Science Friday - Building a digital ant gallery, from the ground up

Episode Date: March 19, 2026

A project called Antscan has generated high resolution images of thousands of ants, representing over 700 species. To make it happen, researchers brought preserved ants from collections around the world to a particle accelerator in Germany. There, a powerful synchrotron X-ray source combined with a vial-swapping robot allowed the researchers to build a collection of 3D ant images, inside and out. Each voxel (like a 3D pixel) has a resolution of 1.22 micrometers—enough to see the tiny hairs on ant bodies, and distinguish individual muscle fibers. Antscan researcher Julian Katzke joins us to describe the background of the project, and how the images could be used for science and art. Check out Antscan images at our website. Guest: Dr. Julian Katzke is a postdoc at the Smithsonian National Museum of Natural History.  He worked on the AntScan project while a PhD student at the Okinawa Institute of Science and Technology. Transcripts for each episode are available within 1-3 days at sciencefriday.com. Subscribe to this podcast. Plus, to stay updated on all things science, sign up for Science Friday's newsletters.

Hi, I'm Flora Lichten, and you're listening to Science Friday. In the seminal insect film, a bug's life, we're given a bug's eye view of the world. Now stay calm. We are going around the leaf. Around the leaf? I don't think we can do that. Oh, nonsense. This is nothing compared to the twig of 93. We come face to mandible with ants, their antenna, their spindly legs, everything. And they're delightful, but perhaps the accuracy left a little something to be desired. Enter the AntScan Project, which has generated high-resolution X-ray images of over 2,000 real ants from over 700 species.

And I gotta say the results really are stunning. This is ants in all of their beautiful and frightening glory. Joining me now to dig into this mound of ant data is Dr. Julian Katzka. He worked on this project while a PhD student at the Okinawa Institute of Science and Technology. Hi, Julian. Hi, Flora. Thanks for having me. Thanks for being here. Okay, so you came into this work comparing the mouth parts of dozens of ants species.

Why mouth parts? So if you're a human and you live anywhere in the world, there's a pretty high chance that like ants are all around you. And there's this now pretty well-known fact that the biomass of all ants equals or surpasses that of like all humans. And the total diversity of ant species is like, is enormous. And that also extends into their like forms and shapes, particularly in the mouthparts because they are like sort of the first tool that ants use

in their daily lives. So like our mascot model is this South American army. end. They have like small ant workers, but then also large soldiers. And these large soldiers have these fish hook like mandibus that you could even use to like staple a wound. They're by the so fierce that they would never let go. And I think the theory here is that these mandibles are really just there to hurt something that is as big as a human. And so really like learning more about the evolution of these different mouth parts was motivating my PhD research. Describe for people who can't see it, and we'll put some on our website at Science Friday.com slash ants.

What do the images look like? So the images, they are like derived from x-ray images, right? So if you've ever taken like an x-ray image at a hospital, you have this see-through x-ray image. And then if you would like turn yourself around while taking these x-ray images like enough times, you could mathematically reconstruct the 3D volume. So that is like the technology behind computer tomography. And that's exactly what we did with the antscan. The data that we end up with are like slices of images that are like gray scale.

And so they contain the anatomy of the ants. And then you put them together and you can color them and make them beautiful too. Yes. So yeah. So the tomography data is like the raw data and there's like plenty of methods that you can use to generate something. of it. So like in the first instance, it might just be like a measurement, like body size or whatever. But you can also use them in a way like Hollywood 3D animators would. Yeah. I mean, that's what caught my eye. They were just amazing. They're sort of they're like

3D, but in this extreme detail. And they look like aliens. I mean, the ants only look like aliens because they look alien to us, right? They are a bit It's strange because they're so small and like our eyes are just not good enough to really take it all in at once. And that's, I think, it's like one of the basic aspects why digital library of animal shapes and forms can be very important is because so much of like the world out there is just so tiny. And for us to to engage with it more, we need them at like the same scale or even at like larger scales than ourselves. Tell me about the tool that you used. I mean, is this something that you could do in your local lab or like with a CT scanner in a hospital or do you need something special? So definitely not with a CT scanner in a hospital because the resolution is like usually much lower.

But what we use here was a synchrotron light source. So a certain type of like particle accelerator that generates really high energy synchrotron egg. X-ray radiation, and we harness that to do a micro-CT scan very, very fast. But that's not the end of it, right? Because a lot of X-ray radiation in a room would make it very dangerous. So another key part of the technology here that we have a robot that exchanges the samples for us. And then the last thing is that we need a high-speed camera to cope with all that speed.

When you say it's really fast, how fast are we talking? For one ant. So for one end, the imaging itself would be just about 30 seconds. And then at the time when we recorded just another 30 seconds just to transfer the data. Oh, wow. So very fast. Very fast, like especially compared to like what we do in laboratory microsities for insects. So there we talk about like more like 8 to 12 hours for one end.

