AI transforms microscopy, enabling diagnostic-quality "virtual staining" without chemical dyes

A new study demonstrates how artificial intelligence (AI) can generate high-resolution, diagnostically useful pathology images directly from unstained tissue samples. Published in BME Frontiers, the research combines a computational imaging technique with deep learning to create a faster, cheaper, and more accessible alternative to traditional chemical staining.

The method, called Virtual Staining Fourier Ptychographic Microscopy (VS-FPM), uses a low-cost microscope setup to capture detailed "phase images" of label-free tissue. A generative AI model is then trained to instantly translate these images into realistic virtual Hematoxylin and Eosin (H&E) stains—the gold standard in pathology.

The technology revealed no statistically significant difference between the spatial resolution of FPM images captured at 4× magnification and images from a pathology slide scanner at 20× magnification. Visual assessment and image similarity metrics showed that VS-FPM images of unstained tissues closely resemble images of chemically H&E-stained tissues.

Clinical validation by pathologists confirmed VS-FPM’s accuracy in distinguishing normal from dysplastic tissues, minimizing inter-laboratory variability. The LED-based system is low-cost and compatible with existing microscopes, making it accessible for low-resource settings.

“VS-FPM addresses multiple limitations of conventional pathology at once,” said Dr. Michael Shaw, the study's corresponding author from University College London and the National Physical Laboratory. “We are creating a more flexible digital workflow. The ability to image without stains and digitally refocus opens new possibilities for diagnostics.”

This work paves the way for more streamlined lab workflows, reducing turnaround times and toxic chemical waste. Importantly, the original unstained tissue remains intact for further molecular testing.