The Eye Disease testing application surpasses the 'sensitivity of the gold standard', said Baylor University researchers.
The Baylor University researchers' prototype smartphone application – designed to help parents detect early signs of eye diseases in their children such as retinoblastoma, aggressive childhood eye cancer – has passed its first major test.
The CRADLE (ComputeR Assisted Detector LEukocoia) application looks for traces of abnormal reflections from the retina called leukocoria or "white eyes," the main symptoms of retinoblastoma, as well as other common eye disorders. This study, published in the journal Progress of Science, found this application is an effective tool for adding clinical leukocoria screening, which allows parents to screen their children more frequently and efficiently throughout their development.
CRADLE – developed by Baylor University researcher Bryan F. Shaw, Ph.D., professor of chemistry and biochemistry, along with Greg Hamerly, Ph.D., associate professor of computer science – searched for family photographs to look for signs of leukocoria.
According to the study's first author, Baylor University senior researcher Micheal Munson, the researchers determined sensitivity, specificity and accuracy from the prototype by analyzing more than 50,000 photos of children taken before their diagnosis. For children with eye disorders who are diagnosed, CRADLE is able to detect leukocoria for 80 percent of children. The application detects leukocoria in photos taken an average of 1.3 years before their official diagnosis.
The effectiveness of traditional screening during general physical examination is limited, with signs of retinoblastoma through detection of leukocoria in only 8 percent of cases. CRADLE sensitivity for children ages 2 and younger exceeds 80 percent. The 80 percent limit is considered by ophthalmologists as a "gold standard" of sensitivity for the same device, Munson said.
The researchers found the CRADLE application to be more effective only by the breadth and frequency of the sample size: daily family photos, according to the study. Given the number of photos taken by family and friends as well as various environments, there are various opportunities for light to reflect ocular lesions regardless of their location in the eye.
As the application algorithm becomes more sophisticated, its ability to detect even a few leukocoria events has increased.
"This is one of the most important parts of building applications," Shaw said. "We want to be able to detect all the colors and intensity of leukocoria. As a parent of a child with retinoblastoma, I am very interested in detecting traces of leukocoria that appear as pupils 'gray' and difficult to detect with the naked eye. "
Initially, the CRADLE application was used primarily to identify retinoblastoma – a rare eye disease which is the most common form of eye cancer in children up to age 5. Shaw's own experience as a parent of a child with retinoblastoma forms the origin of the application.
Shaw and Hamerly made an application in 2014 for the iPhone and 2015 for Android devices after Shaw's son, Noah lost his right eye, but his left eye could be saved. He is now 11 years old.
"We suspect that the application will detect leukocoria associated with other disorders that are more common and some are rare," Shaw said. "We are right. So far parents, and some doctors, have used it to detect cataracts, retinal myelin nerve fiber layers, refractive errors, Coats disease, and of course retinoblastoma. "
Munson said: "I only remember its purpose: saving visions and the potential lives of children throughout the world," Munson said.
Shaw said they were retraining the algorithm with Baylor scholars currently marking and sorting about 100,000 additional photos. He said they were also looking for additional features to reduce detection of false positives.
This application can be downloaded for free and can be found under the name "White Eye Detector."
Reference: "Early detection of autonomic eye disease in childhood photos" by Micheal C. Munson, Devon L. Plewman, Katelyn M. Baumer, Ryan Henning, Collin T. Zahler, Alexander T. Kietzman, Alexandra A. Beard, Shizuo Mukai, Lisa Diller, Greg Hamerly and Bryan F. Shaw, 2 October 2019, Progress of Science.
DOI: 10.1126 / sciadv.aax6363