Scientists have developed a faster method for creating highly detailed 3D models of ants by combining advanced X-ray imaging, robotics, and artificial intelligence (AI). The technique allows researchers to digitally reconstruct hundreds of species in a fraction of the usual time.
The study, published on 5 March in the journal Nature Methods, was led by researchers including Evan Economo of the University of Maryland and Thomas van de Kamp from the Karlsruhe Institute of Technology in Germany.
For more than a decade, Economo’s laboratory has relied on micro-CT scanners to examine insect morphology, the study of their physical structures. While technology produces extremely detailed 3D images, scanning a single specimen can take up to 10 hours.
To overcome this limitation, the research team introduced a high-throughput system that integrates a synchrotron particle accelerator, X-ray imaging, robotics and AI. This approach enabled scientists to rapidly scan thousands of ant specimens and transform the data into interactive 3D models.
The project, known as Antscan, produced digital reconstructions of about 800 ant species. According to the researchers, completing such a project with conventional CT scanners would have taken roughly six years of continuous work.
Instead, using facilities at KIT, the team scanned about 2,000 specimens in just one week.
Researchers collected ethanol-preserved ants from museums, partner institutions and specialists worldwide. The specimens were then transported to KIT, where a powerful synchrotron beam generated intense X-rays capable of rapidly imaging multiple samples.
A robotic system handled the specimens during scanning, rotating each one and replacing it with another every 30 seconds. The process produced stacks of 2D images that were later combined into detailed 3D reconstructions.
However, many of the scanned ants initially appeared in distorted positions. To correct this, computer science students helped develop AI tools that automatically adjust the posture of the insects, creating lifelike models like how ants appear in nature.
The resulting models reveal internal structures such as muscles, nervous systems, digestive organs and stingers at micrometer-level resolution. Scientists say digital ants can also be animated or placed in virtual reality environments for research, education or even visual media production.
Antscan database has already supported new scientific research. In a separate study published in Science Advances in December 2025, researchers used the data to explore how ant colonies balance worker size and physical strength.
By examining more than 500 species, the team found that colonies investing less in thick exoskeleton armor often support larger numbers of workers. The findings suggest that lower investment in protective cuticles may allow colonies to grow larger and diversify more successfully.
Scientists say the detailed 3D models make it possible to precisely measure structures such as cuticle volume, something that was previously difficult to calculate.
Researchers believe the growing Antscan archive could eventually serve as a digital library of biodiversity. The detailed scans may also help train machine-learning systems to automatically identify ants during field studies.
The team plans to expand the database by scanning more specimens and applying similar AI-based methods to other biological datasets, potentially opening new opportunities for studying the diversity of life on Earth.
