Predictors of residual tricuspid regurgitation after interventional therapy: an automated deep-learning CT analysis

Abstract

AbstractComputed tomography (CT) is used as a valuable tool for device selection for interventional therapy in tricuspid regurgitation (TR). We aimed to evaluate predictors of TR reduction using CT and automated deep learning algorithms. Patients with severe to torrential TR and CTs prior to either percutaneous annuloplasty (PA) or tricuspid transcatheter edge-to-edge repair (T-TEER) were enrolled. CTs were analyzed using automated deep learning algorithms to assess tricuspid valve anatomy, right heart morphology, and function. Outcome parameters comprised post-interventional TR ≤ 1 and all-cause mortality. 84 patients with T-TEER (n = 32) or PA treatment (n = 52) were enrolled. Patients with a post-interventional TR ≤ 1 presented lower tenting heights and smaller tenting angles compared to patients with a TR > 1. Tenting height showed the best accuracy for post-interventional TR > 1 with an AUC of 0.756 (95% CI 0.560–0.951) in the T-TEER and 0.658 (95% CI 0.501–0.815) in the PA group, consistent with a suggested threshold of 6.8 mm and 9.2 mm, respectively. Patients with a post-interventional TR ≤ 1 exhibited a mortality of 4% and those with a TR > 1 of 12% during a follow-up of 331 ± 300 and 370 ± 265 days, respectively (p = 0.124). To conclude, tenting is associated with procedural outcomes and should be considered during screening for interventional TR therapy.