Image Analysis Enhances Pathology Evaluation of Preclinical Studies, not Replaces it

In this episode of the Digital Pathology Podcast, I sit down with Danielle Brown, the General Manager at Charles River Laboratories and an expert in toxicologic pathology. Danielle and I explore the evolving role of image analysis in preclinical studies, diving into how this technology enhances—rather than replaces—the expertise of pathologists.

We discuss the challenges and benefits of integrating image analysis into preclinical research, particularly in a GLP-compliant environment, and the importance of collaboration between pathologists and image analysis scientists.

• The limitations of visual analysis,
• The efficiency and consistency of image analysis compared to traditional methods, and
• The future of AI-driven pathology. 

Danielle also shares insights from her extensive experience in managing image analysis teams and how Charles River Laboratories ensures regulatory compliance in their processes.

This episode is packed with valuable insights and practical advice for anyone involved in preclinical research or interested in the integration of image analysis in pathology. Danielle’s expertise and our discussion provide a roadmap for leveraging technology to improve diagnostic accuracy and efficiency.

📄The paper Aleks and Danielle co-authored: “Developing a Qualification and Verification Strategy for Digital Tissue Image Analysis in Toxicological Pathology” 

📄 Learn more about GLP-compliant tissue image analysis at Charles River Laboratories

Be Part of the Pathology Evolution: Stay informed on the latest in digital pathology innovations. Subscribe for more insights, become a member of the Digital Pathology Club, and get your complimentary copy of “Digital Pathology 101“. Embark on your path to discovery and progress in the fascinating world of pathology.

• Digital Pathology Club Membership
• Free Digital Pathology 101 eBook

Danielle: [00:00:00] So usually the image analysis scientists will work on the algorithm when they think it’s looking good. The pathologist will kind of review it. The pathologist can also provide some annotations where the, you know, to help train the algorithm in the beginning.

So it’s, it’s a partnership all the way through. So the pathologist is involved in the beginning to help. Help the scientists understand, you know, what are the areas we’re looking for? What are the nuances there? And then when the algorithm is ready or when the scientist thinks it’s ready, the pathologist will need to verify it.

Look at, you know, a set of run through a set of, slides with maybe some differences, the range of what you’re going to see of staining and make sure that it’s performing,  adequately.

Aleks: Welcome my digital pathology trailblazers. Today, I have personally special guests. Today, my guest is Dr. Danielle Brown, and she’s personally special to me because she is my mentor at work at Charles River [00:01:00] Laboratories, she is also a leading expert in toxicologic pathology and toxicology with over a decade of, or I don’t know, maybe over 15 years or how many years, a lot of years, more than me.

Of experience in stereology and image analysis, she authored a lot of publications, and she spoke at a lot of international conferences on this topic.  And I was honored to co-author a publication with her as well. And as I said, we work together at Charles River Laboratories. So welcome to the podcast, Danielle.

And we always start with the guests.

Dr. Danielle Brown’s Background and Career

So let’s start with you and let’s tell the digital pathology trailblazers who you are. What’s your background? And, you know, your current role is a little different than the image analysis background that you’re famous for. Tell them about you. 

Danielle: Yeah, thanks for having me.  It’s great to be able to chat with you and all the listeners.  Yeah, so I started as a veterinary pathologist in 2009, so it’s been about [00:02:00] 15 years. And my first job was, was to get an image analysis lab kind off the ground. At the position I had as well as an immunohistochemistry lab. And so I really have done.

Quantitative pathology my whole career starting right out of my residency and even in my residency. I did a lot of it, which is why I started in that position. So,  and then over time, I grew the, the lab and the specialty area and went into kind of a management side of things because I had several image analysis scientists reporting to me. 

Aleks: You did too well in that role. So… 

Danielle: Yeah, I, 

Aleks: …you move on.

