Lopez-Jimenez, Francisco

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.

Abstract Aims Heart failure is prevalent; however, there is no cost-effective screening option. Against formal echocardiography, we assessed the diagnostic performance of focused cardiac ultrasound (FoCUS), performed by novice users guided and interpreted by artificial intelligence (AI) for the assessment of left ventricular ejection fraction (LVEF). Methods and results We prospectively enrolled 496 adults referred for diagnostic echocardiography. Novice operators (without clinical or imaging experience) underwent a 4-h imaging workshop, then performed FoCUS using a point-of-care device (Philips Lumify) utilizing real-time AI-image guidance (UltraSight). Images were assessed by AI-image interpretation software (Mayo Clinic), and two expert blinded echocardiologists. Median age was 67, 39% female, and body mass index range 17–56 kg/m2. Forty-one subjects (8.3%) exhibited moderate or greater LV dilatation and 28 (5.6%) had LVEF &lt;40%. The median scan time was 4 min (IQR 3–5). Adequate views were achieved in 95.0% and 97.4% of subjects for AI interpretation and expert analysis, respectively. A two-step diagnostic screening process for low EF in which only abnormal AI reads or uninterpretable scans were reviewed by experts (15.1%) had a sensitivity 96.2%, specificity 95.4%, PPV 54.3%, and NPV 99.8%. A subsequent prospective validation study of 344 subjects demonstrated similar results (sensitivity 100%, specificity 99.4%). Conclusion In this early feasibility investigation, AI-guidance technology embedded on a handheld device, enabled novice users without clinical or imaging experience to acquire sufficient quality images to accurately assess LVEF with minimal training. The technology evaluated in this study may hold promise for AI-guidance and interpretation to facilitate low-cost screening for cardiac dysfunction.

Abstract Aims Early detection of structural heart disease (SHD) improves patient outcomes. However, population-based screening is not recommended due to the lack of accurate and cost-effective tools. We evaluated the costs of artificial intelligence–enabled electrocardiogram (AI-ECG) alone vs. AI-ECG followed by handheld cardiac ultrasound (HCU) for SHD screening. Methods and results We performed a model-based cost analysis using data from 286 adult patients who underwent ECG and same-day HCU performed by a novice operator. Transthoracic echocardiogram (TTE) was the reference standard. We compared two screening strategies: (i) AI-ECG alone and (ii) a stepwise approach (AI-ECG followed by HCU). We assessed costs per diagnosis of aortic stenosis (AS), increased left ventricular wall thickness (ILVWT), and left ventricular systolic dysfunction (LVSD). Sensitivity analyses were conducted for varying disease prevalence. The stepwise approach decreased the cost per diagnosis of AS from $6386 (AI-ECG alone) to $2746 (57.0% savings), ILVWT from $4448 to $2895 (34.9% savings), and LVSD from $1469 to $1296 (11.8% savings). Overall, the cost per diagnosis for all SHDs combined decreased from $1940 to $1570 (19.1% savings). Sensitivity analysis demonstrated that cost savings were inversely proportional to disease prevalence. Nevertheless, stepwise screening remained cost-saving compared with AI-ECG alone until prevalence exceeded ∼55.9% for AS, 28.9% for ILVWT, 20.7% for LVSD, and 40.8% for all SHDs combined. Conclusion A stepwise screening strategy incorporating HCU after a positive AI-ECG reduces the immediate costs of SHD detection by minimizing unnecessary TTEs. This approach may enhance the feasibility of population-based SHD screening, particularly in lower-prevalence settings.

