UPV AI Model Hits 89% Accuracy in Atrial Abnormality Detection

Researchers at Spain's Universitat Politècnica de València (ITACA) developed a graph neural network model that uses surface electrical recordings to localize atrial cardiomyopathy tissue abnormalities, achieving 89% accuracy in localization and 84% in damage extent, including on unseen anatomical structures. Published in Discover Computing, it remains a proof of concept on simulated data and needs validation on real patient records before treatment-planning use.