Danielle: I grew the area, the specialty area. So I was, managing several people and then it kind of increased over time. I was managing pathologists and then managing a smaller site for Charles River. And now my newest role, well, I’ve been in it for two and a half years now, which doesn’t feel, it doesn’t feel [00:03:00] like I’ve been in it that long, but I’m the general manager for the Reno site for Charles River, which is a large site about 700 to 750 employees.  And we do a lot of different, areas of toxicology and pathology. So I don’t, I, you know, I don’t. I don’t read slides anymore now that I’m, in the higher level management, but I counsel on scientific, you know, give scientific advice and counsel on study designs and things like that still.

Role of Image Analysis in Preclinical Drug Development

Aleks: So,  let’s, because the topic is image analysis, I want you to tell me why, what is the role of image analysis in preclinical drug development research?

Why should we even care? And the, you know, the diagnostic role is kind of more famous because, you know, diagnostic tools and there was a 510k approval for a particular algorithm. But what is [00:04:00] it for in preclinical? Why should we care? I care. But why should other people care? 

Challenges and Limitations of Visual Analysis

Danielle: Yeah, sometimes, you know, the human eye has limitations, right?

So, as pathologists, we’re great at noticing patterns. Pattern recognition is something the human eye is great at, but what the human eye is not so great at is quantitative differences. That’s because our eyes can be tricked quite easily. So there’s some famous papers out there that really kind of show the different kind of biases and I’m sure everyone’s, you know, taking some of those visual, you know, puzzles or things where it tricks your eyes and you think something’s moving or you think something’s bigger than something else. 

Aleks: So, how many circles when you only see squares and then the moment you see circles, you see the circles.

Danielle: Yeah, yeah. So there’s, there are some limitations there and sometimes we, you know, there may be really subtle differences between treatment groups or there may be, you know, In the earlier research side [00:05:00] of things, you may want to quantify and an endpoint has an immunohistochemical stain, or you want to quantify the number of cells or anything else that, you know, as far as a human eyes concern, it might be a little bit difficult to differentiate, or you may have some bias that comes in, not on purpose. It’s just happens because of the way that our eyes work. So, you know, I think it’s really important for those subtle differences, or when a certain quantitative end point is really important for, you know, the outcome of the study, or really understanding, you know, what you’re trying to look at. 

Aleks: Definitely. And, I mean, because I also have the image analysis background, I, I’ve heard questions like you are experts, you’re a pathologist. And why is it not enough for a pathologist to, to just look at it and state, and you just basically answered my second question, which is like, okay, if it’s,  like, [00:06:00] if estimates that are like clearly visible.

You don’t have to quantify. Yes, you can look at it and you can see this is less. This is more.  when the difference is really pronounced, it’s easy, but, often, it’s not that pronounced. And I think, you know, the puzzles that we mentioned, I don’t think people apply. It’s just like a curiosity, a fun thing.

Oh, we got tricked. But actually,  you can get as tricked in research and specifically when you’re trying to analyze immunohistochemistry stains. So,  let’s talk about that. Like, especially. I would like, because a task that is often,  asked, of pathologists is like, Oh, quant, quantify the intensity.

But we know that quantifying intensity visually is first of all, not that easy. And second, we can be tricked. Yeah. Can you specify this, that this trick that applies here? 

Danielle: Yeah. So when you stain,  I mean, even with any kind of stain, but especially with an immunohistochemical marker where, you know, you’re looking, you know, maybe it’s a dab stain.  So you’re looking at brown staining, your eyes can be really tricked and, and just how much intensity of staining there is, or maybe there’s background staining, and it’s really hard to differentiate that from real differences. So, for example, you know, we always mix up staining runs, right? You never want to run all of your high dose or you know, treated animals or something like that on one. But,  even if you do that, you’ll still have some differences that your eye may have a difficult time picking up whether it’s real, whether it’s background. So, for example, I have a slide that I show in some of my talks, and it’s a stereology study, but it’s the same type of thing. It’s [00:08:00] quantification, right? And look, it’s in the brain of mice. One of the sections, one of has a much darker, stain on it than the other one, and looking at it, you say, well, that one has more staining or more,  positive cells in that region of the brain, but in reality, when you quantify it, it’s, it’s less, it’s a lot less.

