancer. Can you explain what the CMS system is?

Hema:

Sure. So the CMS system sorts colorectal cancers into four main subtypes based on their molecular features.

Toxy:

So it helps us understand how the tumor might behave.

Hema:

Right. And how it might respond to different treatments.

Toxy:

Another step towards personalized medicine.

Hema:

Exactly. And in this study, they found that the different CMS subtypes weren't evenly distributed between the high and low TLS density groups.

Toxy:

So certain subtypes were more common in one group than the other.

Hema:

Yes. They saw a big difference in the number of CMS1 tumors.

Toxy:

Were there more CMS1 tumors in the high TLS density group?

Hema:

There were. And those CMS1 tumors are known for having a strong immune response. So it fits with the overall findings.

Toxy:

It does. Maybe TLS density is especially useful for predicting outcomes for patients with CMS1 tumors.

Hema:

It could be. You know, this study really suggests that TLS density could be really important for making decisions about diagnosis, prognosis, and treatment.

Toxy:

It does. But it's important to remember that every study has limitations.

Hema:

Right.

Toxy:

What are some of the limitations we should keep in mind with this one?

Hema:

Well one big one is that it was a retrospective study, meaning they were looking back at data that had already been collected. It would be great to see perspective studies where they follow patients over time and see how TLS density changes.

Toxy:

Especially to see how TLS density is affected by treatment.

Hema:

Exactly. We could see how TLS density changes as the cancer progresses, too.

Toxy:

And of course, there's still that potential for AI to really revolutionize this. Oh,

Hema:

definitely. Imagine being able to identify different subtypes of TLS automatically. That would give us even more detailed information about the immune system's response.

Toxy:

It's incredible to think about all the possibilities. This research is really paving the way for some exciting advancements in digital pathology.

Hema:

It is. It shows how much potential there is for digital pathology to change how we diagnose and treat cancer.

Toxy:

This has been a really fascinating conversation.

Hema:

It has.

Toxy:

But there was one more thing I wanted to touch on before we finish up. The researchers in this study didn't differentiate between different subtypes of TLS.

Hema:

That's true.

Toxy:

What are your thoughts on that? Do you think looking at those subtypes could be helpful?

Hema:

It could be really valuable. We know from other research that there are different types of TLS, and those different types might have different roles in the immune response.

Toxy:

So lumping them all together might have hidden some important details.

Hema:

It might have. Maybe some TLS subtypes are better at predicting prognosis or treatment response than others.

Toxy:

That'd be fascinating to find out.

Hema:

It would, and it could lead to even more targeted treatments for patients.

Toxy:

This deep dive has really opened my eyes to the incredible potential of digital pathology.

Hema:

Mine too. We're just starting to understand all that TLS can tell us about the relationship between the tumor and the immune system. It's such an exciting field.

Toxy:

It is. This has been a truly insightful discussion. I'm sure our listeners are eager to learn more about this. Dr. Alex, we've gone deep into the fascinating world of tertiary lymphoid structures and their potential as a tool in colorectal cancer. Back to you for any final thoughts.

Hema:

Wow. That was a really deep dive into TLS. I think our listeners have a much better understanding now of how important they could be in colorectal cancer.

Toxy:

I agree. And this study was a great example of that, really showed the potential clinical impact.

Hema:

Definitely. You know, one thing I was thinking about was, uh, how TLS density could even change how we stage colorectal cancer. We talked about it a bit before, but I was hoping you could elaborate on that. Sure.

Toxy:

Right now, we mainly use the TNM system.

Hema:

Which looks at things like tumor size and lymph node involvement.

Toxy:

Right, and metastasis.

Hema:

But it doesn't tell us everything.

Toxy:

No, it doesn't really give us any information about the patient's immune response. And we know that's so important, especially when it comes to how the cancer progresses and how well someone responds to treatment.

Hema:

That's where TLS density could really come in. It could. It gives us a way to measure the immune activity, you know, quantitatively within the tumor. And that could add a whole new level of detail to our staging systems.

Toxy:

So you could have two patients, same TNM stage, but totally different TLS densities.

Hema:

Exactly. And that could mean totally different prognosis, maybe even different treatment approaches.

Toxy:

We could identify patients who might need more aggressive treatment right away.

Hema:

Or closer monitoring, even if their TNM stage says they don't.

Toxy:

It's all about personalization.

Hema:

It is, and I think this research is a big step towards making that a reality.

