Validation of Non-invasive Detection of Hyperkalemia by Artificial Intelligence Enhanced Electrocardiography in High Acuity Settings.

BACKGROUND: Artificial intelligence (AI) electrocardiogram (ECG) analysis can enable detection of hyperkalemia. In this validation, we assessed the algorithm's performance in two high acuity settings. METHODS: An emergency department (ED) cohort (February-August 2021) and a mixed intensive care unit (ICU) cohort (August 2017-February 2018) were identified and analyzed separately. For each group, pairs of laboratory-collected potassium and 12 lead ECGs obtained within four hours of each other were identified. The previously developed AI ECG algorithm was subsequently applied to leads I and II of the 12 lead ECGs to screen for hyperkalemia (potassium > 6.0 mEq/L). RESULTS: The ED cohort (N=40,128) had a mean age of 60 years, 48% were male, and 1% (N=351) had hyperkalemia. The area under the curve (AUC) of the AI-ECG to detect hyperkalemia was 0.88, with sensitivity, specificity, positive predictive value, negative predictive value and positive likelihood ratio of, 80%, 80%, 3%, 99.8% and 4.0, respectively, in the ED cohort. Low-eGFR (<30 ml/min) subanalysis yielded AUC, sensitivity, specificity, positive predictive value, negative predictive value and positive likelihood ratio of 0.83, 86%, 60%, 15%, 98% and 2.2, respectively, in the ED cohort. The ICU cohort (N=2,636) had a mean age of 65 years, 60% were male, and 3% (N=87) had hyperkalemia. The AUC for the AI-ECG was 0.88 and yielded sensitivity, specificity, positive predictive value, negative predictive value and positive likelihood ratio of 82%, 82%, 14%, 99% and 4.6, respectively in the ICU cohort. Low-eGFR subanalysis yielded AUC, sensitivity, specificity, positive predictive value, negative predictive value and positive likelihood ratio of 0.85, 88%, 67%, 29%, 97% and 2.7, respectively in the ICU cohort. CONCLUSION: The AI-ECG algorithm demonstrated a high negative predictive value, suggesting that it is useful for ruling out hyperkalemia, but a low positive predictive value, suggesting that it is insufficient for treating hyperkalemia.

An electrocardiogram-based AI algorithm for early detection of pulmonary hypertension.

BACKGROUND: Early diagnosis of pulmonary hypertension (PH) is critical for effective treatment and management. We aimed to develop and externally validate an artificial intelligence algorithm that could serve as a PH screening tool, based on analysis of a standard 12-lead electrocardiogram (ECG). METHODS: The PH Early Detection Algorithm (PH-EDA) is a convolutional neural network developed using retrospective ECG voltage-time data, with patients classified as "PH-likely" or "PH-unlikely" (controls) based on right heart catheterisation or echocardiography. In total, 39 823 PH-likely patients and 219 404 control patients from Mayo Clinic were randomly split into training (48%), validation (12%), and test (40%) sets. ECGs taken within 1 month of PH diagnosis (diagnostic dataset) were used to train the PH-EDA at Mayo Clinic. Performance was tested on diagnostic ECGs within the test sets from Mayo Clinic (n=16 175/87 998 PH-likely/controls) and Vanderbilt University Medical Center (VUMC; n=6045/24 256 PH-likely/controls). Performance was also tested on ECGs taken 6-18 months (pre-emptive dataset), and up to 5 years prior to a PH diagnosis at both sites. RESULTS: Performance testing yielded an area under the receiver operating characteristic curve (AUC) of 0.92 and 0.88 in the diagnostic test set at Mayo Clinic and VUMC, respectively, and 0.86 and 0.81, respectively, in the pre-emptive test set. The AUC remained a minimum of 0.79 at Mayo Clinic and 0.73 at VUMC up to 5 years before diagnosis. CONCLUSION: The PH-EDA can detect PH at diagnosis and 6-18 months prior, demonstrating the potential to accelerate diagnosis and management of this debilitating disease.

