Artificial intelligence can identify risk of familial hypercholesterolemia

Chloe Kent 12 April 2019 (Last Updated December 23rd, 2019 10:23)

Researchers at Stanford University School of Medicine have developed a new artificial intelligence algorithm that can identify whether or not a patient is at risk of developing familial hypercholesterolemia, a cholesterol-raising genetic disease, with up to 88% accuracy.

Artificial intelligence can identify risk of familial hypercholesterolemia
The algorithm can detect the presence of the disease with up to 88% accuracy. (Credit: Shutterstock)

Researchers at Stanford University School of Medicine have developed a new artificial intelligence (AI) algorithm that can identify whether or not a patient is at risk of developing familial hypercholesterolemia (FH), a cholesterol-raising genetic disease, with up to 88% accuracy.

The disease hinders the ability of the body to clear harmful low-density lipoprotein (LDL) cholesterol. It can cause early and potentially fatal heart problems, but is often misdiagnosed merely as high cholesterol.

This oversight can be serious, as FH patients are three times more likely to develop early heart disease than patients who present with just high cholesterol. Around 30% of women with FH will have a heart attack by the age of 60 without appropriate treatment, and the rate rises sharply to 50% by the age of 50 in men.

Stanford University assistant professor of cardiovascular medicine Joshua Knowles said: “We think that less than 10% of individuals with FH in the United States actually know that they have it.”

The programme was developed using data from 197 patients who had FH and 6,590 who did not, allowing the AI to learn the difference between the two. The researchers trained the algorithm to pick up on a combination of family history, current prescriptions, lipid levels, lab tests and more to understand what details signal the disease.

The algorithm was then run through a set of 70,000 anonymous patient records it hadn’t encountered before. The research team reviewed 100 of these patient charts and from these extrapolated that the algorithm had detected FH patients at 88% accuracy.

They then teamed up with the Geisinger Healthcare System to test the algorithm on 466 FH patients, and 5,000 non-GH patients. Predictions in this second round of testing came back with 85% accuracy.

The software could be implemented in doctor’s offices to screen family members of FH patients to see if they have inherited the mutation, alongside patients presenting with potential symptoms.

Stanford University associate professor of medicine and of biomedical data science Nigam Shah said: “Theoretically, when someone comes into the clinic with high cholesterol or heart disease, we would run this algorithm.

“If they’re flagged, it means there’s an 80% chance that they have FH. Those few individuals could then get sequenced to confirm the diagnosis and could start an LDL-lowering treatment right away.”

The research was published online on 11 April in npj Digital Medicine.

The project was funded by the American Heart Association and Amgen, and researchers from Atomo Health in Texas, the University of Pennsylvania, Yale University and Georgia State University contributed to the study.