Google is kind of an insane corporation, isn’t it? I mean, what don’t they do? Now, they are trying to come up with a way to fight cardiovascular disease, but they’re doing it through the use of machine learning. In a paper published in Nature’s Biomedical Engineering Journal, it’s been revealed that Google’s health subsidiary Verily detailed a method to predict risks of heart diseases. How exactly? Well, they are going to predict the risk of these disease by scanning the rear interior wall of your eye.
I just had an eye exam last week, and they scanned the rear interior of the eye. Maybe not the interior wall, but at least the back of my eye. So I wonder how this will be different. To start, my eye doctor was scanning the back of my eye to determine if I had any eye diseases specifically. I think this is all very fascinating. But how will they take this same information to determine other types of diseases?
How they’re doing this is both interesting, and also kind of simple if you think about it. When photographed, with the aid of a microscope and camera, that portion of your eye, known as the fundus, can give doctors an idea about the patient’s age, blood pressure, cholesterol levels, and whether or not they smoke. All of these can factor into whether you are likely to suffer from heart disease. Which makes sense, in a way. I mean, if they’re able to “see” all of that, just by scanning the back of your eye, it’s just a short jump to determine if you’re predisposed to certain heart diseases. All of these things tell a doctor if you’re unhealthy, but that requires the use of a blood test.
Where this gets really interesting though is how Verily is going to be able to detect this data. Verily is going to use machine learning to look at the data of nearly 300,000 patients. Of the data, they looked at one image of a person who suffered a cardiovascular event in the five years following the image, and they looked at one who didn’t. They were able to identify this correctly 70% of the time. This is comparable to alternate methods that require blood tests, in terms of accuracy. 70% seems low, but you have to take into consideration how machine learning works. It takes that information and teaches itself when it can identify these cases. I think it won’t take any time at all for that number to get into the high 90s.
Can you imagine though? Instead of having to have invasive testing done (or multiple tests) in order to determine a particular disease or condition, you may be able to get the same results using machine learning. A lot of people are scared of what AI could do in the future, and I think we should all be somewhat concerned. But at the same time, AI has the ability to bring us some pretty amazing technology that could save lives. So what should we be doing?
I think that we should be exploring how these technologies can actually change our world for the better. We are constantly living in fear of robots and while I think that there is some crazy stuff that could happen, I think we should focus on how the technology can help us and improve our lives. In this case, it might be saving lives by giving us valuable insight into what kinds of treatable health conditions we might be living with.