The prognostic value of artificial intelligence to predict cardiac amyloidosis in patients with severe aortic stenosis undergoing transcatheter aortic valve replacement.

Aims: Cardiac amyloidosis (CA) is common in patients with severe aortic stenosis (AS) undergoing transcatheter aortic valve replacement (TAVR). Cardiac amyloidosis has poor outcomes, and its assessment in all TAVR patients is costly and challenging. Electrocardiogram (ECG) artificial intelligence (AI) algorithms that screen for CA may be useful to identify at-risk patients. Methods and results: In this retrospective analysis of our institutional National Cardiovascular Disease Registry (NCDR)-TAVR database, patients undergoing TAVR between January 2012 and December 2018 were included. Pre-TAVR CA probability was analysed by an ECG AI predictive model, with >50% risk defined as high probability for CA. Univariable and propensity score covariate adjustment analyses using Cox regression were performed to compare clinical outcomes between patients with high CA probability vs. those with low probability at 1-year follow-up after TAVR. Of 1426 patients who underwent TAVR (mean age 81.0 ± 8.5 years, 57.6% male), 349 (24.4%) had high CA probability on pre-procedure ECG. Only 17 (1.2%) had a clinical diagnosis of CA. After multivariable adjustment, high probability of CA by ECG AI algorithm was significantly associated with increased all-cause mortality [hazard ratio (HR) 1.40, 95% confidence interval (CI) 1.01-1.96, P = 0.046] and higher rates of major adverse cardiovascular events (transient ischaemic attack (TIA)/stroke, myocardial infarction, and heart failure hospitalizations] (HR 1.36, 95% CI 1.01-1.82, P = 0.041), driven primarily by heart failure hospitalizations (HR 1.58, 95% CI 1.13-2.20, P = 0.008) at 1-year follow-up. There were no significant differences in TIA/stroke or myocardial infarction. Conclusion: Artificial intelligence applied to pre-TAVR ECGs identifies a subgroup at higher risk of clinical events. These targeted patients may benefit from further diagnostic evaluation for CA.

Artificial Intelligence Electrocardiography to Predict Atrial Fibrillation in Patients With Chronic Lymphocytic Leukemia.

Background: The use of an artificial intelligence electrocardiography (AI-ECG) algorithm has demonstrated its reliability in predicting the risk of atrial fibrillation (AF) within the general population. Objectives: This study aimed to determine the effectiveness of the AI-ECG score in identifying patients with chronic lymphocytic leukemia (CLL) who are at high risk of developing AF. Methods: We estimated the probability of AF based on AI-ECG among patients with CLL extracted from the Mayo Clinic CLL database. Additionally, we computed the Mayo Clinic CLL AF risk score and determined its ability to predict AF. Results: Among 754 newly diagnosed patients with CLL, 71.4% were male (median age = 69 years). The median baseline AI-ECG score was 0.02 (range = 0-0.93), with a value ≥0.1 indicating high risk. Over a median follow-up of 5.8 years, the estimated 10-year cumulative risk of AF was 26.1%. Patients with an AI-ECG score of ≥0.1 had a significantly higher risk of AF (HR: 3.9; 95% CI: 2.6-5.7; P < 0.001). This heightened risk remained significant (HR: 2.5; 95% CI: 1.6-3.9; P < 0.001) even after adjusting for the Mayo CLL AF risk score, heart failure, chronic kidney disease, and CLL therapy. In a second cohort of CLL patients treated with a Bruton tyrosine kinase inhibitor (n = 220), a pretreatment AI-ECG score ≥0.1 showed a nonsignificant increase in the risk of AF (HR: 1.7; 95% CI: 0.8-3.6; P = 0.19). Conclusions: An AI-ECG algorithm, in conjunction with the Mayo CLL AF risk score, can predict the risk of AF in patients with newly diagnosed CLL. Additional studies are needed to determine the role of AI-ECG in predicting AF risk in CLL patients treated with a Bruton tyrosine kinase inhibitor.

