MARY LOUISE KELLY, HOST:
A father's quest to catch the earliest signs of eye disease and prevent loss of vision is making progress. Five years ago, NPR reported the story of Bryan Shaw's son Noah and how he lost an eye to cancer. Shaw has developed a smartphone app that aims to detect signs of eye disease in family photos taken long before a child ever sees a doctor. Now a prestigious science journal has published the first results. NPR's Joe Palca has more.
JOE PALCA, BYLINE: Doctors diagnosed Noah Shaw's eye cancer when he was 4 months old. To make the diagnosis, the doctor shined a light into Noah's eye and got a white reflection from the tumor growing at the back of his eyeball. Noah's dad Bryan is a scientist. And as he told me in 2014, he wondered if he could see that same white reflection in flash pictures his wife was always taking of their baby son.
BRYAN SHAW: I told my wife, give me all your pictures. Give me all the pictures. I need all the pictures.
PALCA: Now, often, when you take a flash picture, you get red eye - a red reflection from the blood vessels at the back of the eye. Shaw looked to see if there were examples of a white reflection in Noah's baby pictures, and there were months before his diagnosis.
SHAW: We had white eye showing up in pictures at 12 days old.
PALCA: Now, Shaw is a chemist not an eye doctor or a computer scientist, but he decided to create software that could scan photos for signs of this white reflection.
SHAW: If I would've had some software telling me, hey, go get this checked out, that would have sped up my son's diagnosis. And the tumors would have been just a little bit smaller when we got to them. There might have been fewer.
PALCA: And now that software exists. Along with colleagues at Baylor University in Waco, Texas, Shaw created an app called CRADLE. Using those pictures of his son and others he collected from other parents, he trained the app using artificial intelligence to find white eye. To test the app, they analyzed more than 50,000 pictures taken of 40 children. Half had no eye disease. Half had been diagnosed with eye cancer or some other debilitating eye disease.
SHAW: On average, the app detected white eye in pictures collected 1.3 years before diagnosis.
PALCA: In other words, the app could give an early alert to parents that something might be amiss with their child. The results appear in the journal Science Advances. Now, Shaw says the app isn't perfect. It sometimes misses white eye when it's there and sometimes says it's there when it's not.
SHAW: That's a problem.
PALCA: Those so-called false positives occur infrequently - less than 1% of the time. But ophthalmologist Sean Donahue of Vanderbilt University says that's not good enough because with 4 million children born in the United States each year, a 1% false positive rate would mean tens of thousands of children showing up at doctors' offices unnecessarily.
SEAN DONAHUE: It's going to have to be substantially better than 1%.
PALCA: Still, Donahue is upbeat about the promise of the app.
DONAHUE: This is exciting new technology. And this is, I think, how we're going to go with screening for a number of different diseases in the future.
ALISON SKALET: There's certainly promise here, and it makes sense to me to be harnessing the technology that we have.
PALCA: Alison Skalet specializes in eye cancer at the Oregon Health and Science University. She expects the app will get more accurate as time goes on and its artificial intelligence gets smarter. Bryan Shaw would like to see that. But to train the app to better recognize white eye, he needs something.
SHAW: Pictures - we need more pictures, especially of kids from Africa and Asia.
PALCA: That way, Shaw says, he can make the app more globally relevant and save more children's vision.
Joe Palca, NPR News.
(SOUNDBITE OF MUTUAL BENEFIT SONG, "TERRAFORM")
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.