Google AI and Diabetic Retinopathy, Hope lies Ahead

One of the primary health divisions at Google, Verily, has introduced a new series of projects around Diabetic Retinopathy. The primary project is an algorithm that’s focus is to help determine heart disease.

For those who aren’t aware, diabetic retinopathy is an eye disease when blood vessels are damaged. This happens when there are high blood sugar levels, and the blood vessels either swell or leak.

Testing Results

Verily believes that it can use deep learning algorithms to “increase the accuracy of diagnoses”. This has been backed up after using the algorithm on more than 284,000 patients.

However, the results did not just come from a singular database. Instead, more than 48,000 came from the UK Biobank database. Then, more than 236,000 results came from the EyePACS database.

The results of this testing revealed “surprisingly high accuracy” in two sets of datasets. This includes being able to distinguish between smokers and non-smokers with a success rate of 71%.

How does it work?

All of the work is done by simply scanning the patient’s eyes. Then the algorithm will determine various cardiovascular risk factors.

The factors that the algorithm takes into account include the following:

  • Age
  • Gender
  • Smoking
  • Blood Pressure

The post states that the algorithm was able to identify that the patient had suffered a major cardiovascular event. Additionally, it was able to successfully identify another patient that not experienced an event.

What makes this so exciting is that the accuracy of these tests nearly lineup with those of blood tests. The difference between scanning the eye and blood tests means that results could be provided much quicker.

To the future

There is also the hope that these tests could present results that differ from the standard ones. Mainly, in the realm of being able to accurately predict any cardiovascular disease that the patient is susceptible to.

Moving even further, this new algorithm hopes to help doctors be able to easily identify future issues that may arise. The “black box” provided by Verily aims to give doctors more confidence in these findings.

The work by Verily is not finished, and you won’t be seeing these arrive in the doctor’s office just yet. In the announcement, it’s stated that more work must be done “to develop and validate these findings”.

Verily aims to continue to the work on this newly founded algorithm. The company also hopes to find new ways to make this more useful for patients in a myriad of ways.

Would you be more open to seeing new methods for doctor’s to help identify issues? Especially if there is no need to draw blood and still get the same results.

Sound off in the comments below and let us know what you think about Verily’s findings.

SOURCEAI Google Blog
Previous articleAmazon’s Alexa to Compete with Apple’s Health Service offerings
Next articleDoes Google own a patent on Smart Diapers?
Sudz Niel Kar
I am a technologist with years of experience with Apple and wearOS products, have a BS in Computer Science and an MBA specializing in emerging tech, and owned the popular site AppleToolBox. In my day job, I advise Fortune 500 companies with their digital transformation strategies and also consult with numerous digital health startups in an advisory capacity. I'm VERY interested in exploring the digital health and fitness-tech evolution and keeping a close eye on patents, FDA approvals, strategic partnerships, and developments happening in the wearables and digital health sector. When I'm not writing or presenting, I run with my Apple Watch Ultra or Samsung Galaxy Watch and closely monitor my HRV and other recovery metrics.

1 COMMENT

  1. Great information and thanks for sharing. Well, I am an optometrist and practice near Illinois areas currently working as an Optometrist near Naperville and I also write about eyes health tips Diabetic retinopathy: Causes, symptoms, and treatments.

    If you have time then kindly go through my post and let me know should I add more points?

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.