Unlocking the potential of artificial intelligence in electrocardiogram biometrics: age-related changes, anomaly detection, and data authenticity in mobile health platforms.

Aims: Mobile devices such as smartphones and watches can now record single-lead electrocardiograms (ECGs), making wearables a potential screening tool for cardiac and wellness monitoring outside of healthcare settings. Because friends and family often share their smart phones and devices, confirmation that a sample is from a given patient is important before it is added to the electronic health record. Methods and results: We sought to determine whether the application of Siamese neural network would permit the diagnostic ECG sample to serve as both a medical test and biometric identifier. When using similarity scores to discriminate whether a pair of ECGs came from the same patient or different patients, inputs of single-lead and 12-lead medians produced an area under the curve of 0.94 and 0.97, respectively. Conclusion: The similar performance of the single-lead and 12-lead configurations underscores the potential use of mobile devices to monitor cardiac health.

Artificial intelligence-based classification of echocardiographic views.

Aims: Augmenting echocardiography with artificial intelligence would allow for automated assessment of routine parameters and identification of disease patterns not easily recognized otherwise. View classification is an essential first step before deep learning can be applied to the echocardiogram. Methods and results: We trained two- and three-dimensional convolutional neural networks (CNNs) using transthoracic echocardiographic (TTE) studies obtained from 909 patients to classify nine view categories (10 269 videos). Transthoracic echocardiographic studies from 229 patients were used in internal validation (2582 videos). Convolutional neural networks were tested on 100 patients with comprehensive TTE studies (where the two examples chosen by CNNs as most likely to represent a view were evaluated) and 408 patients with five view categories obtained via point-of-care ultrasound (POCUS). The overall accuracy of the two-dimensional CNN was 96.8%, and the averaged area under the curve (AUC) was 0.997 on the comprehensive TTE testing set; these numbers were 98.4% and 0.998, respectively, on the POCUS set. For the three-dimensional CNN, the accuracy and AUC were 96.3% and 0.998 for full TTE studies and 95.0% and 0.996 on POCUS videos, respectively. The positive predictive value, which defined correctly identified predicted views, was higher with two-dimensional rather than three-dimensional networks, exceeding 93% in apical, short-axis aortic valve, and parasternal long-axis left ventricle views. Conclusion: An automated view classifier utilizing CNNs was able to classify cardiac views obtained using TTE and POCUS with high accuracy. The view classifier will facilitate the application of deep learning to echocardiography.

Assessing Biological Age: The Potential of ECG Evaluation Using Artificial Intelligence: JACC Family Series.

Biological age may be a more valuable predictor of morbidity and mortality than a person's chronological age. Mathematical models have been used for decades to predict biological age, but recent developments in artificial intelligence (AI) have led to new capabilities in age estimation. Using deep learning methods to train AI models on hundreds of thousands of electrocardiograms (ECGs) to predict age results in a good, but imperfect, age prediction. The error predicting age using ECG, or the difference between AI-ECG-derived age and chronological age (delta age), may be a surrogate measurement of biological age, as the delta age relates to survival, even after adjusting for chronological age and other covariates associated with total and cardiovascular mortality. The relative affordability, noninvasiveness, and ubiquity of ECGs, combined with ease of access and potential to be integrated with smartphone or wearable technology, presents a potential paradigm shift in assessment of biological age.

Electrocardiography-based Artificial Intelligence Algorithms Aid in Prediction of Long-term Mortality After Kidney Transplantation.