So our, our brain or our eyes see something, but it can really be due to staining variations, which is a background that are, it’s really difficult to weed out that background. But when you train a computer, you can train the computer on the entire range of staining. And then the, the computer can actually pick out, you know, what is real and what is not real a little bit better.

Importance of Pathologist and Image Analysis Collaboration

But you really need a pathologist to recognize the change in the first place, right? It’s not going to, it doesn’t take over what the pathologist role is  The [00:09:00] pathologist is still very important to identify changes. And then the image analysis is kind of an adjunct tool to augment the results and really put an objective quantification to it. 

Aleks: Yeah, let’s let’s elaborate on that a little bit because often it’s like, Oh, if I have the pathologist, why do I need image analysis? Or if I have image analysis, why do I need the pathologist?  the reality is like what you just said. You need both,  and like the simple way is, okay, you need the amount in image analysis to quantify, but you need the pathologist to know where to quantify.

Danielle: Yes. 

Aleks: Did you have, example, or do you recall any examples or situations where,  like the image analysis was top notch, but it was like from totally the wrong place on the slide or the other way around the [00:10:00] pathologist did the best they could, but actually they were visually or like neurobiologically not able to quantify?

Do you have any examples of these situations? 

Danielled: Yeah. So when I was managing, you know, when I was managing image analysis scientists, it was really important for me to work closely with them to make sure, because they weren’t pathologists,  they’re very good at what they did. Very good at creating algorithms, very good at,  you know, she’s seeing those, but as a pathologist, you know, my role was to make sure that they were looking in the right places and understood, you know, is the stain supposed to be nuclear or cytoplasmic?

So understanding what part of the cell should be stained to help differentiate real staining versus background, or I did, I’ve done a lot of neuropathology and with the neuropathology, you know, you have different areas of the brain. You may want to look at, and sometimes you have a stain. That will stain in several areas, but you only want to quantify in one.

So, helping them understand [00:11:00] the anatomy,  and where to draw, you know, regions of interest. is very important as well. So it’s really a partnership, you know, between the pathologist and, you know, the image analysis scientist. If the pathologist isn’t the one running the algorithms, which we often have support staff to help us, it’s really a partnership.

And the pathologist take plays a key role in that. And I think that’s why it’s really important to have an image analysis. laboratory that also has pathologists on staff. 

Aleks: Yeah, I think it’s a key component. Me personally, because I was part of such team as well, even before joining Charles River, and now I am also occasionally help out with those projects.

I don’t believe these projects we did were a pathologist didn’t look at it because, you know, simple things are okay. A stain was quantified in an area of necrosis or background stain was quantified instead of [00:12:00] like a targeted stain, but there are also a lot more nuanced things. So one time I was part of a project where,  the stain was very specific in the samples where,  there was lymphoid tissue where the stain was developed and the stain was staining the lymphoid cells and that was what we were looking for.

And then in the tumor samples that were analyzed the same stain, stained, very plump endothelial cells. And the image analysis, the algorithm was like very precisely quantifying the endothelial cells. And if a pathologist wouldn’t look at it.  Nobody would know that these are not the cells that we are looking for.

And it turned out later after some research that there are instances where this marker can also stain endothelium. We ended up not generating data on this particular [00:13:00] tumor because it was not what we were looking for. Even though the stains, the cells were perfectly stained, the quantification was like really good.

The pathologist said, “No, this is not what we’re looking for.” And, they ended up scoring it visually, which I think is also. an advantage of a project like this because this was a shortcoming that like there was no way to overcome it. The best you can do is ask the pathologist, okay, are there a lot or, or very few,  and not generate artificial data, even though if you don’t know what the, what the cells are, it looks perfectly fine.

So that was also my lesson learned from more nuanced stuff. 

