Modern imaging systems struggle when light has to pass through materials like deep tissue or dense fog, creating a scattering effect that produces a blurry picture. But researchers at the University of Rochester, in the state of New York, have made a breakthrough with an AI-enhanced imaging system that not only offers clearer images, but could also be much less expensive than existing systems.

The applications are manifold, but particularly promising in biomedical imaging for cancer, where dense tissue can scatter light.

When light enters clear air, it tends to follow a straight path, but when it encounters something more complicated, such as the layers of tissue inside the human body, some of the light gets knocked off course. The resulting image can look as though someone smeared a camera lens.

Near-infrared light is particularly useful because it can penetrate deeper into biological tissue than visible light, making it valuable for medical imaging. But detecting that light typically requires specialized sensors that can be considerably more expensive than conventional silicon-based cameras.

The University of Rochester system, announced August 17, 2026, uses a technique called time-gating, which can be thought of as an extraordinarily fast shutter. Instead of allowing all the light reaching a detector to contribute to an image, the system opens a gate for an almost unimaginably short period of time, allowing it to distinguish useful light from much of the light that has been scattered along the way.

“Time-gating essentially works like the shutter in a camera,” said Yang Xu, the lead author and a Rochester Ph.D. “In this case, we use light to control light.”

The AI is not simply applying a digital sharpening filter to a blurry photograph. It has learned patterns that allow it to reconstruct a larger and clearer scene from the optical information it receives. And it relies on the much less expensive silicon-base.

A useful analogy is a jigsaw puzzle. Instead of having every piece, the system receives some of the pieces and uses what it has learned about the picture to determine how the missing portions fit together. The approach dramatically expands the system’s field of view and improves the quality of the reconstructed images. The AI is not being asked to create information from nothing; it is working with information captured by an optical system designed to preserve useful signals while rejecting much of the noise caused by scattering.

The potential applications in the medical field are broad, the technology could eventually help researchers develop better ways to image tissue without invasive procedures.

“Before applying artificial intelligence, we could see only a limited field of view,” Xu said. “By adding our collaborators’ methods, we can essentially reconstruct a much larger target area, enlarging the field of view our ultrafast time-gating technique can capture.”

Other University of Rochester collaborators involved in the studies include optics alumna Saumya Choudhary, Ph.D., and physics doctoral student Long Nguyen. The US Office of Naval Research, the National Science Foundation, and the Department of Energy provided funding for the research.

The technology is still in the research stage, however, and is not a clinical diagnostic tool today. More testing would be required before such a system could be routinely used on patients.

The system also has a practical application with autonomous vehicles, which use LiDAR, short for light detection and ranging, to measure distances and build a picture of their surroundings. The technology sends out pulses of light and measures how long they take to return after hitting an object.

But fog, dust and other particles can scatter that light, making it harder for the system to distinguish a person, vehicle or other obstacle from its surroundings. A sensing system that can better separate useful light from scattered light could help autonomous vehicles and other LiDAR-equipped systems operate more effectively when visibility is poor.

At NYU, researchers are developing AI methods to reconstruct MRI scans from less raw data, with the goal of making MRI dramatically faster. Other systems are being developed to help radiologists detect breast cancer, analyze brain images and identify early signs of diseases such as Alzheimer’s. The idea is to use AI throughout the imaging process, from capturing and reconstructing an image to helping physicians interpret it.

That does not necessarily mean humans will no longer be part of the process. Research into AI-assisted radiology points toward collaboration between humans and machines. AI can examine enormous amounts of visual information without fatigue, while physicians bring clinical experience and judgment, including the ability to recognize when a machine may be wrong.