Scanning 2,000 ants in a week: how a particle accelerator built a 3D insect library

It would have taken six years of continuous operation with a standard lab scanner. Instead, it took a week.

A research team led by Evan Economo at the University of Maryland and Thomas van de Kamp at the Karlsruhe Institute of Technology combined a particle accelerator, robotic sample handling, and artificial intelligence to scan 2,000 ant specimens and build an open-access database of high-resolution 3D models covering 800 ant species. The project, called Antscan, is described in a paper published in Nature Methods on March 5, 2026.

From specimen vials to synchrotron beam

Building the database started with old-fashioned curation. Ethanol-preserved ant specimens were sourced from partner institutions, museum collections, and experts worldwide, then sorted by species and caste at the Okinawa Institute of Science and Technology (OIST), where Economo previously held his primary appointment. The sorting alone took weeks of manual work.

"At times, this work involved our entire lab -- we sat for two weeks straight sorting each ant by hand after a full month of cataloging," said Julian Katzke, the study's first author and a graduate of Economo's lab at OIST.

Once sorted into standardized trays, the specimens traveled to KIT in Germany for imaging. There, a synchrotron particle accelerator produced a high-intensity X-ray beam for micro-CT scanning -- essentially a medical CT scan scaled down to insect size and magnified enormously. A robotic arm swapped specimen vials automatically, producing 3,000 X-ray images of each ant every 30 seconds. Those 2D image stacks were then reconstructed into 3D tomographic models.

What the scans reveal

These are not surface photographs. The micro-CT scans capture internal structures at micrometer resolution: muscles, nervous systems, digestive systems, and stingers are all visible. The level of detail makes the models useful for studying locomotion, biomechanics, and comparative anatomy in ways that were previously impossible without painstaking manual dissection.

"We've estimated that if we were to carry out this project with a lab-based CT scanner, it would take six years of continuous operation," Katzke said. "With the setup at KIT, we scanned 2,000 specimens in a single week."

AI for pose correction

The raw scans presented a cosmetic problem. Ethanol-preserved specimens often have contorted limbs and unnatural positions. To make the models useful for animation, education, and public engagement, the team needed a way to transform awkward poses into natural ones.

That task fell to computer science students at the University of Maryland, where Economo now chairs the Department of Entomology. Students in Associate Professor James Purtilo's software engineering course are developing AI-based "pose estimation" tools that automatically adjust specimen postures. "This problem was a doozy," Purtilo said.

Already producing science

The database has already demonstrated its research value. In a December 2025 paper in Science Advances, Economo and collaborators used Antscan data to investigate a fundamental question about ant societies: do colonies fare better with many lightly armored workers or fewer heavily armored ones?

By measuring cuticle volume -- the thickness of the protective exoskeleton, which is nutritionally expensive to produce -- across more than 500 species, the team found a strong negative correlation between armor thickness and colony size. Colonies that invested less in individual ant quality could afford more workers, leading to larger and more diverse societies.

Cuticle volume had been extremely difficult to measure before Antscan. Combined with a separate 2025 study in Cell that produced high-quality genomes for most ant genera, the 3D morphological data opens possibilities for studying how shape and genetics relate across the ant family tree.

Open access by design

All Antscan raw files are freely downloadable, and the portal includes a built-in viewer for browsing any ant in 3D. The decision to make everything open was deliberate.

"One of our goals was to democratize access to high-resolution micro-CT scans, which can be prohibitively expensive, especially for smaller institutions or non-institutional experts like citizen scientists, local collectors, or artists and educators," Katzke said.

The workflow itself serves as a blueprint. While this project focused on ants, the same combination of synchrotron scanning, robotics, and AI-based processing could be applied to beetles, flies, spiders, or any small organism preserved in museum collections worldwide.

"When specimens are digitized, we can build libraries of organisms that can streamline their use from scientific laboratories to classrooms to Hollywood studios," Economo said.