BACKGROUND: Predicting long-term mortality postkidney transplantation (KT) using baseline clinical data presents significant challenges. This study aims to evaluate the predictive power of artificial intelligence (AI)-enabled analysis of preoperative electrocardiograms (ECGs) in forecasting long-term mortality following KT. METHODS: We analyzed preoperative ECGs from KT recipients at three Mayo Clinic sites (Minnesota, Florida, and Arizona) between January 1, 2006, and July 30, 2021. The study involved 6 validated AI algorithms, each trained to predict future development of atrial fibrillation, aortic stenosis, low ejection fraction, hypertrophic cardiomyopathy, amyloid heart disease, and biological age. These algorithms' outputs based on a single preoperative ECG were correlated with patient mortality data. RESULTS: Among 6504 KT recipients included in the study, 1764 (27.1%) died within a median follow-up of 5.7 y (interquartile range: 3.00-9.29 y). All AI-ECG algorithms were independently associated with long-term all-cause mortality (P < 0.001). Notably, few patients had a clinical cardiac diagnosis at the time of transplant, indicating that AI-ECG scores were predictive even in asymptomatic patients. When adjusted for multiple clinical factors such as recipient age, diabetes, and pretransplant dialysis, AI algorithms for atrial fibrillation and aortic stenosis remained independently associated with long-term mortality. These algorithms also improved the C-statistic for predicting overall (C = 0.74) and cardiac-related deaths (C = 0.751). CONCLUSIONS: The findings suggest that AI-enabled preoperative ECG analysis can be a valuable tool in predicting long-term mortality following KT and could aid in identifying patients who may benefit from enhanced cardiac monitoring because of increased risk.

Artificial intelligence-enhanced electrocardiogram for arrhythmogenic right ventricular cardiomyopathy detection.

Aims: ECG abnormalities are often the first signs of arrhythmogenic right ventricular cardiomyopathy (ARVC) and we hypothesized that an artificial intelligence (AI)-enhanced ECG could help identify patients with ARVC and serve as a valuable disease-detection tool. Methods and results: We created a convolutional neural network to detect ARVC using a 12-lead ECG. All patients with ARVC who met the 2010 task force criteria and had disease-causative genetic variants were included. All case ECGs were randomly assigned in an 8:1:1 ratio into training, validation, and testing groups. The case ECGs were age- and sex-matched with control ECGs at our institution in a 1:100 ratio. Seventy-seven patients (51% male; mean age 47.2 ± 19.9), including 56 patients with PKP2, 7 with DSG2, 6 with DSC2, 6 with DSP, and 2 with JUP were included. The model was trained using 61 case ECGs and 5009 control ECGs; validated with 7 case ECGs and 678 control ECGs and tested in 22 case ECGs and 1256 control ECGs. The sensitivity, specificity, positive and negative predictive values of the model were 77.3, 62.9, 3.32, and 99.4%, respectively. The area under the curve for rhythm ECG and median beat ECG was 0.75 and 0.76, respectively. Conclusion: Our study found that the model performed well in excluding ARVC and supports the concept that the AI ECG can serve as a biomarker for ARVC if a larger cohort were available for network training. A multicentre study including patients with ARVC from other centres would be the next step in refining, testing, and validating this algorithm.

Predictive Value of Artificial Intelligence-Enabled Electrocardiography in Patients With Takotsubo Cardiomyopathy.

BACKGROUND: Recent studies have indicated high rates of future major adverse cardiovascular events in patients with Takotsubo cardiomyopathy (TC), but there is no well-established tool for risk stratification. This study sought to evaluate the prognostic value of several artificial intelligence-augmented ECG (AI-ECG) algorithms in patients with TC. METHODS AND RESULTS: This study examined consecutive patients in the prospective and observational Mayo Clinic Takotsubo syndrome registry. Several previously validated AI-ECG algorithms were used for the estimation of ECG- age, probability of low ejection fraction, and probability of atrial fibrillation. Multivariable models were constructed to evaluate the association of AI-ECG and other clinical characteristics with major adverse cardiac events, defined as cardiovascular death, recurrence of TC, nonfatal myocardial infarction, hospitalization for congestive heart failure, and stroke. In the final analysis, 305 patients with TC were studied over a median follow-up of 4.8 years. Patients with future major adverse cardiac events were more likely to be older, have a history of hypertension, congestive heart failure, worse renal function, as well as high-risk AI-ECG findings compared with those without. Multivariable Cox proportional hazards analysis indicated that the presence of 2 or 3 high-risk findings detected by AI-ECG remained a significant predictor of major adverse cardiac events in patients with TC after adjustment by conventional risk factors (hazard ratio, 4.419 [95% CI, 1.833-10.66], P=0.001). CONCLUSIONS: The combined use of AI-ECG algorithms derived from a single 12-lead ECG might detect subtle underlying patterns associated with worse outcomes in patients with TC. This approach might be beneficial for stratifying high-risk patients with TC.

