AI-Powered Microscopy Enables Real-Time Imaging of Living Cells

Using artificial intelligence, engineers at the University of California San Diego have developed a new way to watch the inner workings of living cells in real time. The process both captures images that are twice as sharp as conventional microscopes and is fast enough to play as smooth video.

The advance, published in Nature Communications, relies on an algorithm that transforms a once slow and computationally intensive process into one that produces reliable, high-quality images instantly and without introducing false details. It could make cutting-edge microscopy more practical for everyday research.

The technique builds on a widely used imaging method called structured illumination microscopy (SIM), which enhances image detail by shining patterned light onto a sample and combining a small number of images. SIM is especially useful for studying live cells because it works quickly and minimizes light exposure, which can damage cells. However, some SIM systems are difficult to use because they require precise calibration of light patterns. Even small errors in these light patterns can reduce image quality. Meanwhile, simpler SIM systems that use random light patterns often suffer from slow image processing that can take seconds or minutes per frame.

To overcome these challenges, a team led by Zhaowei Liu, professor in the Department of Electrical and Computer Engineering at the UC San Diego Jacobs School of Engineering, developed an upgraded version of the technique, called unrolled blind-SIM (UBSIM). By integrating artificial intelligence into the image reconstruction process, UBSIM produces high-quality images hundreds to thousands of times faster while maintaining simpler hardware. This means that scientists can view detailed images as they are captured rather than wait for processing to finish.

And because the method is built from the physics of how images are formed, it also avoids the risk of introducing misleading details sometimes seen in traditional AI-based approaches.

"One of the most exciting advancements with this algorithm is the removal of artifacts and hallucinations," said study co-first author Zachary Burns, an electrical and computer engineering Ph.D. alumnus from Liu's lab. "Currently, many neural network-based models can imagine fake structures when they are applied to new data. This is a major problem for scientists who use these AI models - they need to trust that the structures in the cell they are observing are real. By integrating optical physics, our model removes these issues and builds confidence that it can be used accurately."

In tests with live cells, UBSIM produced high-resolution video at up to 50 frames per second. The video revealed rapid changes in structures such as the endoplasmic reticulum in real time.

With UBSIM, a super-resolution image can be reconstructed and displayed in real-time without any supervision, making super-resolution microscopy as convenient as a traditional light microscope. This will significantly improve the user experience and the effectiveness of using super-resolution microscopy for discoveries."

Zhaowei Liu, Professor, Department of Electrical and Computer Engineering, UC San Diego Jacobs School of Engineering

Researchers say future work will focus on further improving resolution.