Oh, Jae K.

Background Bicuspid aortic valve (BAV) is the most common congenital heart defect, associated with a lifetime morbidity burden exceeding 80% and an estimated 7% probability of transmission to first-degree relatives. Hence, echocardiographic screening of relatives is recommended. The aim of our study is to develop an artificial intelligence-enabled electrocardiogram (AI-ECG) model for the detection of BAV. Methods Between January 1, 1990, and June 30, 2023, the Mayo Clinic healthcare system database was searched for patients aged ≥18 years with echocardiograms performed for family history of BAV, abnormal auscultation, a BAV-related diagnosis, or to rule out the condition, and with an electrocardiogram (ECG) obtained within 6 months. Cases were patients with a confirmed BAV diagnosis, and controls were those with normal aortic valves. ECGs were randomly assigned into training (80%), internal validation (10%), and test (10%) groups. The AI-ECG model was built using Keras under TensorFlow. Results A total of 13,065 valid ECG–echocardiogram pairs were included (mean [SD] age, 58 [17] years; 44% women), with a median time of 1 day (interquartile range, 0–3 days) between studies. With a 45% BAV prevalence in the test group ( n = 1301), the area under the curve was 0.704, positive predictive value (PPV) 60.1%, and negative predictive value (NPV) 69.4%. At a prevalence of 7%, the NPV increased to 96.1%, but the PPV decreased to 12.2%. Conclusions In this retrospective cohort and using the current AI-ECG model, BAV detection showed modest diagnostic performance and limited standalone clinical utility, suggesting that ECG-only screening for BAV may require additional modalities or more refined phenotyping.

Background Diastolic dysfunction is common in patients with aortic stenosis and may influence outcomes following surgical aortic valve replacement. We aimed to examine the association of preoperative artificial intelligence (AI)‐generated diastolic function grades with early and late outcomes following aortic valve replacement and how postoperative progression influence prognosis. Methods We identified 5503 patients undergoing aortic valve replacement between 2000 and 2023. Diastolic function was assessed using a validated deep‐learning AI model applied to 12‐lead ECGs done preoperatively and on postoperative follow‐up. Diastolic grades were classified by AI into Grades 1 to 3. Longitudinal trend analyses and multivariable regression models were used to assess study end points. Results Among 5503 patients (mean age 72.4±10.8 years; 39% female), higher AI ECG diastolic grades were associated with greater comorbidity burden, including diabetes, renal disease, and heart failure. AI ECG diastolic Grade 3 was independently associated with higher in‐hospital mortality (odds ratio, 2.5; P =0.007) and other complications. At 5‐year follow‐up, patients with Grade 3 showed the least improvement in diastolic function by both ECG and echocardiography. Grades 2 and 3 diastolic function at baseline were independently associated with increased late mortality (hazard ratio, 1.3 and 2.45, respectively; both P <0.001). Additionally, lack of improvement in AI ECG diastolic grade by 1 year was also independently associated with late mortality. Conclusions AI ECG–derived diastolic function grades strongly correlates with early complications and long‐term mortality and diastolic progression after aortic valve replacement. AI ECG provides a powerful, noninvasive tool for risk stratification and longitudinal monitoring of patients with aortic stenosis.

Importance Timely identification of aortic stenosis (AS) is essential for appropriate clinical management, yet screening remains limited by dependence on comprehensive echocardiography and trained imaging personnel. Objective To develop and validate a deep learning algorithm for detection of moderate or greater AS and prospectively evaluate its performance using artificial intelligence (AI)–guided focused cardiac ultrasound (FoCUS) acquired by novice operators. Design, Setting, and Participants This diagnostic study included retrospective algorithm development and validation and prospective evaluation of AI-guided FoCUS across Mayo Clinic sites in the Midwest, Arizona, and Florida. The model was developed using 6753 patients and evaluated in internal validation (n = 852), internal test (n = 844), and validation (n = 1912) cohorts. Performance was assessed on FoCUS acquired by experienced sonographers (n = 602) and prospectively by novice operators (n = 1302). The retrospective model development and validation cohorts comprised studies performed from January 2005 through September 2022. Prospective study was conducted in 2 enrollment periods from June to August 2024 and from June to September 2025. Participants from both periods were combined to comprise the final prospective cohort. Exposure AI-guided FoCUS acquisition and automated deep learning–based assessment for detection of moderate or greater AS. Main Outcomes and Measures Detection of moderate or greater AS. Performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, and positive predictive value. Results The model demonstrated excellent discrimination in the internal test cohort (AUROC, 0.99; 95% CI, 0.98-1.00) and geographically distinct validation cohorts in Arizona (AUROC, 0.99; 95% CI, 0.97-1.00) and Florida (AUROC, 0.99; 95% CI, 0.96-1.00). Among FoCUS examinations acquired by experienced sonographers, sensitivity was 95% (95% CI, 82-99) and specificity was 97% (95% CI, 95-98). In the prospective novice-operator cohort, 1258 of 1302 examinations (96.6%) were suitable for automated analysis. Sensitivity was 93% (95% CI, 82-99) and specificity was 96% (95% CI, 95-97). Expert review of AI-positive and uninterpretable examinations increased the positive predictive value from 49.4% to 91.1%, with sensitivity of 85.4%. Conclusions and Relevance A deep learning algorithm accurately detected moderate or greater AS across validation cohorts in this study. In prospective evaluation, novice operators were able to acquire AI-guided FoCUS examinations that enabled accurate detection of moderate or greater AS. These findings support a scalable strategy that may expand access to AS detection in settings with limited echocardiography resources.

