Prospective evaluation of the learning curve and diagnostic accuracy for Pre-TAVI cardiac computed tomography analysis by cardiologists in training: The LEARN-CT study

Background: To investigate the learning curve and the minimum number of cases required for a cardiologist in training to acquire the skills to an accurate pre-TAVI cardiac CT (CCT) analysis using a semi-automatic software. Methods: In this prospective, observational study, 40 CCTs of patients scheduled for TAVI were independently evaluated twice by 5 readers (80 readings each, 400 in total): a certified TAVI-CT specialist served as the reference reader (RR) and 4 cardiology fellows (2 interventional and 2 non-invasive cardiac imaging) as readers. The primary outcome was the minimum number of cases required to achieve an accuracy in imaging interpretation ≥80%, defined as the agreement between each reader and the RR in both balloon and self-expandable valve size choice. The secondary outcomes were the intra- and inter-observer variability. Results: After 50 readings (25 cases repeated twice) cardiology fellows were able to select the appropriate valve size with ≥ 80% of accuracy compared to the RR, independently of valve calcification, image quality and slice thickness. Learning curves of both interventional and non-invasive cardiac imaging fellows showed a similar trend. Cardiology fellows achieved a very high intra- and inter-observer reliability for both perimeter and area assessment, with an intraclass correlation coefficient (ICC) ranging from 0.96 to 0.99. Conclusions: Despite the individual differences, cardiology fellows required 50 readings (25 cases repeated twice) to get adequately skilled in the pre-TAVI CCT interpretation. These results provide valuable information for developing adequate training sessions and education protocols for both companies and cardiologists involved.