Science

Artificial intelligence lets smartphone microscopes produce lab-quality images

Artificial intelligence lets smartphone microscopes produce lab-quality images
Image of a blood smear from a smartphone camera (left), following enhancement by the algorithm (center), and taken by a lab microscope (right)
Image of a blood smear from a smartphone camera (left), following enhancement by the algorithm (center), and taken by a lab microscope (right)
View 2 Images
Image of a blood smear from a smartphone camera (left), following enhancement by the algorithm (center), and taken by a lab microscope (right)
1/2
Image of a blood smear from a smartphone camera (left), following enhancement by the algorithm (center), and taken by a lab microscope (right)
An illustration showing how the technology enhances images
2/2
An illustration showing how the technology enhances images

There are now various attachments that allow you to capture microscope-scale images with your smartphone. Unfortunately, however, the limitations of the phone's lens and image sensor mean that those images still won't be as good as those obtained by a lab-grade microscope. That could change, however, thanks to a recent advance in artificial intelligence.

Led by Prof. Aydogan Ozcan, a team at the University of California Los Angeles (UCLA) started by taking microscope photos of lung tissue samples, blood and Pap smears. They did so two times, first using a standard laboratory-grade microscope, and then using a smartphone with a microscope attachment which they made with a 3D printer for under US$100.

Both sets of images were then uploaded to a laptop, which was running a deep-learning-based algorithm created by the scientists. Using the lab microscope photos as a reference, that algorithm learned how to rapidly enhance the smartphone photos to the point that they were similar in quality to those reference photos, possessing the level of resolution and color details needed for a laboratory analysis.

An illustration showing how the technology enhances images
An illustration showing how the technology enhances images

The system has since been shown to be capable of accurately enhancing other low-quality images, even when it doesn't have a corresponding high-quality image to use as a guide. It is now hoped that the technology could find use in resource-poor regions, where microscopes are in short supply.

"Using deep learning, we set out to bridge the gap in image quality between inexpensive mobile phone-based microscopes and gold-standard bench-top microscopes that use high-end lenses," says Ozcan. "We believe that our approach is broadly applicable to other low-cost microscopy systems that use, for example, inexpensive lenses or cameras, and could facilitate the replacement of high-end bench-top microscopes with cost-effective, mobile alternatives."

A paper on the research was recently published in the journal ACS Photonics.

Source: UCLA

3 comments
3 comments
Bob
I'm not sure I agree that this is making a quality image. It is merely inserting pixels that it thinks should be there. Once that is done it will of course look like the quality image but is it? The photos above made the enhanced photo even smoother and less pixelated than the lab microscope. Let's take a low quality photo of an average woman and fill in the missing pixels with photos of super models. Will photos of average women all look like super models? It will probably sell like hot cakes but be far from accurate.
aki009
This is really neat stuff, but from a medical perspective likely mostly useless.
The net is designed to fill in missing data by making educated guesses. Those guesses are determined by the data it was taught with. Even if the teaching set includes anomalies, they will easily go without enhancement in real use given the many manifestations that they have.
Naturally this tech might be better than nothing, but a tool that gives doctors false negatives can be worse than knowing that there is an unknown.
Douglas Bennett Rogers
A neural net is ideally suited to identifying something like a missile target. It knows exactly what it is looking for and it is very fast.