Artificial intelligence-enabled ECG for left ventricular diastolic function and filling pressure.

Assessment of left ventricular diastolic function plays a major role in the diagnosis and prognosis of cardiac diseases, including heart failure with preserved ejection fraction. We aimed to develop an artificial intelligence (AI)-enabled electrocardiogram (ECG) model to identify echocardiographically determined diastolic dysfunction and increased filling pressure. We trained, validated, and tested an AI-enabled ECG in 98,736, 21,963, and 98,763 patients, respectively, who had an ECG and echocardiographic diastolic function assessment within 14 days with no exclusion criteria. It was also tested in 55,248 patients with indeterminate diastolic function by echocardiography. The model was evaluated using the area under the curve (AUC) of the receiver operating characteristic curve, and its prognostic performance was compared to echocardiography. The AUC for detecting increased filling pressure was 0.911. The AUCs to identify diastolic dysfunction grades ≥1, ≥2, and 3 were 0.847, 0.911, and 0.943, respectively. During a median follow-up of 5.9 years, 20,223 (20.5%) died. Patients with increased filling pressure predicted by AI-ECG had higher mortality than those with normal filling pressure, after adjusting for age, sex, and comorbidities in the test group (hazard ratio (HR) 1.7, 95% CI 1.645-1.757) similar to echocardiography and in the indeterminate group (HR 1.34, 95% CI 1.298-1.383). An AI-enabled ECG identifies increased filling pressure and diastolic function grades with a good prognostic value similar to echocardiography. AI-ECG is a simple and promising tool to enhance the detection of diseases associated with diastolic dysfunction and increased diastolic filling pressure.

Artificial intelligence electrocardiogram as a novel screening tool to detect a newly abnormal left ventricular ejection fraction after anthracycline-based cancer therapy.

AIMS: Cardiotoxicity is a serious side effect of anthracycline treatment, most commonly manifesting as a reduction in left ventricular ejection fraction (EF). Early recognition and treatment have been advocated, but robust, convenient, and cost-effective alternatives to cardiac imaging are missing. Recent developments in artificial intelligence (AI) techniques applied to electrocardiograms (ECGs) may fill this gap, but no study so far has demonstrated its merit for the detection of an abnormal EF after anthracycline therapy. METHODS AND RESULTS: Single centre consecutive cohort study of all breast cancer patients with ECG and transthoracic echocardiography (TTE) evaluation before and after (neo)adjuvant anthracycline chemotherapy. Patients with HER2-directed therapy, metastatic disease, second primary malignancy, or pre-existing cardiovascular disease were excluded from the analyses as were patients with EF decline for reasons other than anthracycline-induced cardiotoxicity. Primary readout was the diagnostic performance of AI-ECG by area under the curve (AUC) for EFs < 50%. Of 989 consecutive female breast cancer patients, 22 developed a decline in EF attributed to anthracycline therapy over a follow-up time of 9.8 ± 4.2 years. After exclusion of patients who did not have ECGs within 90 days of a TTE, 20 cases and 683 controls remained. The AI-ECG model detected an EF < 50% and ≤ 35% after anthracycline therapy with an AUC of 0.93 and 0.94, respectively. CONCLUSION: These data support the use of AI-ECG for cardiotoxicity screening after anthracycline-based chemotherapy. This technology could serve as a gatekeeper to more costly cardiac imaging and could enable patients to monitor themselves over long periods of time.

