Artificial intelligence is able to detect diabetic retinopathy more effectively than ophthalmologists and specialists, according to research conducted by seven MDs and PhDs published in the Sept. 29 edition of the Ophthalmology Science Journal.
The study reviewed 521 patients and 999 eyes from 10 different eye care centers in the U.S.
Retina specialists had a 59.5 percent sensitivity rate, correctly identifying 22 of 37 eyes as positive for more than mild diabetic retinopathy, and 182 of 184 as negative in the set they were given.
The EyeArt AI system had a 97 percent sensitivity rate, correctly identifying 36 of 37 positive eyes and 162 of 184 negative eyes in the same set.
General ophthalmologists had a 20.6 percent sensitivity rate, correctly identifying 35 of 170 eyes as positive and 607 of 608 eyes as negative in a different set.
On that set, EyeArt AI had a 96.5 percent sensitivity rate, correctly identifying 164 of 170 positive eyes and 525 of 608 negative eyes.