Abstract A recent discussion that highlighted the differences between the reported effective orifice area and corresponding incidence of prosthesis-patient mismatch within the 1- and 7-year results of the PERIcardial SurGical AOrtic Valve ReplacemeNt (PERIGON) Trial of the Avalus valve has once again emphasized the instability of such measurements. The effective orifice area, calculated using the continuity equation through measurements of the left ventricular outflow tract diameter, is highly susceptible to measurement variability, as previously demonstrated in literature through both intra- and interobserver variability. To illustrate this problem, a simulation was performed in which the left ventricular outflow tract diameter was systematically altered by −2 mm, −1 mm, 0 mm, +1 mm, and +2 mm. Subsequently, the proportion of severe prosthesis-patient mismatch was recalculated and varied between 13.3%, 6.2%, 3.1% (original), 1.2%, and 0.3%, respectively. These findings demonstrate that small but clinically realistic variability in left ventricular outflow tract measurements can significantly affect the presence of prosthesis-patient mismatch. A measurement variability of 1 mm can nearly double or halve the prevalence of severe prosthesis-patient mismatch. This illustrates that prosthesis-patient mismatch, in its current form, is highly unstable and that indexed effective orifice area-based prosthesis-patient mismatch should not be used in isolation.

Abstract Echocardiography is the most widely used cardiac imaging modality, yet artificial intelligence-enabled interpretation remains limited by the inability of existing models to integrate visual assessment, quantitative measurement, and clinical reasoning within a unified framework. Here we present EchoAtlas, the first autoregressive vision-language model developed for echocardiographic interpretation. Trained on over 12.9 million question-answer pairs derived from approximately 2 million echocardiogram videos, EchoAtlas achieves 0.966 accuracy on multiple-choice questions in our internal test set and establishes a new state-of-the-art on the public MIMIC-EchoQA benchmark (0.699 vs. 0.508 previously). EchoAtlas also provides accurate quantitative measurements, segment-level regional wall motion assessment, longitudinal comparison, and diagnostic reasoning across diverse question formats — capabilities not previously demonstrated in this domain. These results highlight the potential of autoregressive vision-language models as a foundation for interactive echocardiographic interpretation, representing an early step toward scalable, auditable artificial intelligence systems in cardiology practice.

Background Artificial intelligence (AI)–enabled electrocardiography has emerged as a tool for detecting cardiac dysfunction. The prognostic relevance of AI‐enabled electrocardiography–derived diastolic dysfunction (DD) in patients undergoing transcatheter aortic valve replacement had not been assessed. Methods We analyzed 3197 patients undergoing transcatheter aortic valve replacement for severe aortic stenosis between 2010 and 2023 with baseline 12‐lead ECGs processed by a validated AI model to classify diastolic function (grades 0–3). Multivariable Cox models and nested model comparisons assessed associations with all‐cause mortality, including a prespecified analysis in patients with indeterminate echocardiographic grading. Trajectories were defined by change in AI‐enabled electrocardiography DD grade (<2 versus ≥2) from baseline to 30‐day or 1‐year follow‐up. Results At baseline, 21% had grade 3, 57% had grade 2, 8% had grade 1, and 14% had grade 0 AI‐enabled electrocardiography DD. Higher grades were associated with adverse cardiac remodeling and comorbidities. Over a median follow‐up of 3.4 years, grade 3 AI‐enabled electrocardiography DD independently predicted mortality (hazard ratio [HR], 1.80 [95% CI, 1.47–2.20]; P <0.001). AI‐enabled electrocardiography DD improved prognostic discrimination beyond clinical and echocardiographic measures (ΔHarrell concordance statistic, 0.016; Δ χ 2 =57; P <0.001). Among 1259 patients with indeterminate echocardiographic grading, AI‐enabled electrocardiography added prognostic value (ΔHarrell concordance statistic, 0.02; Δ χ 2 =13; P =0.006). Worsening or persistently high‐risk trajectories were associated with increased mortality (HRs, 1.45–1.80; all P <0.05). Conclusions AI‐enabled electrocardiography–derived DD independently predicts mortality after transcatheter aortic valve replacement, adds value beyond echocardiographic grading, and enables dynamic risk stratification through longitudinal tracking.

