- Google’s new AI scans a person’s retina to detect risk of heart disease
- The deep learning algorithm was trained on data from 284,335 patients
- Validated on two independent datasets and has an accuracy of 70%
Here’s yet another potential breakthrough in the field of healthcare. And Google is at the forefront again.
Google has developed an AI algorithm that scans an individual’s retina and tells them if they’re at risk of contracting a heart disease. According to an article published in Nature journal Biomedical Engineering, Google researchers trained a deep learning algorithm to predict cardiovascular risk by using data points from 284,335 patients, which included retina scans and medical data.
To test the algorithm, the deep learning model was applied on two independent datasets of 12,026 and 999 patients. The model was shown images of the retina of individuals who suffered a heart attack within five years of the study and those who did not have a cardiac event. The algorithm was able to predict heart attacks and cardiovascular events 70% of the time, a good accuracy but not good enough to be used for any practical purpose yet.
This result is similar to testing methods using a patient’s blood. Google AI can not only predict heart disease, but also the likelihood of a cardiovascular event, such as a heart attack or stroke. Additionally, the model can tell an individual’s age, blood pressure, and whether or not the patient smokes.
Google’s Lily Peng, an MD and lead researcher on the project, hopes that the AI can be applied to other areas of scientific discovery, including cancer research.
Our take on this:
The accuracy of the model is a bit low but the optimism is quite high about the future of this project. Ms. Peng admitted that the data used for this study was smaller than expected but as more data points are added, she expects the accuracy of the model to increase. Should the results improve, physicians worldwide will be able to run quick scans to detect any risks of heart diseases in a matter of seconds.
However, expectations should be tempered a bit at this point. Ms. Peng admitted that it’s a matter of a few years, rather than months, before this technology can be made available for practical purposes.
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
You can also read this article on our Mobile APP