López-Jiménez, Francisco

Introduction: Artificial intelligence applied to electrocardiograms (ECG-AI) enables early and efficient detection of cardiovascular disease. The technical and clinical impact of different approaches to designing ECG-AI remains under-explored. We consider left ventricular hypertrophy (LVH) and examine the impact of disease-specific labeling of minority etiologies of left ventricular wall thickening- hypertrophic cardiomyopathy (HCM) and cardiac amyloidosis (CA) - on ECG-AI performance. Hypothesis: ECG-AI designed to detect HCM or CA outperforms ECG-AI broadly trained to identify LVH due to differences in attention to key ECG features associated with HCM or CA. Methods: Independent neural networks were trained to identify HCM, CA, and LVH. HCM and CA patients were identified using the full depth of EHR data and validated via clinician adjudication (N = 3,025 and 4,251 patients, respectively). Moderate and severe LVH patients (N = 98,144) were identified using echocardiogram results and international LVH criteria. We compared the ability of disease-specific ECG-AI vs. LVH ECG-AI to detect HCM (or CA) by examining model performance within each of the three test sets. Results: 52.6% of HCM patients and 49.5% of CA patients met LVH criteria. Within the HCM test set, the AUROC, sensitivity, specificity, and positive predictive value (PPV) of the HCM ECG-AI were 0.94, 0.88, 0.85, and 60%, respectively, whereas the same for LVH ECG-AI were 0.83, 0.83, 0.64, and 37%. Within the CA test set, CA ECG-AI achieved AUROC, sensitivity, specificity, and PPV of 0.94, 0.81, 0.91, and 69%, respectively, whereas the same for LVH ECG-AI were 0.75, 0.73, 0.68, and 31%. ECGs positive for HCM exhibited higher P wave area, QRS deflection and R wave amplitude compared to LVH-positive ECGs (Cohen's D > 0.75). ECGs positive for CA exhibited lower R wave amplitude/area, QRS deflection and QRS areas (Cohen's D > 0.6). Conclusion: Models trained using carefully labeled HCM or CA identify distinct signals associated with the diseases, resulting in improved disease discrimination compared to a model trained to identify echocardiographic left ventricular thickening. Labeling choices should be communicated transparently to clinicians to ensure safe and effective use of ECG-AI.

Introduction: Cardiotoxicity is a serious side effect of anthracycline treatment, most commonly manifesting as a reduction in left ventricular ejection fraction (LVEF) on cardiac imaging. Cost-effective alternatives to cardiac imaging are currently not available. Artificial intelligence (AI) models have been developed to detect a reduced LVEF from 12-lead electrocardiograms (ECGs) in the general population. Hypothesis: AI-ECG model can detect a reduced LVEF after anthracycline therapy (i.e. cardiotoxicity). Methods: We included an unbiased oncological population of 989 female breast cancer patients receiving (neo)adjuvant anthracycline chemotherapy. Average follow-up time was 9.83±4.2 years. Patients with incomplete transthoracic echocardiography (TTE) data, HER-2-directed therapy, significant comorbidities, upfront metastatic disease, second primary malignancy or pre-existing cardiovascular disease were excluded from the analyses. Patients with LVEF decline for reasons other than anthracycline-induced cardiotoxicity were also excluded. Primary readout was the diagnostic performance of AI-ECG by area under the curve (AUC) for various levels of LVEF <50%. Results: Using 1,877 ECGs from 703 unique patients, the AI-ECG model detected an LVEF <50% and ≤35% after anthracycline therapy with an AUC of 0.93 and 0.94, respectively. Conclusions: These data support the use of AI-ECG for cardiomyopathy (cardiotoxicity) screening after anthracycline-based chemotherapy. This technology could serve as a gatekeeper to more costly testing and could enable patients to monitor themselves over long periods of time.