Efficiency and Cost of Image Analysis vs. Pathologist Scoring

Aleks: Next question question is, how does the image analysis compared to pathologist scoring in terms of efficiency, efficiency, and turnaround cost [00:14:00] when researchers are considering deciding whether to use image analysis, for a project, which I say they should, but obviously there are more,  things to consider than just me liking the technology.  Right? So,  what do they need to consider, in terms of efficiency cost and all these logistical things? 

Danielle: So, I mean, there’s different aspects to that and with the image analysis, if it’s a new stain or something that’s very unique or complex, the method development, working up the algorithm, verifying and qualifying that algorithm can take time.

So, that’s the part of image analysis that’s potentially inefficient.  however, once you have that developed, the time it takes to run through all the slides and classify them is very fast, much faster than a pathologist scoring.  if, if it’s something that maybe it’s been, you have an algorithm already developed and you just have to tweak it a little for [00:15:00] that range of Staining or that range of samples, it can be quite efficient.

The pathologist scoring, you know, can take more time. Pathologist labor time is expensive. So if, you know, if it is something that’s already been worked up, it can be more efficient and less expensive for the image analysis.  

Aleks: And also I would say, more consistent, and it’s also not because, Oh, the pathologist is lazy and it’s not consistent.  It’s another like neurobiological, thing that we have when we work a study and the, the diagnostic drift where we get more nuanced with our scoring towards the end of the study. And it’s, you know, inherent bias that we have, the, the way to mitigate this is like to go back and try to rescore it.  That adds time on top of what we already did. [00:16:00] And you can totally eliminate this with image analysis.

Validation and Qualification of Image Analysis Algorithms

And you mentioned the validation and qualification of those algorithms. How do you make sure that we are identifying what we’re, what we’re supposed to identify? What is the process of validating such an algorithm or making sure that, we are quantifying what’s supposed to be quantified? How was it in your team? And it’s still the same. 

Danielle: Yeah. So usually the image analysis scientists will work on the algorithm when they think it’s looking good. The pathologist will kind of review it.  The pathologist can also provide some annotations where the, you know, to help train the algorithm in the beginning. So it’s, it’s a partnership all the way through. So the pathologist is involved in the beginning to help. Help the scientists understand, you know, what are the areas we’re looking for? What are the nuances there? And then when the algorithm is ready or when the scientist thinks it’s ready, the pathologist will need to verify [00:17:00] it.

Look at, you know, a set of run through a set of, slides with maybe some differences, the range of what you’re going to see of staining and make sure that it’s performing, adequately. 

Aleks: Okay. So basically a partnership both, with the image analysis scientists, which by the way, are always trained by the pathologist at the beginning of the project.  So they know what they’re looking for. But to mitigate, everything else and all the things that we just mentioned before, there is a pathologist working with them. And so this partnership is both. Can be quantitative when we do annotations, both for training and both for checking and then the visual assessment, which, by the way, this is what we described in the paper that we co-authored.

Itt’s a process for,  the qualification of those algorithms and I’m going to link to this paper below. [00:18:00] So that was a cool. Cool work that we did together and with, with some other colleagues from Charles River and across the industry. 

Danielle: Yeah, and I did want to mention to when you talked about consistency, you having more consistency in the pathologist having diagnostic drift and things like that.

But I think the other thing that’s very important to call attention to is the type of data that you get. So when you have a pathologist scoring you get. You know, data that a categorization, you know, you have things are scored a one or two or three and the everything’s lumped together. So you have, you know, it’s a different type of statistical analysis to perform on that.

Then when you do image analysis where it’s quantitative and it spans, you know, you don’t, it’s not just integral or it’s like a one or two, three spans an entire, you know, infinite number of results. And so the statistical [00:19:00] analysis is different and you have a lot more nuance between rather than lumping everything into categories.

So I think that’s an important thing to distinguish as well, especially when you’re trying to look for differences that are small between groups and, you know, really pull out any kind of statistical significance. 

