How deep learning can help us identify cancer cells

AI-microscopeTackling cancer remains notoriously difficult, not to mention expensive.  A team from UCLA have developed a clever technique for identifying cancer cells much faster than is currently the case.

The method, which was documented in a recent study, takes a different stance to the current labeling approach that can fall down due to inaccuracy.

Instead, the new method captures an image of the cells without harming them.  It’s capable of detecting 16 different physical characteristics, such as its biomass, granularity and size.

Smarter microscopes

The method takes advantage of a couple of UCLA inventions, including a photonic time stretch microscope and a deep learning algorithm that is capable of identifying cancer cells with high levels of accuracy.

The team believe that the microscope is just one of the possible applications of the technology.  It works in a similar way to a flash on a camera, and takes pictures of the blood cells using laser bursts.

Of course, such a process is done in nanoseconds so would ordinarily be much too fast to be digitized by normal equipment.  The new microscope is capable of overcoming this by using smart optics that both enhance the clarity of the images and slow them down long enough to be detected.  The deep learning algorithm then goes to work on identifying the cancerous cells from their healthy peers.

“Each frame is slowed down in time and optically amplified so it can be digitized,” the researchers say. “This lets us perform fast cell imaging that the artificial intelligence component can distinguish.”

Working in the dark

The approach is also clever in the way it works with light.  Ordinarily, the light generated by taking such pictures would destroy the cells, but the photonic time stretch approach allows pictures to be taken with low-level illumination.

It’s certainly a fascinating approach, and the team believe that it could result in a much more data driven approach to diagnosis.

Check out the video below for more information.

Facebooktwittergoogle_plusredditpinterestlinkedinmail

Leave a Reply

Your email address will not be published. Required fields are marked *