ABSTRACT Background Acute heart failure (AHF) exhibits marked heterogeneity in diastolic hemodynamics, yet comprehensive echocardiographic assessment of diastolic function (DF) and filling pressure (FP) is often infeasible. We evaluated whether artificial intelligence–enabled electrocardiography (AI-ECG) could provide scalable DF grading and FP estimation in hospitalized AHF patients. Methods We retrospectively studied adults hospitalized for AHF across Mayo Clinic sites (2013–2023) who received ≥1 dose of intravenous loop diuretic and had paired 12-lead ECG and TTE. The previously validated AI-ECG DF model was applied without retraining to generate four DF grades and a continuous FP probability. Clinical outcomes were all-cause mortality and heart failure rehospitalization. Associations with clinical severity markers and echocardiographic indices were examined. Kaplan–Meier survival analysis and adjusted multivariable Cox proportional hazards models were performed. Exploratory analyses examine the kinetics of change in FP probability and impact on mortality. Results Among 11,513 patients (median age 75 years, 39% female), AI-ECG DF grading was feasible in 100%, whereas echocardiographic DF was indeterminate in 44% of clinically eligible patients. In 2,582 patients with determinate echocardiographic DF, AI-ECG FP probability discriminated TTE Grade 2–3 dysfunction with AUC 0.85 (95% CI 0.83 – 0.86). Higher AI-ECG DF grades were associated with higher comorbidity burden, worse NYHA class, elevated NT-proBNP, higher MAGGIC scores, elevated PCWP, and more advanced structural remodeling. After multivariable adjustment, AI-ECG DF remained independently associated with mortality (hazard ratio [HR] 1.25, 95% CI 1.16–1.35 for Grade 2; HR 1.44, 95% CI 1.33–1.56 for Grade 3 versus Normal/Grade 1). Combining AI-ECG DF with MAGGIC scores yielded ordered risk gradients, with highest mortality in patients with both high MAGGIC and Grade 2–3 DF. Among patients with serial ECGs, improvement in FP probability was independently associated with lower mortality (HR 0.85, 95% CI 0.79–0.91), whereas worsening did not show a consistent adverse gradient beyond baseline DF. Conclusions In a large, geographically diverse AHF cohort, AI-ECG DF grading was universally feasible, correlated with established hemodynamic severity markers, and provided independent prognostic information beyond established risk factors, supporting its role as a pragmatic, scalable diastolic biomarker in AHF. CLINICAL PERSPECTIVE What Is New? In 11,513 hospitalized acute heart failure (HF) patients, artificial intelligence-enabled electrocardiography provided diastolic function grading in 100% of patients from a single 12-lead ECG without requiring additional clinical variables, compared with 56% feasibility for guideline-based echocardiography grading. AI-ECG diastolic function grades correlated with established marker of severity (NYHA functional class, NT-proBNP, MAGGIC risk scores, and pulmonary capillary wedge pressure) and remained independently associated with both mortality and HF rehospitalization after multivariable adjustment. Serial AI-ECG measurements identified post-discharge filling pressure trajectories, with improvement independently associated with 15% lower mortality, a first demonstration that longitudinal ECG assessment can track post-discharge hemodynamic recovery. What Are the Clinical Implications? AI-ECG transforms the universally obtained 12-lead ECG into an actionable hemodynamic biomarker that addresses the critical gap when echocardiographic diastolic function assessment is indeterminate or unavailable in acute HF patients. Despite markedly different hemodynamic severity and long-term outcomes across AI-ECG diastolic function grades, hospitalization length of stay did not differ, suggesting advanced diastolic dysfunction represents occult risk not easily recognized during routine acute care and highlighting the need for improved post-discharge risk stratification. The continuous filling pressure probability metric enables longitudinal monitoring of post-discharge hemodynamic status using serial routine ECGs, potentially identifying patients requiring intensified follow-up or specialist referral.

Abstract Aims Artificial intelligence electrocardiography (AI-ECG) algorithms are emerging tools for identifying individuals at risk of atrial fibrillation (AF). We evaluated the predictive performance of a validated AI-ECG algorithm for incident AF in UK Biobank participants with AF risk factors, irrespective of prevalent cardiovascular disease, and its incremental value when added to clinical predictors. Methods and results The AI-ECG tool was applied to sinus rhythm ECGs from UK-Biobank participants with risk factors for AF but no AF. Model performance was evaluated using time-dependent ROC-AUC and Harrell’s C-index. Multivariable Cox regression was used to identify clinical risk factors associated with incident AF and to quantify the contribution of AI-ECG. A total of 21 842 participants (56% male) were included. The median follow-up time was 3.7 years (IQR 0.5–5.4) The ECG-AI tool achieved a ROC-AUC of 0.73 (95% CI 0.68–0.78) at 1 and 0.69 (95% CI 0.66–0.72) at 3 years. A multivariable Cox regression model using clinical parameters achieved a ROC-AUC of 0.71 (95% CI 0.66–0.75) at 1 and 0.71 (95% CI 0.68–0.74) at 3 years. By adding ECG-AI to the clinical Cox regression model, the ROC-AUC increased to 0.75 (95%CI 0.71–0.80) at 1 and 0.74 (95% CI 0.71–0.77) at 3 years. AI-ECG showed a hazard ratio of 1.23 per decile increase (95% CI 1.18–1.27). Conclusion An AI-ECG algorithm improved the prediction of incident AF when added to a clinical parameter-based model over a median follow-up time of 3.7 years among individuals with comorbidities predisposing to AF who may benefit from targeted screening and preventive strategies.

