Smart health DeepMind can detect 50 ocular pathologies

March 31, 2019

Smart health DeepMind , Google’s Artificial Intelligence, can now detect 50 eye diseases as accurately as a doctor.

Millions of anonymous data from Moorfields Eye Hospital in London will be used by Google’s artificial intelligence to detect visual impairments.

As we know, medicine is one of the main fields of application of new technologies in the coming years. Because it represents a real gold mine for companies but also because this sector fascinates researchers. New example with DeepMind, the artificial intelligence of Google and its use in ophthalmology.

Google’s Artificial Intelligence and the British National Health Service (NHS) announce a new collaboration at Moorflieds Ophthalmic Hospital in London. Objective: to detect early warning signs of visual impairment and other eye diseases.

Smart health DeepMind learned from 3D scans

94% accuracy. This is the very good figure obtained by Google’s artificial intelligence in the detection of 50 eye diseases. The least we can say is that DeepMind is a multiform AI and very diverse features. DeepMind has been nurtured thanks to a very large database. 3D scans of more than 7500 patients. That’s 15,000 eyes and the diagnoses related to them.

To do this, DeepMind will use the technique of machine learning – or machine learning. More than a hundred anonymous images will be analyzed to generate an algorithm capable of identifying two very particular anomalies.

The first, referred to as “age-related macular degeneration,” is a disease that affects the retina and results in a progressive loss of central vision.

“98% of visual acuity loss could be avoided”

The second, diabetic retinopathy, can occur during diabetes-related complications and impact the retinal vessels to the point of loss of sight.

With imaging and eye scans, Google has promised that anonymous data does not allow any personal identification. In a question / answer form available on the Moorfields Hospital website, DeepMind even states that “patients will be able to withdraw their data from the sharing system by contacting the health service”.

Tracked in time, these diseases can still be cured, and that’s the challenge of DeepMind’s mission. “If you have diabetes, you are 25 times more likely to become blind, and if you can detect that, 98% of visual acuity loss can be avoided.”

A learning work that is bearing fruit since DeepMind is now particularly accurate. 3D technology scans the retina, the AI ​​records mapping and Google’s algorithm, coupled with optical coherence tomography, establishes a diagnosis. However, artificial intelligence also goes further in explaining its diagnosis and provides patients with advice on the most appropriate care for each disease. The AI ​​even allows to assign a percentage of confidence to the various treatments offered. We can imagine that the relationship is not yet its strong point, but when it does, ophthalmologists can return home.