Toxy:

It's been such a fascinating conversation. We've learned so much about TLS, how they work in colorectal cancer, and what this all means for patients. It

Hema:

has been really interesting.

Toxy:

But before we wrap up, I want to leave our listeners with one last thought. This study really shows how important it is for different fields to work together. We have pathologists, oncologists, immunologists, computer scientists, all contributing to these advancements.

Hema:

I totally agree. It's amazing to see what we can accomplish when everyone collaborates.

Toxy:

Absolutely. I'm excited to see what the future holds for this kind of research, and I'm sure we'll see even more effective cancer treatments as a result.

Hema:

I'm excited too. It's such an incredible time to be working in cancer research and digital pathology.

Toxy:

Well, thank you so much for joining us for this deep dive. It's been a pleasure discussing this with you. Dr. Alex, back to you to wrap things up.

Aleksandra:

Thank you so much to my AI co hosts and to you, my digital pathology trailblazers for listening to this experimental episodes. There are a few coming out. Let me share with you what I think. So let's start with the content. Good. I like it. It is an easy way to learn about literature, specifically when everybody is so busy. There is no time for anybody who is a parent, has a full time job or any kind of full time activity to sit and read papers every day, unless this is part of your job, right? My academic digital pathology trailblazers, you might be a little annoyed with, the form of this. But for me, I'm probably in between, you know, I read enough papers, but this is not part of my job description. So to me, this is useful. In parallel, I'm still using a software called Natural Reader to read papers to me. And obviously it depends how you consume information, if audio information is for you, but I am assuming yes, because you're listening to the audio version of this podcast. So anyway, I had this software, I'm going to link to the YouTube video about it, Natural Reader, that would read papers to me. There's always this hierarchy level of how difficult it is to do something versus the benefits. So like, uh, for runners, any run is better than no run. Along those lines, any type of scientific literature reading is going to be better than none. That's why I started with DigiPath Digest as the lowest barrier to entry just reviewing abstracts. this is going to be my next step of actually I can dive a little bit deeper into the full paper. Obviously the next level is going to be to actually read this paper in full. and highlight and do the normal stuff that you do with papers when you analyze them. And I know that whenever we have the bandwidth to do it, we will do it for the papers that are relevant. But this is a super cool informational kind of screening tool. So that's how I see it. I like the content. Regarding the form, there is going to be a little bit of getting used to it, I think. I named my AI co hosts, the female voice is Hema and the male voice is Toxy, Hema-Toxy. I had different names for them in the previous experiment, so I don't guarantee they will stay the same. But for now, it's Mrs. Hema, and Mr. Toxy. There's gonna be getting used to it, like there was for me, getting used to the automated voice of the natural reader. But again, for me, better to listen to this and get the information from the paper in this form than not. And I also think this is going to be good for the digital pathology trailblazers outside of the academic environment that are not native to scientific literature review. So for you, this is beneficial because my AI co hosts use this. conversational easy language. But when I was prompting notebook ML to create this episode, I said, my audience is an audience of professionals use technical language focused on histopathology. So I gave them instructions. I didn't give them the instructions to let me have a break or a plugin in the middle of the episode. So that was cool surprise. But basically my AI co hosts already know who you are that it's not just a lay person, but people who are already interested in digital pathology. But if you just came across this podcast and maybe you're a patient diagnosed with colorectal cancer, the information is going to be useful to you as well, which I love. This year I heard this quote, Never underestimate your audience's intelligence, but never overestimate their prior knowledge. I thought this was a fantastic guideline for educational content creation because you are highly intelligent, smart digital pathology trailblazers. And the only difference is that you might not have the prior knowledge that I have or the authors of the papers have. Other than that, of course, let me know what you think about this. Do you like it? Is it annoying? Does it require some getting used to? Whatever you think about this. So this particular episode is only an audio. There's no video with it. Later I will figure out if there is a streamlined way to create video. I might create an AI avatars for my co hosts for YouTube. For now it's just on the podcast. And when you listen to the podcast, there is an option to send me a text. In the show notes, you're going to see the option to send me a text. Use this option, it's going to convert the text into emails, and I will be able to get your feedback like you don't have to go to any other app. I mean, if we're connected on LinkedIn, feel free to send me messages there. If you are on my email list, you can email me anytime. But if you're just listening to the podcast, and that's your primary way of learning about digital pathology, there is an option to send me a text. So send me a text with the feedback. I will love it. I will incorporate it in the future episodes and I talk to you in the next one.