Google Artificial Intelligence Diagnoses a Disease by Photograph

Artificial Intelligence and Machine TechnologyLearning captures more and more new areas of human activity. AI demonstrated high efficiency in the diagnosis of skin diseases using photographs. Specialists trained the Google neural network to identify skin diseases solely in appearance, without the tedious collection of tests. An accurate diagnosis can be made after studying an AI photograph of the affected area of ​​the skin.

Specially trained AI is able to diagnosenumerous skin diseases, including psoriasis, melanoma, acne. Diagnostic accuracy is within 93%, and AI studies a simple photograph of the affected part of the body.

During pre-training, the neural network“Got acquainted” with 17,777 images of skin areas with identified signs of the disease. AI “read” over 50 thousand diagnoses made by more than 40 dermatologists. The AI ​​algorithm examines a photograph of an infected area of ​​the skin and selects the three most likely diagnoses.

For more AI informationstudies up to 45 different types of metadata, including medical history, symptoms, age. Google Health experts are working to improve the accuracy of diagnostics, which should soon be equal to the accuracy of highly qualified dermatologists.

Source: googleblog