Russian developers have taught AI to recognize visual defects, representatives of the Don State Technical University told Izvestia. Now the neural network can identify pathologies such as cataracts, glaucoma, myopia and age-related macular degeneration.
“The algorithm works as follows: the doctor uploads an image taken with a fundus camera (a device capable of photographing both the anterior segment of the eye and the fundus of the eye – ed.) into Windows applications. Then the artificial intelligence analyzes it within two to three seconds, after which the program issues a preliminary diagnosis,” said project author Mikhail Shurduk.
The program is made for personal computers with the Windows operating system. Other operating systems such as Linux and macOS are not currently supported.
The authors of the project claim that the accuracy of image recognition is up to 95%. The AI is built on the VGG-19 architecture – programmers use these libraries to train neural networks when working with images.
In the future, programmers want to improve the functionality of the software. To do this, they will continue to train the AI on other images with confirmed diagnoses, so that the neural network can recognize similar cases even more accurately.
In addition, the authors of the algorithm intend to expand the list of recognized diseases, as well as provide the ability to store existing images in the software.
Read more in the exclusive material from Izvestia:
Vigilant character: a virtual ophthalmologist was created in Russia
#virtual #ophthalmologist #created #Russia