AI-Enhanced Microscopy Captures Real-Time Cellular Video

Engineers at the University of California San Diego have developed an artificial intelligence (AI) method to observe live cells in real time. This new technique produces images twice as sharp as conventional microscopes. It also generates video fast enough for smooth playback.

The advancement, detailed in *Nature Communications*, uses an algorithm to streamline a previously slow and computationally intensive process. The algorithm creates reliable, high-quality images instantly. It does so without introducing false details. This innovation could make advanced microscopy more practical for daily research.

The technique builds on structured illumination microscopy (SIM). SIM enhances image detail by shining patterned light onto a sample. It then combines a small number of images. SIM is useful for studying live cells due to its speed and minimal light exposure, which prevents cell damage. However, some SIM systems require precise calibration of light patterns. Errors in these patterns can reduce image quality. Simpler SIM systems, which use random light patterns, often have slow image processing times.

To address these issues, a team led by Zhaowei Liu developed an upgraded version called unrolled blind-SIM (UBSIM). UBSIM integrates AI into the image reconstruction process. This allows it to produce high-quality images hundreds to thousands of times faster. It also maintains simpler hardware requirements. Scientists can now view detailed images as they are captured, eliminating processing delays.

The method is based on the physics of image formation. This design helps prevent misleading details sometimes seen in traditional AI-based approaches. In tests with live cells, UBSIM produced high-resolution video at up to 50 frames per second. This video showed rapid changes in structures like the endoplasmic reticulum in real time. The researchers state that UBSIM makes super-resolution microscopy as convenient as a traditional light microscope.