Bryan Shaw set out to develop the app five years ago after his son Noah was diagnosed with retinoblastoma and ultimately lost his eye. Shaw discovered that signs of the disease were evident in pictures taken of Noah at just 12 days old, where there was a pale glow in one of his eyes that doctors call “white eye.”
Shaw developed the app, called CRADLE, with help from colleagues at Baylor University in Waco, Texas. CRADLE makes use of artificial intelligence to look at baby pictures taken with a flash and pick out instances of white eye. When they tested it on 50,000 images of 40 children, half of whom had been diagnosed with an eye disease, it was able to identify white eye in pictures taken up to 1.3 years before the child was diagnosed.
The app isn’t perfect — it has a 1 percent false positive rate, which many consider to be far too high — but as AI systems get smarter and Shaw receives more images on which to train the app, he expects that rate to improve.
