Machines can detect diabetic retinopathy – a leading cause of blindness from diabetes – from retinal scans to the same degree of accuracy as ophthalmologists, according to a paper published by Google Brain.
Recent advances in machine learning and computer vision have boosted the potential for machines to aid in medical diagnosis.
Diabetic retinopathy (DR) is a disease that can affect people with Type 1 or Type 2 diabetes. The retina is made up of a thin layer of cells at the back of the eye, and converts light into electrical signals for the brain to pick up so it can perceive vision. It needs a constant supply of blood pumped into the blood vessels.
When the sugar content in the blood is too high, it can strain the blood vessels, causing them to burst, which damages vision. If the disease is not caught early on it can lead to blindness.
Ophthalmologists look for signs of hemorrhages in retinal scans, and grade the images to determine the severity of the disease.
Machines can do this too. A paper published on The Journal of the American Medical Association (JAMA) shows that Google Brain’s algorithm could identify DR with over 90 per cent accuracy.
A dataset of 128,000 retinal images was graded three to seven times by a panel of 54 ophthalmologists from the US and India, and was used to train the neural network to pick out signs of DR from the pixels in the images.
The algorithm was tested on two sets of data – one had 9,963 and the other had 1,748 images. The accuracy levels achieved by the algorithm were 97.5 per cent and 96.1 per cent for sensitivity and 93.4 per cent and 93.9 per cent for specificity respectively.
Although these are “exciting results,” Lily Peng, product manager and Varun Gulshan, research engineer at Google Brain, said “there is still a lot more work to do.”
“We are working with doctors and researchers to study the entire process of screening in settings around the world, in the hopes that we can integrate our methods into clinical workflow in a manner that is maximally beneficial. Finally, we are working with the FDA and other regulatory agencies to further evaluate these technologies in clinical studies,” they wrote in a blog post.
Google continues to be interested in applying machine learning for healthcare. DeepMind, Google’s British AI firm, plans to alleviate some of the pressures faced by the UK’s National Health Service by aiding clinicians in their jobs of diagnosing diseases. ®