The AI-powered eye exam could replace the Snellen eye char in the future. The StAT test has proven to be more accurate. (Image Credit: Gary Cassel/Pixelbay)
Classic eye exams are poised to get an upgrade in the future. Researchers from Stanford University have created an online vision test powered by AI. This system is capable of producing diagnoses that are more accurate than the sheet of capital letters used in today’s eye exams. This eye test could be used by patients suffering from eye diseases, helping them to accurately monitor their vision at home. Users can take the AI-powered eye exam at myeyes.ai, but it’s not yet meant to replace visits to the optometrist. The team presented their findings in the Proceedings of the AAI Conference on Artificial Intelligence.
Chris Piech, a computer scientist at Stanford University, says the easy-to-use Snellen chart administrated in traditional eye exams isn’t perfect and has some flaws. Patients would often start to guess the letters whenever they became blurry during the examination, causing inaccuracies. This makes it more likely for patients to score differently each time they participate in the test.
To eliminate human error from the Snellen exam, while improving its accuracy, Piesch and his team created an online test. Before beginning the exam, users are required to calibrate their screen size by adjusting a box on the webpage to match the size of a credit card. They will then need to input how far away they are from the screen, and after entering the distance, an “E” text appears on the page in one of four orientations. The algorithm will then use statistics to estimate a vision score once an answer is provided by the user. As the exam continues, the algorithm can make a more accurate prediction of the score. It asks 20 questions per eye, taking just a few minutes to complete the test.
As the team ran the Stanford acuity test (StAT) through 1000 computer simulations acting like real patients, diagnosis errors were reduced by 74% compared to the Snellen test. The simulations use a known acuity score and factors in the different errors humans make while they take the test. Afterward, it virtually gathers the various eye exams to compare how accurate they are. The team used this method instead of real patients because it begins with the true acuity, which is something not known in a human.
Room lighting or screen brightness would also need to be taken into consideration because those two conditions could affect the scores of the eye exam. The StAT test could potentially replace the Snellen chart, but it would be challenging to convince eye care professionals to agree to a new standard. However, it could also be adopted by more than 80% of ophthalmology practices.
Have a story tip? Message me at: cabe(at)element14(dot)com