Lo Russo, Gerardo

Objective Data on cardiac index (CI) in patients undergoing coronary angiography (CAG) can enhance decision-making and improve clinical outcomes. However, this necessitates an additional invasive procedure, and therefore, it is frequently omitted. This study explores the novel application of convolutional neural networks (CNNs) to predict CI from routine CAGs. Methods and analysis Patients who underwent CAG and simultaneous right heart catheterisation at Mayo Clinic between 2002 and 2023 were included. A three-dimensional (3D)-CNN model (X3D architecture) was developed to predict binary CI categories (normal: CI≥2.2 vs abnormal: CI<2.2 L/min/m²) from CAG videos. CAGs were randomly split into training (70%), validation (15%) and test (15%) sets. All available videos from each study were entered into the model. The model’s performance was evaluated using the area under the curve (AUC), sensitivity, specificity and F1 score. Results A total of 15 297 CAG studies were considered. After applying a quality-check algorithm to exclude low-quality or irrelevant CAGs, 11 475 studies from 9103 patients were included. CI was abnormal in 25.2% of studies. The model achieved an AUC of 0.83 (95% CI 0.81 to 0.85), sensitivity of 0.72 (95% CI 0.68 to 0.77), specificity of 0.78 (95% CI 0.75 to 0.80) and F1 score of 0.62 (95% CI 0.58 to 0.65) on the test dataset. Limiting the input to 4, 3 or 2 prespecified CAG videos representing standard views did not significantly affect the model’s performance, supporting its applicability in varied clinical workflows. Conclusions This study demonstrates the feasibility of leveraging deep learning models to predict CI from CAG videos. The proposed 3D-CNN model shows promise as a tool for providing physicians with real-time data to aid decision-making during CAG procedures.

Abstract Aims To assess the reproducibility of interpreting hypoattenuated thickening (HAT) and peridevice leak (PDL) using cardiac computed tomography (CT) imaging following Watchman FLX left atrial appendage closure (LAAC). Methods and results In this multicentre retrospective reproducibility study, 100 anonymized post-LAAC cardiac CT scans were evaluated within the same cardiac phase by an experienced and a novice rater blinded to prior evaluations. All scans were evaluated twice by each rater, assessing overall HAT and PDL categories as well as specific associated findings based on suggested algorithms for post-LAAC interpretation. Inter- and intra-rater agreement and reliability were evaluated using absolute agreement, Cohen’s kappa, and Kendall’s tau for categorical variables, and mean difference, Bland–Altman plots, limits of agreement (lower and upper), and intra-class correlation coefficients (ICCs) for continuous variables. Within overall categories of both HAT and PDL, substantial agreement (kappa ≥0.61) and reliability (Kendall’s tau-b ≥ 0.75) were observed. Specifically, identifying high-grade HAT (kappa ≥0.78) and distal patency (kappa ≥0.85) displayed the highest agreement within HAT and PDL interpretation. Meanwhile, measuring the height of the proximal screw hub cove represented the least reliable HAT assessment among both inter- and intra-rater comparisons (ICC &lt; 0.75), while suspected leak mechanism represented the least reproducible PDL measure. Conclusion Despite only minimal training of one rater, overall high levels of inter- and intra-rater agreement and reliability were observed across the chosen algorithms for interpretation of HAT and PDL following Watchman FLX LAAC. The prognostic implications of the included variables are to be explored in future trials and registries.

Background Acute ischemic stroke complicates 2% to 3% of transcatheter aortic valve replacements (TAVRs). This study aimed to identify the aortic anatomic correlates in patients after TAVR stroke. Methods and Results This is a single‐center, retrospective study of patients who underwent TAVR at the Mayo Clinic between 2012 and 2022. The aortic arch morphology was determined via a manual review of the pre‐TAVR computed tomography images. An “a priori” approach was used to select the covariates for the following: (1) the logistic regression model assessing the association between a bovine arch and periprocedural stroke (defined as stroke within 7 days after TAVR); and (2) the Cox proportional hazards regression model assessing the association between a bovine arch and long‐term stroke after TAVR. A total of 2775 patients were included (59.6% men; 97.8% White race; mean±SD age, 79.3±8.4 years), of whom 495 (17.8%) had a bovine arch morphology. Fifty‐seven patients (1.7%) experienced a periprocedural stroke. The incidence of acute stroke was significantly higher among patients with a bovine arch compared with those with a nonbovine arch (3.6% versus 1.7%; P =0.01). After adjustment, a bovine arch was independently associated with increased periprocedural strokes (adjusted odds ratio, 2.16 [95% CI, 1.22–3.83]). At a median follow‐up of 2.7 years, the overall incidence of post‐TAVR stroke was 6.0% and was significantly higher in patients with a bovine arch even after adjusting for potential confounders (10.5% versus 5.0%; adjusted hazard ratio, 2.11 [95% CI, 1.51–2.93]; P <0.001). Conclusions A bovine arch anatomy is associated with a significantly higher risk of periprocedural and long‐term stroke after TAVR.