Google AI is taking gadgets to a whole new level as the deep learning algorithm from it has the potential to change how heart conditions are diagnosed just with a picture of your eyes.
What if a simple photo of your eyeball could predict your chances of having a heart attack? That is what a new deep learning algorithm from researchers at Google and Verily, a data science and healthcare technology operation run by Google’s parent company Alphabet hopes to achieve.
The scientists are using a high-definition photograph of a patient’s eyeball and hoping the algorithm can predict patients’ cardiovascular risk factors, such as smoking status and blood pressure and chances of a future heart attack at the same accuracy level as traditional tests.
The technology was tested on data from nearly 300,000 patients, and the encouraging findings were reported Monday in the journal Nature. More so, Verily’s algorithm stands to make cardiovascular testing a whole lot more streamlined and less invasive.
How Does The Technology Work?
For starters, Verily’s A.I.-powered tool requires a high-definition photo of the blood vessels in the eyeball’s interior lining, known as the retinal fundus. You maybe surprised but your eyes actually hold an unbelievable amount of information about your overall health.
The tool is a deep learning algorithm, which means that it learns from images it has seen before and uses that knowledge to inform its predictions. It teaches itself to look for certain things in the blood vessels of the retinal fundus, then crunches that data to draw conclusions about the patient’s cardiovascular health.
Why Does It Matter?
At a typical visit to a cardiologist, your doctor will often look into your daily behaviour, diet, age, genetic history, ethnicity, and a whole list of other factors to determine if you’re at risk.
These factors let doctors assign a numbered score, which represents a patient’s likelihood of a cardiac event. The algorithm has the potential to go a step beyond the traditional methods.
The risk score could be a starting point, but getting a direct measure of your vascular health, should be able to allow the doctor to more precisely measure your risk of a heart attack.
At the moment, the algorithm is “on par” with current methods, but it has the potential to surpass them in predictive capabilities if it gets more data or in simpler terms, more pictures of eyeballs.
However, the equipment needed for a retinal fundus isn’t exactly common. A lot of hospitals the world over don’t have it but the process is also much simpler than a full CT scan or other methods of imaging blood vessels.
Scientists are nonetheless excited about the algorithm’s ability to create “attention maps” which highlight certain portions of the image that it used to make predictions, some of which doctors had not thought to analyze before. This lets doctors learn more about how the algorithm works, so human and machine can learn from one another.
While we do not know what the future holds, the future of this tech device will depend on how quickly the tech can get to market. Cardiovascular diagnoses could get a whole lot more streamlined, and more accurate which should make every heart doctor and patient happy.