Aleks: And I would also add to this one also inter-study comparability because you might have different pathologists scoring different studies, but you want to compare the results.

And we’re humans and, 1, 2, 3, 4, 5 can be different for one pathologist and, then another pathologist. And it could, it, and not because like anybody’s wrong, it’s, it may be that, the range in one study is different. So you kind of, subjectively divided into those five buckets, according to the range in the study, whereas in the other one,  it’s going to be [00:20:00] different and your buckets are going to be defined by the range in the study.

And if you do image analysis, you basically get the numbers and you can. And you know, okay, and this one, it was so many, and the other one, it was so many.  So that’s a super important thing. Question, next question. So, we started with,  discussing how image analysis is, used in discovery. And I think in the early research, this is, a tool that everybody uses, often because they have a lot more flexibility than in the regulated environment.

And a lot of our studies, the, the Charles River CRO studies, they need to be conducted in the,  good laboratory practice-regulated environment.

GLP Compliance and Regulatory Considerations

Can image analysis be done according to GLP?  How, how do we do it and why is it important? [00:21:00]

Danielle: Yeah, so, that’s. We, that’s a really good question. We’ve actually, validated the software that we use.

So we use the Visiopharm software mainly. I mean, there’s other softwares that are used throughout the company, but we have validated that software to use in a GLP environment,  through a pretty strict, master validation plan,  and part 11 compliance, it really. It’s a, you know, a lot of interaction with the vendor to make sure that that they’re helping you along the way to get to that compliant level.

and then also for the slide scanners that we use to scan the, to create the digital slides. I think it’s important because, you know, it, it helps us bring image analysis into the regulatory space helps us bring image analysis into studies that are for, you know, R&D submission to the FDA. And we don’t have to take a GLP exception for certain areas of [00:22:00] the study that we’ve done so, you know, we went through that process and I, I think it was very helpful and worth it. 

Aleks: I’ve heard of a story recently where,  there was a submission made to regulatory authorities with semi-quantitative data or qualitative data, for something that should have been analyzed with image analysis.

And, basically that was the regulatory. comment, Hey, you need to provide us with quantitative data, that was generated in the regulated environment. So,  there was kind of an attempt to skip this part and then it, kind of resulted in doing the work twice. So, so I think that’s also something to consider.

Okay, will the regulators ask for quantification when quantification is needed? 

Danielle: And sometimes if there’s, you know, [00:23:00] if it’s a known pathway with a known change,  and, you know, regulators will ask for that depending on,  what the precedence is. So, you know, even for stereology, which is, you know, also a quantitative technique, but a little bit different than the image analysis because it quantifies and 3D, regulators will ask for that to be performed for certain types of compounds in certain pathways. Because of just known changes that can occur. 

Aleks: So, yeah, is there a list or is there like, oh, okay, I wish, we had a list, right?

Danielle: Anecdotally know what I’ve done a lot of work on, but,  it’d be nice if there was a list or if we can. Even, help them, with a list of what we think should be quantified. I don’t know.

Aleks: Maybe, maybe that should be another thing that we should work on together. 

Danielle: Maybe a better, yeah, a paper. [00:24:00]

Aleks: So, at Charles River, at the CRO in general,  such as Charles River,  You might get questions about experience with specific states, like, multiplex techniques, special histology methods.

Method Development for Specific Stains and Techniques

So how do you approach projects that may require new method development, or specific niche procedures? So we already have, like, kind of, if we haven’t done projects with image analysis for a certain thing, we already know that there’s gonna be an algorithm method development…

Danielle: Right.

Aleks: …part. What do you do if there is a very specific request for something that has to be developed from scratch? 

Danielle: Yeah. So I mean, that’s, we get those all the time. A lot of times, you know, we have hundreds and hundreds of immunohistochemical, stains that we’ve done, that we’ve qualified. , but everything’s always different.

