AI Tool Can Scan a Retina, Predict Heart Disease Risk in Under a Minute

Doctors have long been able to assess a person’s risk of high blood pressure, diabetes, and other diseases by manually looking at the retina. Some existing tests similarly use images of someone’s retinal vasculature to detect whether they’re currently experiencing a disease. However, human observation lacks the precision necessary to determine a person’s risk of developing (or worse, dying from) heart disease later in life, while existing predictive tests require extra steps like blood pressure measurements and blood tests.

QUARTZ can reportedly fill this gap. A team of researchers from Cambridge University, St George’s University of London, and University College London created the tool by obtaining images of 95,463 participants’ retinas, which came from UK Biobank (a voluntary biomedical database) and the European Prospective Investigation into Cancer (EPIC). They used these images to train a machine learning algorithm to analyze the width and tortuosity of the retinas’ blood vessels. Then the team combined these analyses with the participants’ known cardiac outcomes to create a heart risk prediction model.

Now the researchers can scan anyone’s eye and obtain key cardiac insights in a matter of seconds. In their study published last week in the British Journal of Ophthalmology, the researchers write that a person’s retinal vasculature snapshot, combined with risk factors like smoking, previous heart attacks, and drugs to treat high blood pressure, can be used to predict that person’s risk of cardiovascular disease at the same accuracy rate as current, less convenient methods.

As we mentioned last week in our report on the White House’s proposed “AI Bill of Rights,” machine learning algorithms are as biased as the data they’re based on, and QUARTZ is no exception. The Verge pointed out that UK Biobank’s data happens to be 94.6 percent white, meaning there’s little room to determine whether QUARTZ predicts heart disease risk as accurately among people of color. The researchers also note in their study that UK Biobank and EPIC data are from a healthier population than is representative of the UK (and the world) at large.

But it’ll be a while before QUARTZ finds its way into real-life medical settings. As with any other novel medical tool or system, QUARTZ requires rigorous testing before it can become eligible for use on regular patients. The team hopes to confirm the tool’s efficacy with a wider, more diverse population during this time.

Continue reading

Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption

Intel's higher-end desktop CPU TDPs no longer communicate anything useful about the CPUs power consumption under load.

Analysts Predict Rapid DDR5 Adoption by 2023

Analysts are predicting a fast ramp for DDR5 when it debuts late this year — much faster than the DDR3-to-DDR4 transition.

New Study Predicts Teeny Tiny Mountains on Neutron Stars

A new analysis of neutron star physics predicts the "mountains" on the surface are less than a millimeter tall.

Predictive Policing Software Shown to Entrench Bias, not Address It

A new analysis by Gizmodo has concluded that such software disproportionately targets poor communities and communities of color.