Engineers at the University of California San Diego have developed a groundbreaking microscopy technique that uses artificial intelligence to produce crisp, real-time video inside living cells. Published in Nature Communications, the new method, known as unrolled blind-structured illumination microscopy (UBSIM), overcomes traditional limitations to deliver high-quality images at speeds previously unattainable. The technology captures images twice as sharp as conventional microscopes while processing them instantly to create smooth, live video. The technique builds on structured illumination microscopy (SIM), a method that enhances detail by projecting patterned light onto a sample. While SIM is ideal for studying live cells due to its speed and minimal light exposure, existing systems face significant hurdles. Precise calibration of light patterns is difficult, and small errors can degrade image quality. Conversely, simpler systems using random patterns often suffer from slow processing times that can take seconds or minutes per frame, making real-time observation impossible. To address these issues, the team led by Professor Zhaowei Liu from the Jacobs School of Engineering integrated artificial intelligence directly into the image reconstruction process. This approach creates UBSIM, which generates reliable, high-quality images hundreds to thousands of times faster than standard methods without requiring complex hardware. A key advantage of UBSIM is its adherence to the physics of image formation. By anchoring the algorithm in physical laws rather than relying solely on data-driven neural networks, the system avoids the risk of hallucinations or fake details often seen in other AI models. Zachary Burns, the study's first author and a Ph.D. student in Liu's lab, highlighted the importance of this accuracy. He noted that many neural network models can invent structures when analyzing new data, a critical flaw for scientists who need to trust their observations. By removing these artifacts, UBSIM ensures that the structures observed within cells are real, building confidence in the results. In tests involving live cells, the system achieved high-resolution video at up to 50 frames per second. This speed allowed researchers to observe rapid biological changes, such as movements within the endoplasmic reticulum, as they happened. Professor Liu stated that the ability to reconstruct and display super-resolution images in real time without supervision makes the technology as convenient as a traditional light microscope. This advancement is expected to significantly improve user experience and accelerate discoveries in cell biology by making high-end microscopy practical for everyday laboratory use. The innovation effectively bridges the gap between high-resolution capabilities and the speed required for dynamic live-cell analysis.