So we get those a lot, especially very complex things. So sometimes it will be maybe [00:25:00] we’re multiplexing immunohistochemistry with in situ hybridization to look at protein and DNA on the same section. Maybe it’s a, it’s a multiplex fluorescent stain with, several markers, to quantify, you know, what’s co-locating and things like that.

So we get those all the time. And it’s we have a great, at several sites. We have great immunohistochemistry labs with both technical staff and scientific staff that really understand immunohistochemistry, the science behind it, and how to work up new antibodies as well as for the in situ hybridization for DNA.

So, so we’re used to that. It can take, you know, That adds some time to the project because we, we need to develop the stains and do method development runs and, and qualify a new stain, but we do that all the time. So that’s nothing new. 

Aleks: I think sometimes the, expectation might be, Oh, if it’s already described in this paper, then just like take this and, [00:26:00] and do the same.

Why do you need to develop? And at least, from my IHC experience, whenever you’re starting to do it in the new lab with new reagents, even like for the simplest thing, there is going to be a method development step. You know, if it’s something widely available and easy to do, it’s going to be short.

If it’s new and, and then each procedure, it’s going to be longer. 

Danielle: Yeah. I mean, even the antibodies, if it’s a new lot of antibody, especially if it’s a polyclonal antibody, they’re so different. The lots are very different. So, you know, even lot to lot variation. So when we have a study, if we, we work up a method in one lot, we try to buy enough of that same lot through the entire study.

Because otherwise, it can cause variation within the study. And we don’t want that. So looking at a paper, often the methods aren’t especially detailed. They don’t say, did we do this type of antigen retrieval, they use different immunohistochemical staining, staining instruments, [00:27:00] even the, the water in the lab or the humidity levels.

There’s so, immunohistochemistry is very finicky. So there’s so many aspects that can affect it, that it, You know, if it’s not something that’s been worked up in that lab or and with that lot of antibody,  it needs to be worked up again. And if it’s been worked up the method and it’s a new lot, you know, usually the, you know, it’s more of a requalification of the method, make sure it still works.

But like I said, there can be variation and you may have to change things. 

Aleks: And also, that’s another like argument for, okay, there can be so much variability in the preclinical, and once you have it, have it dialed in, why do you want to add more with, just doing it visually?  Do the image analysis and like, have your numbers, have your, for IHC, right?

I think this is a superpower of any specialty pathology, [00:28:00] method and IFC being big part of this, but also special stains, the immunofluorescence, everything. It visually is so much easier for quantification.  So then like skipping this quantification part with an algorithm is,  depriving yourself of a lot more granularity of data.

Danielle: Yeah, and objectivity, right? Yeah. 

Aleks: Yes, very much.

Future of Image Analysis in Pathology

So looking ahead, how do you see the role of image analysis evolving in pathology and preclinical research? Are there any emerging technologies or techniques that you’re excited about or, do you think the regulators are going to start asking for more quantification on the things that we already know should be quantified?

What are your thoughts on that? 

Danielle: Yeah, I think it’s come a long way. Even when I [00:29:00] first started, the algorithms and type of, you know, decision tree and different things that are used,  you know, it used to be thresholding where you just say, “Okay, If it’s darker than this, it’s positive. If it’s lighter than that, it’s not.

Which was very simple and very binary. Now it’s a lot more complex with a lot of decisions being put into it, behind the scenes, you know, by the computer when you, when you create an algorithm. So it’s already advanced quite a bit and, you know, a new kind of artificial intelligence-based image analysis has really helped us to quantify things not only in immunohistochemistry but also even in an H regular H and E stain.

And I think that helps, because sometimes, you know, we may not have a special stain to mark something, but it still is difficult to tell differences just, you know, with our eye alone. So, a good example of that is hepatocellular hypertrophy. So in, you know, central [00:30:00] lobular hypertrophy in the liver, it’s a very common finding in toxicological studies.

