Artificial Intelligence Model TARGET-AI Enhances Efficiency of Heart Scans

Developed by Oikonomou and colleagues, the TARGET-AI model was designed to examine whether electronic health records can guide the selective application of artificial intelligence-supported EKG screenings. A retrospective evaluation showed that the model reduced false positive results in the detection of heart diseases and improved the balance of precision.

Within the scope of the study, 754,533 EKG-echocardiography pairs belonging to 159,322 adult individuals were retrospectively examined. The targeted screening approach was compared with the method of screening all patients with an existing EKG. The results revealed that the TARGET-AI model improved the balance of precision and sensitivity across 26 structural heart disease labels, while also significantly reducing false positive results.

According to the data obtained, the model achieved an AUROC of 0.90 for left ventricular systolic dysfunction and 0.85 AUROC for severe aortic stenosis. Targeted use provided a median increase of 0.25 in F1 scores per label while decreasing false positive results by a median of 303. The research team stated that pre-selecting candidates using the artificial intelligence model could increase screening efficiency, while emphasizing that prospective evaluations are necessary to clarify potential improvements in patient outcomes.