Villa Etchegoyen, Cecilia

Objective To compare the diagnostic accuracy of four available automated electronic medical record (EMR) retrieval methods, including a large language model (LLM)-assisted workflow, against manual chart adjudication for identifying cardiovascular events. Design Retrospective diagnostic accuracy study. Setting Three sites within a single US tertiary health system. Participants Two adult cohorts with previously adjudicated cardiovascular outcomes were included. Cohort 1 included 2258 patients treated with immune checkpoint inhibitors, and Cohort 2 included 1426 patients who underwent transcatheter aortic valve replacement. Primary and secondary outcome measures The reference standard was clinician manual chart adjudication. Outcomes included ischaemic stroke or transient ischaemic attack, myocardial infarction (MI), heart failure (HF) exacerbation or hospitalisation and a composite major adverse cardiovascular events (MACE) outcome. Automated retrieval methods included International Classification of Diseases (ICD) codes, primary diagnosis, problem list and a zero-shot LLM workflow. Area under the (receiver operating characteristic) curve (AUC), sensitivity, specificity and net reclassification improvement were assessed. Results In Cohort 1, the LLM achieved the highest AUC for stroke (0.920; 95% CI 0.881 to 0.958), MI (0.938; 95% CI 0.905 to 0.971) and composite MACE (0.880; 95% CI 0.854 to 0.907), whereas ICD-based retrieval had a higher AUC for HF (0.882; 95% CI 0.845 to 0.918 vs 0.873; 95% CI 0.831 to 0.914). In Cohort 2, the LLM achieved the highest AUC for all evaluated outcomes: stroke (0.915; 95% CI 0.862 to 0.968), MI (0.928; 95% CI 0.839 to 1.000), HF (0.844; 95% CI 0.803 to 0.884) and composite MACE (0.862; 95% CI 0.829 to 0.895). In Cohort 1, differences in AUC between the LLM and ICD methods were not statistically significant across outcomes, whereas in Cohort 2 the LLM showed significantly higher AUC for stroke and composite MACE. Conclusion In this multisite retrospective validation study, the LLM-assisted workflow showed strong but context-dependent performance for identifying cardiovascular events from the EMR. Performance varied by outcome and cohort, and ICD-based retrieval remained competitive for some use cases. These findings support a complementary role for LLM-assisted extraction in retrospective cardiovascular outcomes research.

BACKGROUND: Accurate assessment of left ventricular outflow tract (LVOT) gradients is critical for hypertrophic cardiomyopathy management, yet Doppler-based measurements are technically demanding and require expertise. The objective of this work was to develop a multi-view deep learning model capable of classifying LVOT obstruction (>20 mm Hg) using routine 2-dimensional echocardiographic windows without reliance on Doppler imaging. METHODS: We trained and externally validated a cross-attention-based video-to-video fusion framework that integrated EchoPrime-derived video representations from 3 standard transthoracic echocardiographic views to classify LVOT gradients. RESULTS: Training was performed on a derivation cohort (N=1833) from a tertiary care system in the United States, with model performance evaluated on an internally held-out test set (N=275) and a Korean external validation cohort (N=46). Single-view baselines showed limited discrimination (external area under the receiver operating curves, 0.47–0.70). Conversely, the domain-specific foundational model (EchoPrime) achieved superior single-view performance (area under the receiver operating curves, 0.75–0.80 internal; 0.79–0.83 external), highlighting the importance of echo-specific pretraining and temporal modeling. The proposed multi-view fusion further enhanced predictive performance, with the late fusion model reaching an area under the receiver operating curve of 0.84 on the external cohort with significant population-shift. CONCLUSIONS: These results suggest LVOT physiology is encoded in routine 2-dimensional imaging and can be leveraged for clinically relevant gradient classification without Doppler input. The proposed artificial intelligence-guided strategy demonstrates substantial cost savings compared with the screen-all approach. By integrating complementary spatial-temporal information across multiple views, our approach generalizes robustly across populations and may enable real-time decision support, extend LVOT assessment to portable or resource-limited settings, and complement Doppler-based evaluation for longitudinal hypertrophic cardiomyopathy management.

Background:The development of symptoms or left ventricular dysfunction in severe aortic stenosis (AS) is associated with high morbidity and mortality. While elective transcatheter aortic valve replacement (TAVR) is supported by robust randomized evidence, data on TAVR performed in non-elective settings (urgent, emergent, or salvage) remain limited and heterogeneous. In these settings, outcomes may be influenced more by baseline clinical severity than by procedural factors. This study aimed to systematically evaluate the safety and effectiveness of non-elective TAVR and to compare the associated outcomes with those of elective TAVR.Methods:This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines and registered with PROSPERO. The PubMed, Embase, and Scopus databases were searched for studies published after 2002 evaluating urgent, emergent, or salvage TAVR in adult patients with severe AS. Eligible studies were required to include elective TAVR as a comparator. Outcomes included mortality and major procedural complications. Prespecified era-based analyses and sensitivity analyses excluding studies at critical risk of bias were performed. Risk of bias was assessed using the ROBINS-I tool.Results:A total of 17 observational studies published between 2015 and 2025 were included, comprising 215,141 patients. Compared with elective patients, those undergoing non-elective TAVR had more advanced heart failure, higher NYHA class, greater comorbidity burden, and higher surgical risk scores, whereas baseline echocardiographic severity of AS was similar. Pooled outcomes for non-elective TAVR showed in-hospital, 30-day, and 1-year mortality rates of 5.1%, 12.4%, and 26.7%, respectively. Non-elective TAVR was associated with higher mortality at all time points compared with elective procedures; meanwhile, rates of stroke, vascular complications, and permanent pacemaker implantation were similar, while major bleeding and acute kidney injury were more frequent. Era-based analyses showed a stable relative mortality risk over time.Conclusions:Although non-elective TAVR is associated with worse outcomes than elective procedures, these differences appear to be largely driven by baseline clinical severity rather than by the transcatheter intervention. In unstable patients with severe AS, non-elective TAVR remains the most effective definitive treatment option compared with balloon aortic valvuloplasty or conservative management.The PROSPERO Registration:CRD420251239620, https://www.crd.york.ac.uk/PROSPERO/view/CRD420251239620.