Artificial intelligence electrocardiogram can be used to screen for an abnormal heart function after anthracycline chemotherapy, opening the door to new ways of cost-effective screening of cancer survivors at risk of cardiotoxicity over long periods of time.

Realtime Diagnosis from Electrocardiogram Artificial Intelligence-Guided Screening for Atrial Fibrillation with Long Follow-Up (REGAL): Rationale and design of a pragmatic, decentralized, randomized controlled trial.

BACKGROUND: Atrial fibrillation (AF) is associated with increased risks of stroke and dementia. Early diagnosis and treatment could reduce the disease burden, but AF is often undiagnosed. An artificial intelligence (AI) algorithm has been shown to identify patients with previously unrecognized AF; however, monitoring these high-risk patients has been challenging. Consumer wearable devices could be an alternative to enable long-term follow-up. OBJECTIVES: To test whether Apple Watch, used as a long-term monitoring device, can enable early diagnosis of AF in patients who were identified as having high risk based on AI-ECG. DESIGN: The Realtime diagnosis from Electrocardiogram (ECG) Artificial Intelligence (AI)-Guided Screening for Atrial Fibrillation (AF) with Long Follow-up (REGAL) study is a pragmatic trial that will accrue up to 2,000 older adults with a high likelihood of unrecognized AF determined by AI-ECG to reach our target of 1,420 completed participants. Participants will be 1:1 randomized to intervention or control and will be followed up for 2 years. Patients in the intervention arm will receive or use their existing Apple Watch and iPhone and record a 30-second ECG using the watch routinely or if an abnormal heart rate notification is prompted. The primary outcome is newly diagnosed AF. Secondary outcomes include changes in cognitive function, stroke, major bleeding, and all-cause mortality. The trial will utilize a pragmatic, digitally-enabled, decentralized design to allow patients to consent and receive follow-up remotely without traveling to the study sites. SUMMARY: The REGAL trial will examine whether a consumer wearable device can serve as a long-term monitoring approach in older adults to detect AF and prevent cognitive function decline. If successful, the approach could have significant implications on how future clinical practice can leverage consumer devices for early diagnosis and disease prevention. CLINICALTRIALS: GOV: : NCT05923359.

Patient-Level Artificial Intelligence-Enhanced Electrocardiography in Hypertrophic Cardiomyopathy: Longitudinal Treatment and Clinical Biomarker Correlations.

BACKGROUND: Artificial intelligence (AI) applied to 12-lead electrocardiographs (ECGs) can detect hypertrophic cardiomyopathy (HCM). OBJECTIVES: The purpose of this study was to determine if AI-enhanced ECG (AI-ECG) can track longitudinal therapeutic response and changes in cardiac structure, function, or hemodynamics in obstructive HCM during mavacamten treatment. METHODS: We applied 2 independently developed AI-ECG algorithms (University of California-San Francisco and Mayo Clinic) to serial ECGs (n = 216) from the phase 2 PIONEER-OLE trial of mavacamten for symptomatic obstructive HCM (n = 13 patients, mean age 57.8 years, 69.2% male). Control ECGs from 2,600 age- and sex-matched individuals without HCM were obtained. AI-ECG output was correlated longitudinally to echocardiographic and laboratory metrics of mavacamten treatment response. RESULTS: In the validation cohorts, both algorithms exhibited similar performance for HCM diagnosis, and exhibited mean HCM score decreases during mavacamten treatment: patient-level score reduction ranged from approximately 0.80 to 0.45 for Mayo and 0.70 to 0.35 for USCF algorithms; 11 of 13 patients demonstrated absolute score reduction from start to end of follow-up for both algorithms. HCM scores were significantly associated with other HCM-relevant parameters, including left ventricular outflow tract gradient at rest, postexercise, and with Valsalva, and NT-proBNP level, independent of age and sex (all P < 0.01). For both algorithms, the strongest longitudinal correlation was between AI-ECG HCM score and left ventricular outflow tract gradient postexercise (slope estimate: University of California-San Francisco 0.70 [95% CI: 0.45-0.96], P < 0.0001; Mayo 0.40 [95% CI: 0.11-0.68], P = 0.007). CONCLUSIONS: AI-ECG analysis longitudinally correlated with changes in echocardiographic and laboratory markers during mavacamten treatment in obstructive HCM. These results provide early evidence for a potential paradigm for monitoring HCM therapeutic response.

