Google AI Experiments – Detecting Diabetic Retinopathy to prevent Blindness | Digit.in


So I’ll be talking about how we’re using
AI for Diabetic Retinopathy detection. Diabetic retinopathy sweetly can detect
sight threatening changes before it’s too late. And that’s why Ipax for 15
years now has been putting these retinal cameras in primary care clinics that
don’t have access to eye care. In order to be able to capture images like the one
you see here and detect diabetic retinopathy. So what happens with
diabetes? So in diabetes uncontrolled blood sugar damages the small blood
vessels in the back of the eye. And this is the back of the eye here, the retina
and that leads to the death nerves and receptors that’s called
diabetic retinopathy. So with the retinal image in the primary care clinic, those
images can be transmitted to eye care provider who then reads the image and
then provides a consultation. You can tell what if the retina is healthy or if
it’s damaged. This is a diabetic retina. So we’ve been doing this for 15 years and we have
millions of images. Those images were used as ground truth for article and
that’s Google’s deep learning algorithm that detects and raids
diabetic retinopathy with accuracy greater than 90 percent. That’s better
than what human eye doctors can do. And not only that but it gives it
instantaneous read, instantaneous measure of the level of
diabetic retinopathy. Instead of having to wait hours or days, you know the
clinics can get an immediate measurement and know what to do with the patients. So
Arta is currently just used in clinical trials. So but it’s going to be a
game-changer when it’s out there.
AI has also helped us with other other things for example with image quality. We
have a big problem with that. We have to train and retrain the staffs. These are
the clinics that take the pictures and it’s really kind of a tedious
task. So now with Google’s alogrithm adapted, the photographers could get
immediate feedback and know if an image is too dark, too light, its blurry, is
my position right. And and even if there’s a smudge on the lens. So that’s
been extremely helpful. But what’s been even more interesting to us is the
things that AI can do. These are the pictures of the screening programs. Anyway it’s we know that that the
algorithm to see things that humans cannot see. And you can be able to tell
if there is cardiovascular risks, risk of a heart attack or stroke, neurological
conditions and that opens up a lot of possibilities. And we envision the
retinal image being maybe part of the vitamins test. When a patient comes
to the doctor, you get the blood pressure, weight and then also a retinal image. So this is the grading algorithm that we talked about. So another problem too is that
vision impairment is projected to triple by the year 2050. There’s not enough eye
care providers to be able to keep up in already burdened health care system. So
we need to do something to prevent this vicious cycle of blindness of people
getting to the eye doctor when it’s too late. And when the treatments don’t work
anymore. So we’re really excited to be working with Google
apps that truly make a difference on the front lines. Thank you.

Add a Comment

Your email address will not be published. Required fields are marked *