Scientists Use Artificial Intelligence To Prevent Blindness Due To Diabetes [VIDEO]By Khaleb Skye A. Cruz, UniversityHerald Reporter
Researchers at Stanford University discover ways to incorporate artificial intelligence in the fight against diabetes complications. In particular, robots can now prevent blindness and treat diabetic patients.
Published in the journal from the "American Academy of Ophthalmology", the experts at Stanford's Byers Eye Institute formulate plans to potentially reduce the worldwide rate of vision loss through deep-learning methods. For one, an automated algorithm was created to detect diabetic retinopathy (DR). DR, meanwhile, is the condition wherein blood vessels at the back of the eye are damaged.
Theodore Leng, M.D., as reported on Science Daily, said that artificial intelligence can identify which patients need to be referred to an ophthalmologist for further evaluation and treatment. The interesting part is that the robot can do it with "high reliability". Leng, the lead author of the study, added that if properly implemented, the new technology will have the most impact in places experiencing shortage in ophthalmologists.
Meanwhile, the digital assistant does not require any specialized, inaccessible, or costly equipment to function. It can operate on a typical personal computer or even on smartphones with average processors. Dr. Leng's algorithm is able to determine all disease stages, from mild to severe with an accuracy rate of 94 percent.
Diabetes, according to E&T, affects over 415 million worldwide or one in every 11 adults. Roughly 45 percent of patients are likely to lose sight at some point in their lives. Sadly only a few of them are aware of their condition.
Ophthalmologists normally examine diabetic blindness by directly looking at the back of the eye or by evaluation color photographs of the fundus. Given the rate of diabetes on Earth, this process is very time-consuming and expensive. Thus, asking for help from AI's will definitely not hurt. For the record, Dr. leng's team based the algorithm on more than 75,000 images of different patients, and taught a computer to identify who are healthy and who are not.