Back
Science

Berkeley researchers develop MOSAIC microscope and plan AI-powered Cell Observatory

View source

The MOSAIC Microscope: A 12-in-1 Imaging Powerhouse

Researchers at the University of California, Berkeley, have unveiled a revolutionary new tool for microscopy. Named MOSAIC (Multimodal Optical Scope with Adaptive Imaging Correction), the device is detailed in a recent publication in Nature Methods.

The system integrates 12 distinct imaging modes—ranging from traditional phase contrast to advanced lattice light-sheet microscopy—all switchable at the press of a button. This allows scientists to capture high-resolution 5D data (three spatial dimensions, time, and color) from living specimens.

"The impact of MOSAIC will be minimal until we build an AI model to be able to deal with the data that comes out of those systems."
— Srigokul Upadhyayula, assistant professor and lead developer

A Breakthrough in Complexity and Data Volume

The microscope’s capabilities come with a significant challenge: it generates petabytes of data. More than a dozen labs have already replicated the system based on preprints and instructions distributed by the UC Berkeley team over six years.

To handle this flood of information, the team is now focusing on building an AI Large Vision Language Model (LVLM) as part of Berkeley's Cell Observatory Initiative. The goal is to analyze data that simply exceeds human capacity.

Eric Betzig, Nobel laureate and professor at Berkeley, emphasized that "the data volume and complexity exceed human interpretation capacity," underscoring the necessity of the AI integration.

Technical Foundation and Applications

The MOSAIC microscope relies on a sophisticated combination of technologies:

  • Fluorescent labels
  • Light-sheet imaging
  • High-speed computing
  • Adaptive optics

These elements allow the microscope to track subcellular dynamics, map nanoscale features, and image neural architecture in live mice.

Key Contributors and Funding

The study was led by:

  • First Authors: Gaoxiang Liu and Xiongtao Ruan (UC Berkeley), Tian-Ming Fu and Daniel Milkie (Janelia Research Campus)
  • Senior Authors: Srigokul Upadhyayula, Eric Betzig, and Wesley Legant

Major funding was provided by the Howard Hughes Medical Institute (HHMI), Philomathia Foundation, Biohub, Sloan Foundation, and Berkeley Lab.