Identification and risk stratification of coronary disease by artificial intelligence-enabled ECG.

Background: Atherosclerotic cardiovascular disease (ASCVD) is the leading cause of death worldwide, driven primarily by coronary artery disease (CAD). ASCVD risk estimators such as the pooled cohort equations (PCE) facilitate risk stratification and primary prevention of ASCVD but their accuracy is still suboptimal. Methods: Using deep electronic health record data from 7,116,209 patients seen at 70+ hospitals and clinics across 5 states in the USA, we developed an artificial intelligence-based electrocardiogram analysis tool (ECG-AI) to detect CAD and assessed the additive value of ECG-AI-based ASCVD risk stratification to the PCE. We created independent ECG-AI models using separate neural networks including subjects without known history of ASCVD, to identify coronary artery calcium (CAC) score ≥300 Agatston units by computed tomography, obstructive CAD by angiography or procedural intervention, and regional left ventricular akinesis in ≥1 segment by echocardiogram, as a reflection of possible prior myocardial infarction (MI). These were used to assess the utility of ECG-AI-based ASCVD risk stratification in a retrospective observational study consisting of patients with PCE scores and no prior ASCVD. The study period covered all available digitized EHR data, with the first available ECG in 1987 and the last in February 2023. Findings: ECG-AI for identifying CAC ≥300, obstructive CAD, and regional akinesis achieved area under the receiver operating characteristic (AUROC) values of 0.88, 0.85, and 0.94, respectively. An ensembled ECG-AI identified 3, 5, and 10-year risk for acute coronary events and mortality independently and additively to PCE. Hazard ratios for acute coronary events over 3-years in patients without ASCVD that tested positive on 1, 2, or 3 versus 0 disease-specific ECG-AI models at cohort entry were 2.41 (2.14-2.71), 4.23 (3.74-4.78), and 11.75 (10.2-13.52), respectively. Similar stratification was observed in cohorts stratified by PCE or age. Interpretation: ECG-AI has potential to address unmet need for accessible risk stratification in patients in whom PCE under, over, or insufficiently estimates ASCVD risk, and in whom risk assessment over time periods shorter than 10 years is desired. Funding: Anumana.

Saliency maps provide insights into artificial intelligence-based electrocardiography models for detecting hypertrophic cardiomyopathy.

INTRODUCTION: A 12­lead electrocardiography (ECG)-based convolutional neural network (CNN) model can detect hypertrophic cardiomyopathy (HCM). However, since these models do not rely on discrete measurements as inputs, it is not apparent what drives their performance. We hypothesized that saliency maps could be used to visually identify ECG segments that contribute to a CNN's robust classification of HCM. METHODS: We derived a new one­lead (lead I) CNN model based on median beats using the same methodology and cohort used for the original 12­lead CNN model (3047 patients with HCM, and 63,926 sex- and age-matched non-HCM controls). One­lead, median-beat saliency maps were generated and visually evaluated in an independent cohort of 100 patients with a diagnosis of HCM and a high artificial intelligence (AI)-ECG-HCM probability score to determine which ECG segments contributed to the model's detection of HCM. RESULTS: The one­lead, median-beat CNN had an AUC of 0.90 (95% CI 0.89-0.92) for HCM detection, similar to the original 12­lead ECG model. In the independent HCM cohort (n = 100), saliency maps highlighted the ST-T segment in 92 ECGs, the atrial depolarization segment in 12 ECGs, and the QRS complex in 5 ECGs. CONCLUSIONS: Saliency maps of a one­lead, median-beat-based CNN model identified perturbations in ventricular repolarization as the main region of interest in detecting HCM.