The U-M Biomedical Engineering assistant professor will develop scalable computational methods to reconstruct the movements of freely behaving fruit flies and connect behavior with neural activity.
University of Michigan Biomedical Engineering Assistant Professor Kevin C. Zhou has received a Catalyst Grant from the Michigan Institute for Computational Discovery and Engineering (MICDE) to develop computational imaging methods for studying freely moving fruit flies.
Working with co-principal investigator Ravi Allada, the Theophile Raphael M.D. Legacy Professor of Neurosciences and executive director of the Michigan Neuroscience Institute, Dr. Zhou will develop scalable algorithms for four-dimensional reconstruction. In this context, 4D imaging involves capturing three-dimensional volumes over time to visualize dynamic 3D scenes.
“This grant supports an effort to develop scalable 4D reconstruction algorithms,” Dr. Zhou said. “There is a significant push in this direction across computer vision, but we are specifically interested in applying these methods to study freely moving fruit flies as model organisms.”
The one-year project will use conditional Gaussian splatting-based reconstruction to process long, multiview video recordings. The researchers ultimately hope to identify behavioral characteristics associated with different physiological states, including sleep.
Fruit flies are widely used as model organisms because they allow researchers to investigate fundamental biological processes in a relatively simple system. For this project, Dr. Zhou’s laboratory and its collaborators have built an imaging device that records a freely moving fruit fly from several perspectives simultaneously.
“We collect synchronized video from multiple perspectives and reconstruct the fly in 3D as it moves,” Dr. Zhou said. “At the same time, the device can record the fly’s neural activity.”
Allada’s laboratory has genetically engineered the fruit flies to express bioluminescent indicators. These targeted molecular tools glow in response to changes in the animals’ internal physiological states. The approach draws on the same general biological phenomenon that allows fireflies and other organisms to produce light.
“Researchers have found ways to harness bioluminescence as a useful molecular tool for reporting different biological phenomena,” Dr. Zhou said. “We can correlate the 4D behavioral reconstructions with the neural activity reported by these bioluminescent indicators.”
The collaborators in Dr. Allada’s lab are particularly interested in studying sleep and circadian rhythms. By combining detailed reconstructions of the flies’ movements with bioluminescent signals, the research team could validate known indicators of sleep while potentially uncovering previously unidentified behaviors associated with physiological changes.
“Our hope is to identify behavioral phenotypes associated with the neural activity or physiological states reported by the bioluminescence,” Dr. Zhou said.
Studying sleep presents many experimental challenges. Researchers must record the flies continuously for several days to observe behavior across the circadian cycle. The resulting multiview videos contain large amounts of data that can require significant time and computing resources to process.
“If you want to examine the circadian rhythm of sleep, you have to record for multiple days,” Dr. Zhou said. “When you record video for days, analyzing and reconstructing it can take a long time. Our algorithms are intended to facilitate that process.”
The project will develop methods to reconstruct and track a fly’s behavior efficiently across these extended recordings. Conditional Gaussian splatting, the computational foundation for the work, can represent a dynamic 3D scene using collections of adaptable elements. The team will investigate how to apply this approach at the scale required for long-duration biological imaging.
These computational methods could ultimately provide new tools for neuroscience and ethology, the scientific study of animal behavior. By helping researchers examine how observable behaviors correspond with activity inside the nervous system, the platform could enable investigations into relationships that are difficult to study using conventional imaging methods.
The Catalyst Grant advances Dr. Zhou’s laboratory’s broader goal of developing high-speed 4D imaging systems. Although biomedical microscopy has made major gains in spatial resolution and 3D imaging, capturing large volumes rapidly remains difficult.
“A lot of biomedical microscopy and imaging over the past decade has made great advances in improving resolution and enabling 3D imaging, but often at the expense of imaging speed,” Zhou said. “If you have large 3D volumes, you typically image very slowly. If you want to image very quickly, the 3D volume has to be small.”
Dr. Zhou’s team is working to overcome that trade-off through a combination of new imaging hardware and computational algorithms. The Catalyst Grant will seed development of the algorithmic component.
The laboratory is pursuing similar 4D imaging methods in other areas, including projects focused on neural activity, behavior and ophthalmic imaging. Although each application presents different scientific needs, all require systems that can capture changing 3D structures quickly and accurately.
In ophthalmology, for example, involuntary patient movement can introduce artifacts into images of the eye. Imaging more quickly could help researchers maintain a large field of view while reducing those distortions.
“The motivations are different, but the underlying interest is the same: developing 4D imaging systems,” Dr. Zhou said.
MICDE Catalyst Grants help U-M faculty launch early-stage computational projects, test ambitious ideas and generate preliminary results that can position research teams to compete for larger external awards. Since the program began in 2017, individual awards have grown to between $50,000 and $100,000.
MICDE awarded a combined $575,000 to eight U-M research teams during the 2025–26 grant cycle. The selected projects span computational science, artificial intelligence, robotics, materials, neuroscience, cosmology and scientific-computing hardware.
The cycle marked the first time MICDE offered two Catalyst Grant categories: Exploratory grants for bold, early-stage ideas and Readiness grants for more advanced projects preparing to pursue external support. New partnerships with other U-M units also expanded the program’s reach.
“The record response to this year’s Catalyst Grants call shows how deeply computational science and AI are shaping research across the university,” said Karthik Duraisamy, the Arthur B. Modine Professor of Engineering and Samir and Puja Kaul Director of MICDE. “The new award structure allows MICDE to support projects at different stages of maturity, from ambitious early ideas to teams preparing for major external opportunities.”
For Dr. Zhou and his collaborators, that support provides an opportunity to unite computational imaging, neuroscience and biological research—turning days of complex video into a dynamic view of how neural activity and behavior intersect.