An innovative new way to analyze electrocardiogram (ECG) results using artificial intelligence may help doctors quickly and cheaply identify patients most at risk of heart failure. A new study from researchers at the Mayo Clinic demonstrates a newly developed AI system that is able to detect the presence of left ventricular dysfunction, a condition currently only detectable using other more expensive and time-consuming imaging tests.

Asymptomatic left ventricular dysfunction (ALVD) is a major precursor to heart failure, however the condition doesn't generally display prominent signs or symptoms until it is too late. It's estimated that between three and six percent of the general population suffer from ALVD but screening for the condition is difficult. The best diagnostic tools currently available are imaging tests, including CT scans and echocardiograms. A blood test is also available to detect ALVD but the results have been inconsistent.

The Mayo Clinic team set out to see if a well trained neural network could identify ALVD using simple ECG data. ECG tests are cheap, fast and widely available in many doctor's clinics.

"The ability to acquire an ubiquitous, easily accessible, inexpensive recording in 10 seconds – the EKG – and to digitally process it with AI to extract new information about previously hidden heart disease holds great promise for saving lives and improving health," says Paul Friedman, senior author on the new study.

The system was trained on data from over 600,000 patients, pairing ECG and echocardiogram results. The resulting algorithm was subsequently tested on an independent dataset comprised of over 50,000 patients. The results were exceptionally positive, with sensitivity, specificity, and accuracy all sitting around 85 percent. These detection rates were noted as comparable to many other popular screening tests, including breast cancer mammographies and prostate cancer antigens.

Interestingly, the system also successfully picked up patients more likely to develop ventricular dysfunction in the future. Subjects returning a positive AI result, but without other detectable symptoms turned out to be four times more likely to develop ALVD in the future compared to those subjects receiving a negative result.

"In other words," explains Friedman, "the test not only identified asymptomatic disease, but also predicted risk of future disease, presumably by identifying very early, subtle ECG changes that occur before heart muscle weakness."

Applying AI tools to conventional medical imaging systems is quietly revolutionizing modern diagnostic services. From cancer to dementia, these new systems are able to quickly and effectively detect patterns in imaging data that slip past the eye of human experts. While many of these tools are not in common use yet, they undoubtedly will be soon, and this new ECG system will hopefully add to the ways doctors can catch deadly diseases before they progress.

The new research was published in the journal Nature Medicine.

Source: Mayo Clinic