Coronary artery disease (CAD) remains a leading cause of global morbidity and mortality. Accurate diagnosis of ischemia-causing coronary stenoses is essential for guiding revascularization and improving outcomes. Although invasive fractional flow reserve (FFR) remains the gold standard for functional lesion assessment, its use is limited by procedural invasiveness, cost, and complexity. CT-derived FFR (FFRct), based on computational fluid dynamics (CFD), was the first major advance in noninvasive physiological assessment, but its adoption has been hindered by intensive off-site computation and dependence on high-quality imaging. This review summarizes the evolution from invasive FFR to AI-driven functional assessment of coronary lesions. We examine the principles and validation of CFD-based FFRct and then focus on the shift toward artificial intelligence, including both machine learning (ML) and deep learning (DL) approaches. These methods range from models using engineered geometric and plaque features trained on large synthetic datasets to end-to-end systems that learn directly from imaging data. We discuss key validation studies evaluating diagnostic accuracy, prognostic value, and clinical utility, with attention to performance in challenging settings such as intermediate stenoses, heavy calcification, and patients with comorbidities. We also highlight major barriers to widespread adoption, including dependence on input data quality, limited explainability, regulatory hurdles, and integration into clinical workflows. Finally, we outline future directions, including AI-enabled virtual PCI planning, multimodal risk stratification, and broader access to functional cardiac assessment. AI has the potential to transform noninvasive coronary imaging by enabling a single CCTA scan to provide rapid, integrated evaluation of anatomy, plaque characteristics, and physiological significance, supporting more personalized care and better clinical outcomes.

Background In low-gradient severe aortic stenosis (AS), reduced left ventricular ejection fraction (LVEF <50%) is practically used to define low flow and prompt dobutamine stress echocardiography to discern pseudo-severe AS. In patients with preserved LVEF, stroke volume index (SVI) <35 mL/m² is typically used. However, both are volume-based surrogates. Transaortic flow rate (TAFR), calculated as stroke volume divided by left ventricular ejection time, may better reflect true flow. Nonetheless, comparative data between TAFR and established metrics remain limited. We aimed to evaluate the prognostic value of TAFR in symptomatic low-gradient severe AS. Methods We retrospectively identified patients with low-gradient severe AS (AVA ≤1 cm 2 and peak velocity (Vmax) <4 m/s or mean gradient (MG) <40 mm Hg) who underwent transcatheter aortic valve replacement at Mayo Clinic sites (2017–2023). The primary outcome was 1-year all-cause mortality. Survival was assessed using Kaplan-Meier and Cox proportional hazards models. Results Among 475 patients included (mean age 85±8 years; 49% women), 242 (51%) had TAFR <220 mL/s, 165 (35%) had EF <50%, and 221 (47%) had SVI <35 mL/m 2 . Low TAFR was significantly associated with higher 1-year mortality even after stratifying by EF or SVI. In multivariate analysis, TAFR was an independent predictor of mortality (HR 2.38; 95% CI 1.19 to 4.76, p=0.014) after adjusting for reduced LVEF, low SVI, gender, chronic kidney disease and mitral and tricuspid regurgitation. Conclusions In patients with symptomatic low-gradient severe AS, low TAFR, not SVI or LVEF, is independently associated with mortality and may offer more accurate measure of flow state for clinical staging.

Acute exposure to hypoxia will induce right ventricular (RV) hemodynamic changes and may increase the degree of right-to-left shunting, which can contribute to dyspnea at altitude. In this retrospective study, 125 patients (median age 66 years; 50.4% women) with unexplained dyspnea at altitude underwent hypoxic simulation testing (HST) with transthoracic echocardiography (TTE). During simulated hypoxia (mode (Min-Max) altitude: 8000 (6000–18,000) ft, were observed a significant decrease in oxygen saturation (97% (95–98) vs. 88% (82–92), p < 0.001) and RV free wall longitudinal strain (−19.6 ± 3.99% vs. −17.3 ± 4.17%, p < 0.01), an increase in RV systolic pressure (RVSP: 26 (23–30.5) vs. 29 (25–36.5) mmHg, p < 0.001). No significant changes were observed in TAPSE (20 (18–23) vs. 20 (19–24) mm) or S wave (0.12 (0.11–0.14) vs. 0.13 (0.12–0.14) m/s). Right-to-left shunting was present in 47.2% of patients and 11.9% exhibited inducible shunting only under hypoxia. However, under hypoxia, there were no significant differences in RV hemodynamic parameters or saturation between those with and without shunting. TTE with HST is useful to characterize both cardiopulmonary response and the dynamic changes in right-to-left shunt behavior under hypoxic stress.