Nonspecific symptoms and other diagnostic challenges lead to underdiagnosis of cardiac amyloidosis (CA). Artificial intelligence (AI) could help address these challenges, but a summary of the performance of these tools is lacking. This narrative review of published literature describes the performance of AI tools that use data from ECGs and echocardiography to improve identification of CA and challenges that hinder adoption of these tools. Thirteen studies met inclusion criteria with sample sizes ranging from 50 to 2451 patients. Four studies used ECG data, 8 used echocardiography data, and 1 used both. The CA gold standard was typically defined as a CA diagnosis in an institutional or other database but the requirements for these diagnoses were heterogenous across studies, and many did not distinguish among CA subtypes. AI model development varied considerably, and only 4 studies included external validation. The ability of models to predict CA ranged from 0.71 to 1.00, sensitivity ranged from 16% to 100%, and specificity from 75% to 100%. Only 1 study reported model performance across strata of sex, age, race, and CA type. Persistent challenges to AI adoption include usability, cost, value added, electronic health record/information technology interoperability, patient‐related factors, regulation, and privacy and liability. Published studies on AI for improved identification of CA show favorable performance measures but numerous methodologic and other challenges must be addressed before these tools are more widely adopted.

Abstract Background Automated data extraction from echocardiography reports could facilitate large-scale registry creation and clinical surveillance of valvular heart diseases (VHD). We evaluated the performance of open-source large language models (LLMs) guided by prompt instructions and chain of thought (CoT) for this task. Methods From consecutive transthoracic echocardiographies performed in our center, we utilized 200 random reports from 2019 for prompt optimization and 1000 from 2023 for evaluation. Five instruction-tuned LLMs (Qwen2.0-72B, Llama3.0-70B, Mixtral8-46.7B, Llama3.0-8B, and Phi3.0-3.8B) were guided by prompt instructions with and without CoT to classify prosthetic valve presence and VHD severity. Performance was evaluated using classification metrics against expert-labeled ground truth. Mean squared error (MSE) was also calculated for predicted severity’s deviation from actual severity. Results With CoT prompting, Llama3.0-70B and Qwen2.0 achieved the highest performance (accuracy: 99.1% and 98.9% for VHD severity; 100% and 99.9% for prosthetic valve; MSE: 0.02 and 0.05, respectively). Smaller models showed lower accuracy for VHD severity (54.1%-85.9%) but maintained high accuracy for prosthetic valve detection (&gt;96%). Chain of thought reasoning yielded higher accuracy for larger models while increasing processing time from 2-25 to 67-154 seconds per report. Based on CoT reasonings, the wrong predictions were mainly due to model outputs being influenced by irrelevant information in the text or failure to follow the prompt instructions. Conclusions Our study demonstrates the near-perfect performance of open-source LLMs for automated echocardiography report interpretation with the purpose of registry formation and disease surveillance. While larger models achieved exceptional accuracy through prompt optimization, practical implementation requires balancing performance with computational efficiency.

Background Left ventricular diastolic dysfunction (LVDD) predicts mortality in patients in cardiac intensive care units. An artificial intelligence enhanced ECG (AIECG) algorithm can predict LVDD and mortality in general populations but has not been examined in cardiac intensive care units. Methods This historical cohort study included consecutive adults admitted to Mayo Clinic cardiac intensive care unit from 2007 to 2018 with an admission AIECG. The AIECG assigned the LVDD grade (0–3). Medial mitral E/e' ratio >15 on transthoracic echocardiogram (TTE) defined elevated filling pressures. In‐hospital and 1‐year mortality was evaluated, before and after multivariable adjustment. Results We included 11 868 patients (median age 69.5 years, 37.7% female); 48% had heart failure and 44% had acute coronary syndromes. AIECG LVDD grade was 0 (normal), 33%; 1, 7%; 2, 39%; and 3, 21%. In‐hospital and 1‐year mortality increased in each higher AIECG LVDD grade. After adjustment, each higher AIECG LVDD grade was associated with higher in‐hospital (adjusted odds ratio [OR], 1.22 [95% CI, 1.13–1.32]) and 1‐year mortality (adjusted hazard ratio [HR], 1.23 [95% CI, 1.19–1.29]); this persisted after adjustment for TTE measurements. Patients with grade 2 or 3 LVDD by AIECG and medial mitral E/e' ratio >15 by TTE had the highest in‐hospital (adjusted OR, 2.54 [95% CI, 1.69–3.88]) and 1‐year (adjusted HR, 2.03 [95% CI, 1.65–2.48]) mortality, whereas patients meeting either of these criteria had similar, elevated mortality. Conclusions The AIECG LVDD grade was strongly associated with in‐hospital and 1‐year mortality in patients in cardiac intensive care units, even after adjusting for clinical variables and TTE measurements. Patients with concordant AIECG and TTE for elevated filling pressures were at highest risk.