Introduction: Biologic sex and hormonal concentrations shape ECG parameters due to sex hormone effects on cardiac function. AI-enabled ECGs have previously been shown to accurately predict patient sex in adults and correlate with sex hormone levels. Aims: We aimed to test the ability of AI-enabled ECGs to predict pediatric patient sex and explore the influence of pubertal development. Methods: Two AI-enabled ECG models were created using a convolutional neural network trained on pediatric 10-second, 12-lead surface ECGs. The first model was trained de novo using pediatric data. The second model used transfer learning from a previously validated adult data-derived algorithm. We analyzed the first ECG from 89,063 pediatric patients (aged <19 years) recorded from 1988-2022, and divided the cohort into training, validation, and testing datasets. Subgroup analysis was performed on prepubertal (0-7 years), pubertal (8-14 years), and post-pubertal (15-18 years) patients. Results: The cohort consisted of 46.6% males, with 21,434 prepubertal, 26,300 pubertal, and 41,329 post-pubertal children. No significant performance difference was observed between the two models. Both models demonstrated 81% accuracy and an overall AUC of 0.91 in the testing cohorts, with high discriminatory ability in post-pubertal teenagers (AUC = 0.98). (Figure 1) Conclusions: AI-ECG predicts pediatric patient sex with varying performance across pubertal stages. The highest discriminatory ability is observed post-puberty, decreasing in pubertal and prepubertal children. The results suggest hormonal and physiological changes during puberty may influence cardiac electrophysiology measurably. More research is required to interpret CNN-predicted sex discordance with actual sex and its implications for future cardiovascular risk.

Background: Diagnosis of heart failure with preserved ejection fraction (HFpEF) is challenging due to non-diagnostic imaging outputs and discordant clinical features. Using only a single videoclip from a standard transthoracic echocardiogram (TTE), we developed an artificial intelligence (AI) model to differentiate patients with HFpEF from those without HFpEF. In a separate test population, we assessed the model’s ability to predict cardiac mortality. Methods: A 3-dimensional convolutional neural network was developed and trained on apical 4-chamber videoclips to classify patients with HFpEF (cases) versus without HFpEF (controls). Overall and cardiac mortality were obtained from the National Death Index. Results: The model was independently tested in 646 cases and 638 controls from 8 centers in 4 states. Sensitivity (87.8%; 84.5-90.9%) and specificity (81.9%; 78.2-85.6%) were high and sustained in various subgroups, including increased left atrial volume index, average E/e’ ≥15, and pulmonary artery systolic pressure >35 mm Hg. During follow-up (median [IQR] 3.4 [1.7-6.5] years), 540 patients died. There were 135 cardiac deaths attributable to heart failure (47%), coronary artery disease (41%), valve disease (4%), arrhythmia (4%), and others (5%). Compared to a negative output, all-cause mortality was higher in patients with a positive (HR 2.9 [2.4-3.5]) or uncertain output (HR 1.8 [1.2-2.5]), and cardiac mortality (Figure) was also higher in patients with a positive (HR 7.7 [4.5-13.3]) or uncertain output (HR 3.4 [1.4-8.1]). Cardiac mortality differences according to AI model prediction remained after age adjustment (positive: HR 5.9 [3.3-10.4], p<0.001; uncertain: HR 2.8 [1.2-6.8], p=0.02). Conclusion: A novel AI model to detect HFpEF based on a single routinely acquired TTE videoclip also identifies those at increased risk of cardiac mortality.

Introduction: Artificial Intelligence-enabled electrocardiogram (AI-ECG) can estimate the probability of having atrial fibrillation (AFib) while in sinus rhythm. Recurrent stroke in subjects without known AFib may be due to undiagnosed AFib. We evaluated the association between AI-ECG AFib and recurrent ischemic stroke and with incident AFib Methods: Consecutive adults admitted to a Mayo Clinic or Mayo Clinic Health System hospital or to Olmsted Community Hospital for ischemic stroke between January,1,2000 and June,30,2022 were identified using the Rochester Epidemiology Project. Recurrent ischemic strokes were identified after reviewing medical records and imaging studies. We excluded those with known AFib or without ECGs within 3 years prior to the first ischemic stroke. The AI-ECG network was previously developed using 649,931 ECGs from > 180,000 subjects. The primary endpoint was recurrent ischemic stroke; secondary endpoint was incident AFib. Cox model adjusted for age, sex, and CHA2DS2-VASc2 score variables. Results: We included 4,816 subjects, mean age 70±15 yr, 54% male, 92% white, median CHA2DS2-VASc2 score 4 [3-5], median follow-up 3 [1-7] yr. 280 subjects (6%) had a second stroke and 366 (8%) were diagnosed with incident AFib over follow-up. Those with a positive AI-ECG AFib had an increased risk for recurrent ischemic stroke(HR 1.54, 95% CI 1.21-1.96, P<0.001 Figure A) This association remained significant after adjustment for CHA2DS2-VASc2 score(HR 1.37 95% CI 1.07-1.76, P=0.01). These associations were not affected by anticoagulation use over follow-up. AI-ECG AFib was also associated with incident AFib(HR 2.42 95% CI 1.96-2.98, P<0.001 Figure B) Conclusions: AI-ECG AFib predicts recurrent ischemic stroke in the community, possibly reflecting undiagnosed AFib that can be detected by AI-ECG, as it also predicts incident AFib. These results underscore the potential utility of AI-ECG for risk stratification of patients with ischemic stroke.

