Arrhythmias & electrophysiologyCohort study
Accuracy of a Deep Learning Model in Intracardiac Echocardiography
- retrospective multicenter study
Authors' conclusion
Auto-Contour demonstrated robust multistructure segmentation of ICE anatomy with real-time inference, supporting prospective evaluation of artificial intelligence-assisted ICE for procedural standardization, efficiency, and safety.
Study context
- Population
- Deidentified intracardiac echocardiography (ICE) cine loops from procedures of unique patients, including routine clinical cases and the ViewFlex™ X first-in-human studyprocedures of unique patients, 5,496 ICE cine loops, 65,117 segmentations
- Participants
- 249 participants
- Exposure
- Auto-Contour deep-learning segmentation of ICE anatomy against ICE expert annotations
Main finding
As reported in the abstract
Segmentation performance, larger cardiac chambers
Dice score, left atrium
0.94
Dice score, left ventricle
0.82
95th-percentile Hausdorff distance, left atrium
1.18 mm
95th-percentile Hausdorff distance, left ventricle
3.27 mm
Observational data: associations, not effects.
Further findings
Segmentation performance, smaller structures
- Acceptable performance for the left atrial appendage, pulmonary veins, papillary muscles, and aortic cusps
Mean per-frame inference time
- Mean per-frame inference time
- <0.03 seconds
Original publication
Accuracy of a Deep Learning Model in Intracardiac Echocardiography. JACC Adv. Published 15 June 2026. doi:10.1016/j.jacadv.2026.102867
Summarised from the published abstract; every number is checked against it