DeepMind, an artificial intelligence company from Google, has developed an AI system that accurately identifies 50 types other than eye diseases. The system, which is capable of analyzing 3D scans of the retina to detect precursors of glaucoma, diabetic eye disease and macular degeneration, has been developed through research collaboration with the Moorfields Eye Hospital in London during the past 18 months. AI, capable of 94.5% of the time being able to accurately identify different types of diseases through OCT analysis, has learned to discover eye diseases by studying 15,000 anonymous eye examinations. The results of the trial were published in the journal Nature Medicine on Monday.
DeepMind co-founder Mustafa Suleyman, head of DeepMind Health, said on the internet that this AI system can help save people's eyes, adding that one day it could be deployed in hospitals around the world  "Our AI system can quickly and with unparalleled accuracy interpret the most common x-ray of the eyes." He can recommend different treatments to patients, based on 50 ocular diseases that threaten their vision, and with as much precision as the greatest specialists " he wrote on the blog DeepMind.
And to add: "These are only preliminary results, but they already show that our system can handle a wide range of patients in routine examinations." In the long term, we hope that doctors will give priority to patients who need urgent treatment. "
DeepMind intends to leave NHS hospitals (before National Health Service ) use five years a free validated version of this technology, provided the software meets all necessary legal approvals.
Dr. Pearse Keane, an ophthalmologist in Moorfields, London, said in a press release: "The number of eye scans we perform is growing too fast for human experts to be able to interpret them all in time.This can lead to delays in the diagnosis and thus delays in the treatment of diseases that threaten the eyesight of patients. ".
"If we can diagnose and treat ocular problems quickly, this gives us better opportunities to save the sight of a patient, with further research to ensure the consistency and quality of care for patients with eye problems in the future. can improve " he said.
DeepMind's work with the NHS started badly after the Information Commisses Office (ICO), the UK Data Manager, ruled that an RN test with the Royal Free NHS Foundation Trust was not in accordance with the law on data protection. In concrete terms, the ICO reported that approximately 1.6 million sensitive patient files were incorrectly used and that patients were not adequately informed that their data would be used for this test.
But DeepMind sees health care as a field that will bring about a revolution and probably sees the NHS as an important source of income, although it will never openly admit it. He already accuses some NHS organizations of having access to various DeepMind products and eye-scanning technology can become very lucrative if it is used on a large scale.
Google reportedly paid GBP 400 million to purchase DeepMind in 2014, but the aggressive expansion of the AI lab means it is currently working at a loss, according to the latest documents filed with Companies House. In 2016, the company recorded a total loss of £ 164 million, a significant increase from the loss of (only) £ 54 million in 2015.
The majority of DeepMind's expenses were dedicated to recruiting as many worldwide as possible researchers in the field of machine learning. In 2016, while employing approximately 600 people, he spent £ 104.8 million for [Personnel costs and other related costs] .
A report Published by The information last April indicated that some Google members were not satisfied with the lack of revenue from DeepMind.
However, it should be noted that DeepMind has helped the research giant in a number of areas. In July 2016 DeepMind announced that it has developed software that allows Google to reduce the amount of electricity used by cooling units in its data centers by 40%. DeepMind IA has also helped Google make its speech synthesis service more realistic.