And it can be really difficult to, you know, really determine with your eye where that cutoff is or how to grade, grade those changes, especially in some species where their hepatocytes are already kind of variably sized. And having an algorithm that can perform that, you know, Looking at the size of the cells, on H and E staining through the artificial intelligence-based type of algorithms can be really helpful to augment the pathologists, you know, just diagnostic, vigor.

So, I, I, I’m really excited about more of that coming and you know, of course, that future state of having great, artificial intelligence to weed out all the control and normal and just have the pathologist look at abnormal would be wonderful. That’s easier said than done. We’re still a ways from that, but I, you know, we’re getting there with different tissues and [00:31:00] findings, but I think looking for what are the findings that are really difficult for the pathologist sometimes to differentiate between the groups and, and the hepatocellular hypertrophy is a, is one that’s a,  comes to mind pretty quickly for me.

Aleks: Yes. You know what I have noticed? Because I think for us, it’s like, obvious already image analysis is part of the methods that we can use.  But when I talk to people, to researchers, sometimes they don’t know it’s an option. I know. And sometimes they don’t know it’s an option in a regulated framework.

It’s a, it’s a GLP option. And so I think getting the word out there, is going to help those researchers as well.  Because every now and then I have conversations like, Oh, you can do it really under GLP. What can you do? I’m like, depends on what you need to have quantified. So, [00:32:00] so talking about it, you know, in the podcast and publications everywhere. And still has its value. 

Danielle: Yeah, I agree. I think it’s, you know, we have a lot of publications out there, but I think it’s still something people, it hasn’t been ingrained in us for decades, like a pathologist reading slides has, right? So, it’s still something we, we need to get the word out about more about the value of it, about being able to do it in a regulatory regulated environment, you know, and, you know, the, the difference between, you know, Using it for early research and using it for more of that safety assessment.

I mean, it can be used throughout the, the entire span of the program. 

Aleks: And it’s a funnel, you’re going to be using more of it at the earlier stages. And then as you narrow down, your discoveries, as you narrow down the, the amount of compounds you’re working with [00:33:00] and you like to characterize them better, you’re going to be using less of this and being aware that, okay, you can do it at both stages is super important.

And like you mentioned, okay, this, pathologist just looking at normal is a little bit of a different application than quantifying, to know the exact difference between, the doses between treatment groups. So these are like two different applications. Of the method, and they’re going to be running in parallel, so it’s not that, oh, now everybody is using AI to weed out, normal, abnormal, and speed up the read, and now you don’t need to quantify. No, it’s a different application, and you still need it wherever you need it, right? 

Danielle: Yep, I agree, yeah.

Conclusion and Closing Remarks

Aleks: Thank you so much. Thank you so much for joining me today, Danielle. Thank you so much for, for, you know, all your work that you’ve done to [00:34:00] promote this technology, to promote the use of it in pathology, and for being so successful with your image analysis group that you got promoted, but, the scientists there are still, working on this.

This is, what they do, for all different studies. And I sometimes take part in those projects as well. 

Danielle: Well, thank you for being, so, such an advocate and for doing all of these additional things. Above your day job. 

Aleks: And if anybody is interested in actually doing image analysis on preclinical studies on biomarker discovery, whatever you need image analysis for in your biomedical research.

And I’m going to leave a link to the Charles River page where you can inquire and tell us about your needs and we’re going to be more than happy to get in touch with you. 

Danielle: Yeah, definitely. [00:35:00] 

Aleks: You have a wonderful day, Danielle, and, see you later at work or on teams because we both work in different locations, but we communicate regularly on teams.

Danielle: Yeah. Thank you for having me again and, you have a great day as well. 

Aleks: Thank you so much for listening till the end. This episode was very,  preclinical and toxpath specific. So if you are in this space and you’re. If you are interested in doing image analysis projects or want to outsource image analysis projects and be sure to check the link below and inquire about what Charles River Laboratories can do for you in terms of GLP-compliant tissue image analysis.

and I talk to you in the next episode.