Background: Recently, our team developed an end-to-end artificial intelligence (AI) framework that can successfully estimate left ventricular ejection fraction (LVEF) from echocardiogram videos. However, this framework requires specific views (A2C and PLAX), and cannot estimate LVEF for studies from which these views are missing. Goal: To develop a deep learning model to estimate LVEF from any type and number of echocardiogram video(s). Methods: We built a deep learning model for B-mode echocardiogram videos of the left ventricle (A2C, A3C, A4C, or PLAX views), using retrospective transthoracic echocardiography (TTE) data. The model was evaluated with two test datasets: one with random valid (i.e., showing the left ventricle) views, and one with preselected views (both having 433 patients). Later, the model was tested with a prospective patient cohort that had both TTE and point-of-care ultrasound (POCUS) data collected simultaneously (393 patients). Results: On the retrospective TTE test set, we observed excellent performance (five random valid views per study: RMSE of 6.87% and Pearson correlation (PC) of 0.82 on LVEF estimation, and an AUC of 0.96 (95% CI: of 0.94-0.98) on low LVEF classification (<=40%); preselected views: RMSE of 6.75%, PC of 0.83, and an AUC of 0.96 (95% CI: 0.94-0.98)). For both TTE and POCUS data from the prospective cohort, we also observed strong performance (TTE: RMSE of 5.62%, PC of 0.87, and an AUC of 0.97; POCUS: RMSE of 6.12%, PC of 0.84, and an AUC of 0.96). For the prospective test sets as well as for the retrospective TTE test set with preselected views, results reflect the estimated LVEF average of A2C, A3C, A4C, and PLAX videos from each patient. Conclusions: For both TTE and POCUS data, the current model showed strong performance which was comparable to that of our previously established view-specific model. Alongside this comparable performance, the current model surpasses our previous model in terms of flexibility, as it isn't dependent on view classification and allows a broader selection in terms of the number and types of videos per study or patient. This innovative approach could prove particularly beneficial in point-of-care situations in which it may not be feasible to obtain high-quality, view-specific videos.

Introduction: Recently, an artificial intelligence (AI) algorithm was validated for identifying ECG signatures of atrial fibrillation (AF) risk during normal sinus rhythm. Traditionally, socioeconomically disadvantaged populations are less likely to receive novel or guideline-recommended treatments, and have high chronic disease burden, including cardiovascular diseases. There have also been concerns about new technologies creating a digital divide, worsening existing health disparities. Hypothesis: An AI algorithm will detect AF risk unequally across patients with different levels of social determinants of health (SDOH). Methods: The initial prospective, non-randomized, interventional study leveraged EHR and digital technologies to prospectively enroll 1004 patients with a broad range of AI-ECG risks of AF, who received continuous cardiac rhythm monitoring for up to 30 days. The Area Deprivation Index (ADI) was used as a composite measure of SDOH which is comprised of indicators spanning four domains: income, education, employment, and house quality. 904 intervention patients linked to the ADI data were matched to 904 real-world control patients derived from eligible but unenrolled patients. Those with ADI under 33% were defined as having better SODH; and patients with ADI above 33% were defined as having worse SODH. Results: Median age was 74 years and 38% were women. After matching, the intervention and control arms were balanced in all baseline characteristics, including HF, HTN, diabetes, stroke, CAD and ADI. A hazard ratio of 3.11 (P<0.001) was computed when comparing intervention to control for primary AF diagnosis outcome. In patients with worse SDOH, the intervention increased AF detection from 2.6% to 8.8% (HR 3.63 [2.09, 6.32]); in the better SDOH, the intervention increased AF detection from 3.9% to 9.3% (2.41 [1.23, 4.74]). Conclusions: An AI-guided AF screening program increased the detection of AF in both patients with better or worse SDOH. The current study validates the use of AI in historically disadvantaged populations displaying the possibility for broad implementation with decreased possibility of perpetuating disparities.

