How Google is using AI to Prevent Blindness in Diabetics

0
719
google ai to prevent Blindness in Diabetics
Image: Google Stories

Artificial Intelligence is in its development stage, and at many times we find it hard to find the best application of the technology for the betterment of human society. I think it is similar to the situation faced by us 70 years back from now when we didn’t know that what to use a computer for.

Many companies including Google are trying to find ways in which AI can be utilized in the best way possible to help simplify the challenging problems faced worldwide.

This story was recently featured on the homepage of Google. It shares a story about how a team at Google is using AI to help doctors prevent blindness in diabetics.

Google research scientist Varun Gulshan started this project in his home country India.

The Problem

diabetic

With seventy million folks with diabetes, India encompasses a growing drawback with diabetic retinopathy. It creates lesions within the back of the membrane that may result in total visual impairment.

Eighteen per cent of diabetic Indians already suffer from it. With 415 million diabetics in danger for visual impairment worldwide, it may be a world concern.

But the great news is that permanent vision loss isn’t inevitable. For those that square measure diagnosed early enough, medications, therapies, exercise and a healthy diet square measure extremely effective treatments for preventing additional harm.

The Tricky Part

AI diabetic retinopathy
Image: Google Stories

If Google’s artificial intelligence may facilitate build designation diabetic retinopathy easier by accurately deciphering retinal scans, then maybe the visual modality of millions may be saved.

The difficult half was making a data set for the AI model to train on a task that involves grading and labelling all the scans one by one for various grades of severity.

Finding that downside would eventually need an outsized team of ophthalmologists whose grading of the scans would inform the AI model. And the team also requires quality data to provide more accurate information.

Training the model

The team was assisted by ophthalmologists at Aravind and Sankara Nethralaya to label the membrane pictures. when a couple of short months, the model was trained to spot key markers of diabetic retinopathy, like nervous tissue injury, swelling and haemorrhaging.

And with a bigger knowledge set, Gulshan was positive that they might build the model even additional correct.

They were successfully able to train the model and make a device from it.

How the Google AI Model Works to Detect Diabetic Retinopathy

ai model traning
Image: Google Stories
  1. Over 50 ophthalmologists have manually reviewed more than 1 million anonymous retina scans, rating each for the level of diabetic retinopathy present.

  2. Each scan is reviewed multiple times and is graded manually on a scale of 1 (no diabetic retinopathy signs present) to 5 (extreme signs present).

  3. Each scan is reviewed multiple times and is graded manually on a scale of 1 (no diabetic retinopathy signs present) to 5 (extreme signs present).

  4. Once the algorithm has been trained, it can be used to power an application called an Automated Retinal Disease Assessment (ARDA). ARDA allows a user to upload a retina scan for instant analysis of diabetic retinopathy.

Verdict

Getting a diabetic retinopathy diagnosis will cause higher outcomes overall. ‘If you notice retinal unwellness at associate degree early stage after they don’t would like treatment’, Dr. Cuadros says, ‘it’s still a chance for the patient to know that polygenic disorder is starting to have an effect on their body.

This was just one of the ways in which artificial intelligence could be used to solve compelling problems faced by millions of people. Visit Google stories for more such stories.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.