Scientists from the Ningbo Institute of Ophthalmology have created a model of artificial intelligence, which can differentiate between malignal tumors of the eye with 89% accuracy. The development is described in Research magazine.
According to scientists, eye tumors are often unnoticed in the early stages. They are easily confused with allergies or inflammation, as a result of which patients lose sight, and sometimes life. In addition, small cities are absent from qualified ophthalmologists who can recognize dangerous neoplasms.
Artificial intelligence is already used in diagnosing eye diseases, but its teaching requires large volumes labeled data that is difficult to collect in rare pathologies. The solution was offered by Professor Chrongen Lee scientists.
Scientists have created an OSPM model, which first trained on the 760,000 unubitized images of the eye of the eye received from the 10 China clinics. Then an algorithm was given 1455 images with approved diagnosis. As a result, the accuracy of the artificiality of malignant tumors has reached 89%. These figures have been preserved during the analysis of eye images made with different cameras.
The algorithm called OECM also works with ordinary digital photos. According to scientists, this makes remote scrolling possible. The person in the risk group can take photos of his eye and send the image to check. In addition, the model requires less labeled data for teaching, which is especially important to diagnose rare diseases in regions where specialized specialists are limited.
Development is already being tested in three scenarios, large hospitals to accelerate the primary diagnosis, in regional clinics and as a mobile application for self-test.
If the tests are successful, the technology can become available tool for early detection of tumors. However, it is not yet clear how accurate the model will be able to recognize the subtarics of rare tumors, which have not been in the teaching collection.
Translation of: Euromedia24.com