Wow. Okay. So you can do a lot in a short amount of time. I mean, and can you load up your robot with like 2,000? enhance and just press play? At that time, it was a little bit more limited, so we did 50 at a time, and we still had to turn through a few night shifts to make it work within a week. What resolution can you get to?

So just in terms of numbers, the absolute resolution of like a voxel, which is a 3D pixel, is 1.22 micrometers. That's the highest resolution that we have. But in terms of like outer shape anatomy, you can resolve the delicate hairs that you find on like ant bodies. And then on the internal side, you can resolve up to like individual muscle fibers. So you're getting like all the squishy stuff inside the ants, not just the shell. Yeah. And it's really important, like especially for scientific application, that we have a technology where we cannot only look at like the outer. appearance of the ants, right? There's so much insight that's going on, and that used to be very difficult to study, and that, like, is the brain, all the guts and the musculature of the ants,

right? So, like, ants are really well known for their insane strength, and that must come from somewhere, but it's very challenging to study that. Having 3D data to look inside of them, that makes it a lot easier. You made the data freely accessible. How do you want people to use it? the way it's developing right now, I would always see this in like two different ways. So there's the scientific aspect of using the data for like large-scale projects on like ant evolution and end biodiversity. But we also starting to see this kind of like more engagement with people that are not coming from the science word that might just be happy with like learning a little bit more about ants and like ant shapes and stuff. That's us, by the way. That were your use case for that.

Yeah. Nice. I think the number of people working with like 3D data, like video games are getting ever more popular, that would be the other side of it. So that's like people that want to engage with like 3D data of things that they don't like normally see in their daily 3D lives. These images are arresting. Would you describe them as beautiful? I think, yes. I would say it's beautiful and it's in the quiet.

I think the aspects of like symmetry and asymmetry and sculpture and like forms, extreme forms, they are something that like the more you look at it, the more you understand it, the more you come to appreciate it. Is there a world where you're going to try to train AI on this data so that, you know, we can learn more about ant biology or ant evolution using those tools? Yes, for sure. I think as soon as we're done here, I will have to. to sit down and do a bunch of like annotations to train these models.

No one's going on on Instagram and is like posting a picture of an end and saying like, oh, this particularly is like the left hind leg of the end. And then we would have to do this like a million times over to really be able to to generate a model like that. So we need to all start from scratch here and make use the data that we're given now. I mean, how would you use an ant AI model? Like, why would that be useful? So for science, I would just use an AI model that might be able to, like, distinguish in the 3D data what is the exoskeleton of the end and what is the muscles of the end and what is like all the nervous tissue of the end. And then I could do this for all 2,000s of them and then bring in a phylogenetic tree.

And then I can really say something about the evolution of these traits. And then on the society aspect, like when I have a AI model that can tell me that like this, part of the end is its head, then we would have a much easier workflow to getting like better and more accurate animations. Yeah, for a bug's life, you know, two or three. Yeah, it's like a bug's life two and boogaloo. Are you hoping to do this with other creatures? Yeah, so that's really like why we're calling it the pilot study is that really like by using this like key technology that breaks the bottleneck of like scanning time and then arranging around it

the like the collaboration effort but then also the processing efforts we're really trying to show that it's that is possible to scale this up even further and like I mean I'm still I still have a vested interest in doing more of this for ants but there should be nothing that like excludes other small invertebrate groups

or like other insect groups to be scanned like this was there anything that made you as you know as an ant person just go, what? That's a really, let me just think. That is a question like that, that has never come up. No, I think as an ant person, right? I speak a lot about like the diversity of ants, and sometimes it just becomes like an automatic response, right,

that I say that ants are so diverse. But with this data set, like every time I open one of those scans and there's like 2,000s of them, right? So there are still a lot of scans that I've never opened before. That just makes me realize again and again how different these ants actually are. So even to me as like a researcher, I sometimes like brush over the fact that they are so diverse

and so different. And like I don't really fully encompass it. But when I open one of these scans, that really hits me immediately. Right. And it's easy to overlook, as you say, because we can't resolve those details with our eyes. eyes. So to really appreciate that diversity, you need a tool like this. Exactly.

Dr. Julian Katzka, a postdoc at the Smithsonian Museum of Natural History. He worked on this project while he was a PhD student at the Okinawa Institute of Science and Technology. Thank you, Julian. Thank you. And you can see some of this antclam on our website, ScienceFriiday.com slash ants. Seriously, it is worth a look. I promise you will not regret it. This episode was produced by Charles Burkwist, and if this podcast helps you get a different view of the world, please recommend it to a friend, insect, or otherwise, or leave even just like a teeny, tiny little review wherever you get your podcasts. Thank you for listening. I'm Floralinkman.