Through Google’s revolutionary AI system, researchers were able to successfully predict cardiovascular problems including heart attacks and strokes. The approach relies exclusively on retina images.
According to Google officials, the new detection method uses machine learning techniques instead of relying on existing diagnostics to predict cardiovascular issues.
The results of the Google AI research were published in the journal, Nature Biomedical Engineering.
According to the tech company, medical researchers have proven the link between the risk of a major cardiovascular episode and retinal vessels. By analyzing a retinal image, the Google AI was reportedly able to quantify this link and predict with 70 percent accuracy which patient would experience a heart attack or other major cardiovascular incident within five years.
The Google AI algorithm was fed data from 284,335 patients and validated the information on two independent data sets of 12,026 and 999 patients.
According to Lily Peng, MD, a product manager and lead researcher on the project, the algorithm was able to distinguish between the retinal images of a smoker and those of a non-smoker with 71 percent accuracy. This was 21 percent higher than what human experts could achieve.
Peng acknowledged that further data needs to be analyzed in order raise the algorithm’s accuracy.
“We think that the accuracy of this prediction will go up a little bit more as we kind of get more comprehensive data.” Peng said.
More so, the Google AI’s algorithm could delve even further and predict the systolic blood pressure within 11 mmHg on average for patients overall, this including those with high and low blood pressure.
The algorithm currently works by generating a graphical representation of data, called a “heatmap”, revealing which pixels in the retinal image were the most important for predicting a specific health issue. In the case of predicting blood pressure levels, the Google AI focused more on blood vessels.
Image Source: Pixabay