Introduction: Common genetic variation has been linked to an increased risk of developing dilated cardiomyopathy (DCM). However, the phenotypic profile, outcomes, and biological mechanisms by which such genetic background may predispose individuals to DCM are still unknown. Hypothesis: The primary objectives of this study were to develop a DCM polygenic risk score (PRS) based on prior case control-DCM GWAS and to apply this PRS to a well clinical characterized general population. Methods: The DCM-PRS was derived from 13 DCM risk alleles that were assayed in an unselected cohort (n=1897) of unrelated Olmsted County residents age >45 years who did not have a history of heart failure (HF) or CM and underwent physical examination, echocardiography, and blood analyses. Results: The entire population was equally divided into quartiles based on the DCM-PRS. No significant differences were present in terms of age (median 62 years) and sex (48% male). However, a significant correlation was detected between the DCM-PRS and left ventricle ejection fraction (LVEF) (Beta: -2.68, p<0.001), LV end-systolic diameter (Beta: 0.22, p<0.001), and LV end-diastolic diameter (Beta: 0.14, p=0.01). During a follow-up of 16 years (IQR: 15-17), 47 subjects developed DCM. Using a 3:1 non-DCM vs DCM matched cohort, a logistic regression analysis, found a significant correlation between median DCM-PRS and DCM development [OR=2.27 (1.06, 4.85), p=0.03]. However, there was no association between the DCM-PRS and HF biomarkers (aldosterone, natriuretic peptides, ST2, Galectin-3). Conclusions: DCM-PRS correlates with decreased LVEF, increased LV dimensions, and DCM development in a general population without significant differences in baseline clinical features. The absence of correlation between the DCM-PRS and available HF biomarkers suggests that this genetic background affects LV traits and DCM development independent of these biological mechanisms.

Introduction: Left ventricular diastolic function and filling pressure (FP) can be assessed by echocardiography and cardiac catheterization. Recently, a novel artificial intelligence (AI)-enabled ECG has proven to be effective in identifying increased FP determined by echocardiographic assessment of diastolic function. Aims: We aimed to compare a prognostic value of AI-ECG determination of FP compared to echocardiography and catheterization. Methods: This study included 4,213 patients who had a transthoracic echocardiogram and ECG within 30 days before hemodynamic cardiac catheterization. Increased FP was defined as pulmonary artery wedge pressure >15 mmHg by catheterization and grade 2 or 3 diastolic dysfunction by echocardiography based on diastolic function guidelines. Increased FP by ECG was identified using previously developed AI-ECG. AI-ECG performance was evaluated using the area under the curve (AUC) of the receiver operating curve. We assessed the risk of all-cause mortality using Kaplan-Meier estimate and multivariable Cox proportional-hazards models. Results: AI-ECG had AUCs of 0.8 and 0.7 for predicting increased FP determined by echocardiography and catheterization, respectively. Death was observed in 2,776 (34%) patients over a median follow-up of 7.7 years (IQR 7.1, 8.3). Survival of patients with normal and increased FP by AI-ECG, Echocardiography, and catheterization are shown in Figure 1a. Concordant diagnoses of normal and increased FP by AI-ECG and catheterization or echocardiography had the highest and lowest survival rates, respectively. When FP by AI-ECG was discordant, survival was better predicted by AI-ECG compared to echocardiography and catheterization (Figure 1b). Conclusion: AI-ECG for FP has a potential to discriminate prognosis further compared to FP determined by echocardiography or catheterization.

Background: Early diagnosis of heart failure with reduced ejection fraction (HFrEF) can lead to reduced healthcare costs and improved outcomes. In this study, we evaluated the economic effectiveness of an AI algorithm for detecting low ejection fraction using 12-lead ECG data. Aims: The primary aim was to estimate the health economic impact of the AI-ECG algorithm intervention to detect low ejection fraction. Methods: The study population consisted of 113,719 individuals without a prior history of HFrEF who visited a multistate health system between 2010-2018 with 12-lead ECG data and an echocardiogram within 1 year. The AI-ECG algorithm was run retrospectively, and projected costs were estimated assuming each patient incurred the average cost based on their age group and AI-detected heart failure stage. Hospitalization costs were estimated using Medicare reimbursement rates, and economic benefits from increased survival were estimated using a willingness to pay of $50,000 per quality-adjusted life year (QALY) gained. Results: At the optimal model threshold, the total cost savings per patient (including benefits from QALYs gained) for the AI-ECG algorithm vs. standard-of-care on the overall study population were projected to be $79.11 (95% CI: [$53.77, $104.44]) and $124.93 (95% CI: [$83.99, $165.88]) for the 2 and 4 year periods, respectively. For patients over 50 years old, the projected total cost savings were $85.56 (95% CI: [$56.35, $114.78]) and $135.28 (95% CI: [$88.18, $182.38]) per patient for the 2 and 4 year periods, respectively. See Figure. Conclusions: The AI-ECG algorithm to detect low ejection fraction is a cost-effective early screening tool for HFrEF.