Introduction: Diastolic dysfunction is a hallmark of heart failure with preserved ejection fraction, yet echocardiographic grading is challenging in patients with hypertrophic or amyloid cardiomyopathies due to variable structural and functional remodeling. Consequently, there is a need for additional diagnostic tools to identify patients at increased risk of diastolic dysfunction and enhanced filling pressure. Research Question: What is the accuracy of an artificial intelligence (AI)-enabled electrocardiogram (ECG) model to predict diastolic dysfunction in hypertrophic and amyloid cardiomyopathy? Methods: We included patients with diverse phenotypes of hypertrophic cardiomyopathy (HCM) and cardiac amyloidosis and validated Mayo Clinic's AI-enabled ECG model to grade diastolic function (normal: grade 0/1; abnormal: grade 2/3). Echocardiographic analysis of systolic (LVEF) and diastolic function was performed within 14 days of ECG recording. Diastolic function was graded based on E/e’ (>14), left atrial volume index (>34 mL/m2), pulmonary venous atrial reversal flow duration (≥30 ms), and peak tricuspid regurgitation velocity (>2.8 m/s). Filling pressure was considered increased if ≥3 parameters were abnormal, inconclusive if 2 were abnomal, and normal if ≥3 were normal. Echocardiograms with severe mitral regurgitation were excluded. Results: We identified and matched 840 ECGs from non-obstructive HCM patients (n=125, age 58±19y, 70% male), 145 ECGs from obstructive HCM patients (n=29, age 56±18y, 66% male), 225 ECGs from apical HCM patients (n=33, age 66±14y, 61% male), and 382 ECGs from cardiac amyloidosis patients (n=49, age 70±13y, 57% male). The AI-enabled ECG model predicted diastolic dysfunction with an accuracy of 65% (non-obstructive HCM), 67% (obstructive HCM), 65% (apical HCM), and 75% (cardiac amyloidosis), with sensitivities of 94% and 97% for obstructive HCM and cardiac amyloidosis, respectively, and high positive and negative predictive values for cardiac amyloidosis. Systolic dysfunction (LVEF≤40%) was more common in cardiac amyloidosis (29%) than in all HCM patients (5%). Conclusion: The AI-enabled ECG model demonstrates moderate accuracy but high sensitivity for detecting diastolic dysfunction and increased filling pressure, especially in obstructive HCM and cardiac amyloidosis. The AI-enabled ECG is a useful, widely scalable, and low cost tool to identify patients at increased risk for diastolic dysfunction.

Abstract Aims To assess the performance of an artificial intelligence-enabled electrocardiogram (AI-ECG) algorithm in identifying patients with moderate to severe aortic stenosis (AS) in an Asian cohort from a tertiary care centre. Methods and results We identified a randomly selected patients ≥60 years old who underwent echocardiography and ECG within in 31 days between 2012 and 2021 at the Samsung Medical Center in Korea. Patients with previous cardiac surgery, prosthetic valves, or pacemakers were excluded. The AI-ECG model, originally developed and validated by Mayo Clinic in the USA, was applied without fine-tuning. Performance metrics, including the area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy, were calculated to compare AI-ECG predictions with TTE-confirmed AS status. Among 5425 patients, 1095 had moderate to severe AS, and 4330 age- and sex-matched patients without AS were included as controls. The AI-ECG model achieved an AUC of 0.85 (95% CI: 0.84–0.87) in detecting moderate to severe AS. Sensitivity, specificity, PPV, NPV, and accuracy were 0.83, 0.65, 0.37, 0.94, and 68.29%, respectively. The model's performance was consistent across various age and sex subgroups, with sensitivity increasing in older patients. Conclusion The AI-ECG model developed in the USA demonstrated comparable performance in detecting moderate to severe AS in an Asian cohort compared with its original validation population. These findings highlight the potential utility of AI-ECG as a non-invasive screening tool for AS across diverse patient populations.