We are pleased to announce the pre-proof release of our study in Structural Heart: “Comparison of Experienced and Inexperienced Raters Using Automated Deep Learning CT Analysis to Evaluate Tricuspid Valve and Right Heart Morphology” (Mattig et al., doi: 10.1016/j.shj.2025.100488).

In 30 patients with severe tricuspid regurgitation, both seasoned radiologists and CT novices used the heart.ai v1 deep learning platform to measure annular dimensions, chamber heights and more. Despite differing backgrounds, raters achieved excellent agreement (ICC > 0.95) in just 1.8 minutes (experienced) versus 2.6 minutes (inexperienced) per scan. Unadjusted automated results already proved highly reliable, with optional manual tweaks available at any point.

Our findings demonstrate that AI-based CT analysis can deliver fast, precise heart morphology evaluations—regardless of user experience—and holds promise for simplifying the planning of tricuspid interventions in routine clinical practice.

Read the full pre-proof here: https://www.sciencedirect.com/science/article/pii/S2474870625000806