January 01, 2020

Detecting heart failure risk with artificial intelligence

The left ventricle provides most of your heart's pumping power. A decline in this pumping power — called left ventricular dysfunction — is a potential precursor to heart failure. Sometimes, this decline occurs without causing any noticeable symptoms early on. This is a difficult-to-detect condition known as asymptomatic left ventricular dysfunction (ALVD). It's present in an estimated 9% of older adults, and treating it can help prevent heart failure.

New research from Mayo Clinic cardiologists — published in Nature Medicine — shows how artificial intelligence may increase the ability to accurately diagnose and treat the condition before heart failure can occur. A large set of electrocardiogram (ECG) data was used to develop a type of artificial intelligence with which patterns could be recognized. A set of rules (algorithms) developed from this data had a high level of accuracy in detecting people with left ventricular dysfunction as confirmed by follow-up testing.

Notably, one subset of study participants whose ECGs were flagged as abnormal by the algorithms were found to have normal ventricular function in follow-up testing. However, in the years after, this "false-positive" group was found to have a fourfold risk of developing future left ventricular dysfunction. This suggests that the algorithms may be able to detect very early abnormalities.

While this use of artificial intelligence is not part of standard medical practice, Mayo Clinic experts say it could one day be an easy, inexpensive screening tool for ALVD, with the potential to reduce heart failure risk.