New method uses artificial intelligence to improve cell image quality

SALK INSTITUTE (KUSI) – Scientists at the Salk Institute needed a way to improve low-quality images critical for seeing structures inside cells.

Now, they have figured out how to use artificial intelligence to improve microscope images.

Barriers to this included microscopes being extremely expensive, and even with those high-tech microscopes, the low light allowed for the absence of important detail.

High amounts of light could damage the cells, making them unable to be studied.

The solution uses “deep learning,” in which computer software learns to recognize something and improve by studying examples.

That’s precisely what the Salk team did: take high quality images, degrade them, and made them low quality — a system they dubbed “crappification,” or making the images look like “crap.”

Afterwards, the system was then able to learn to improve naturally poor images or “decrappify.”

The scientists’ method worked on images of brain tissue, which would have otherwise taken a much longer amount of time to get an image of at high resolution.

The team hopes this can make microscope imaging more accessible and less expensive.

Dr. Uri Manor, director of the Waitt Advanced Biophotonics Core Facility, joined KUSI to discuss the Salk Institute’s findings.

Categories: Health, Local San Diego News