BACKGROUND New-onset conduction disturbances (CDs) remain the most frequent and clinically relevant complications after transcatheter aortic valve replacement (TAVR). The anatomical mechanisms underlying conduction injury are not fully understood. This study aimed to investigate the association between inferoseptal recess (ISR) anatomy and post-TAVR conduction disturbances, and to validate its predictive value using an interpretable machine learning algorithm. METHODS This retrospective study included 307 patients who underwent TAVR between 2017 and 2024. Pre-procedural computed tomography (CT) was used to quantify ISR morphology and aortic root anatomy. The primary endpoint was new-onset CDs (bundle branch block, high-grade atrioventricular block, or need for permanent pacemaker implantation). Machine learning algorithms were applied to integrate CT-derived anatomical and clinical variables and to validate the predictive value of ISR metrics, with SHapley Additive exPlanations (SHAP) applied for interpretability. RESULTS CDs occurred in 57 of 307 patients (18.6%). Wider WOISR (per 1-mm increment) was independently associated with reduced risk of CDs (OR: 0.76, 95% CI 0.66-0.89; P<0.001), along with counter-clockwise aortic root rotation (OR: 0.91, [95% CI, 0.87-0.95]; P<0.001), smaller ΔMSID (OR: 0.83, [95% CI, 0.76-0.91]; P<0.001), larger left ventricular end-diastolic diameter (P=0.004), with pre-existing right bundle-branch block (P<0.001). Among ML models, the RF model achieved an AUC of 0.91 in the temporal holdout cohort and showed favorable net benefit across a range of clinically relevant threshold probabilities. SHAP analysis identified non-coronary sinus calcification score, LVEDD, and WOISR as dominant contributors to model output. CONCLUSIONS WOISR (≤4 mm) is an anatomical predictor of CDs following TAVR. WOISR was independently associated with post-TAVR conduction disturbances and provided incremental discriminatory information beyond established clinical predictors. These findings support further prospective evaluation of ISR morphology as a complementary component of CT-based conduction-risk assessment before and after TAVR.
