Deep Learning of Electrocardiograms in Sinus Rhythm From US Veterans to Predict Atrial Fibrillation

Unlocking the diagnostic potential of electrocardiograms through information transfer from cardiac magnetic resonance imaging

Cardiovascular diseases (CVD) can be diagnosed using various diagnostic modalities. The electrocardiogram (ECG) is a cost-effective and widely available diagnostic aid that provides functional information of the heart. However, its ability to classify and spatially localise CVD is limited. In contrast, cardiac magnetic resonance (CMR) imaging provides detailed structural information of the heart and thus enables evidence-based diagnosis of CVD, but long scan times and high costs limit its use in clinical routine. In this work, we present a deep learning strategy for cost-effective and comprehensive cardiac screening solely from ECG. Our approach combines multimodal contrastive learning with masked data modelling to transfer domain-specific information from CMR imaging to ECG representations. In extensive experiments using data from 40,044 UK Biobank subjects, we demonstrate the utility and generalisability of our method for subject-specific risk prediction of CVD and the prediction of cardiac phenotypes using only ECG data. Specifically, our novel multimodal pre-training paradigm improves performance by up to 12.19% for risk prediction and 27.59% for phenotype prediction. In a qualitative analysis, we demonstrate that our learned ECG representations incorporate information from CMR image regions of interest. Our entire pipeline is publicly available at https://github.com/oetu/MMCL-ECG-CMR.

Prediction of incident atrial fibrillation using deep learning, clinical models, and polygenic scores

BACKGROUND AND AIMS

Deep learning applied to electrocardiograms (ECG-AI) is an emerging approach for predicting atrial fibrillation or flutter (AF). This study introduces an ECG-AI model developed and tested at a tertiary cardiac centre, comparing its performance with clinical models and AF polygenic score (PGS).

METHODS

Electrocardiograms in sinus rhythm from the Montreal Heart Institute were analysed, excluding those from patients with pre-existing AF. The primary outcome was incident AF at 5 years. An ECG-AI model was developed by splitting patients into non-overlapping data sets: 70% for training, 10% for validation, and 20% for testing. The performance of ECG-AI, clinical models, and PGS was assessed in the test data set. The ECG-AI model was externally validated in the Medical Information Mart for Intensive Care-IV (MIMIC-IV) hospital data set.

RESULTS

A total of 669 782 ECGs from 145 323 patients were included. Mean age was 61 ± 15 years, and 58% were male. The primary outcome was observed in 15% of patients, and the ECG-AI model showed an area under the receiver operating characteristic (AUC-ROC) curve of .78. In time-to-event analysis including the first ECG, ECG-AI inference of high risk identified 26% of the population with a 4.3-fold increased risk of incident AF (95% confidence interval: 4.02-4.57). In a subgroup analysis of 2301 patients, ECG-AI outperformed CHARGE-AF (AUC-ROC = .62) and PGS (AUC-ROC = .59). Adding PGS and CHARGE-AF to ECG-AI improved goodness of fit (likelihood ratio test P < .001), with minimal changes to the AUC-ROC (.76-.77). In the external validation cohort (mean age 59 ± 18 years, 47% male, median follow-up 1.1 year), ECG-AI model performance remained consistent (AUC-ROC = .77).

CONCLUSIONS

ECG-AI provides an accurate tool to predict new-onset AF in a tertiary cardiac centre, surpassing clinical and PGS.

Serum Potassium Monitoring Using AI-Enabled Smartwatch Electrocardiograms

BACKGROUND

Hyperkalemia, characterized by elevated serum potassium levels, heightens the risk of sudden cardiac death, particularly increasing risk for individuals with chronic kidney disease and end-stage renal disease (ESRD). Traditional laboratory test monitoring is resource-heavy, invasive, and unable to provide continuous tracking. Wearable technologies like smartwatches with electrocardiogram (ECG) capabilities are emerging as valuable tools for remote monitoring, potentially allowing for personalized monitoring with artificial intelligence (AI)-ECG interpretation.

OBJECTIVES

The purpose of this study was to develop an AI-ECG algorithm to predict serum potassium level in ESRD patients with smartwatch-generated ECG waveforms.

METHODS

A cohort of 152,508 patients with 293,557 ECGs paired serum potassium levels obtained within 1 hour at Cedars Sinai Medical Center was used to train an AI-ECG model ("Kardio-Net") to predict serum potassium level. The model was further fine-tuned on 4,337 ECGs from 1,463 patients with ESRD using inputs from 12- and single-lead ECGs. Kardio-Net was evaluated in held-out test cohorts from Cedars Sinai Medical Center and Stanford Healthcare (SHC) as well as a prospective international cohort of 40 ESRD patients with smartwatch ECGs at Chang Gung Memorial Hospital.

RESULTS

The Kardio-Net, when applied to 12-lead ECGs, identified severe hyperkalemia (>6.5 mEq/L) with an AUC of 0.852 (95% CI: 0.745-0.956) and a mean absolute error (MAE) of 0.527 mEq/L. In external validation at SHC, the model achieved an AUC of 0.849 (95% CI: 0.823-0.875) and an MAE of 0.599 mEq/L. For single-lead ECGs, Kardio-Net detected severe hyperkalemia with an AUC of 0.876 (95% CI: 0.765-0.987) in the primary cohort and had an MAE of 0.575 mEq/L. In the external SHC validation, the AUC was 0.807 (95% CI: 0.778-0.835) with an MAE of 0.740 mEq/L. Using prospectively obtained smartwatch data, the AUC was 0.831 (95% CI: 0.693-0.975), with an MAE of 0.580 mEq/L.

CONCLUSIONS

We validate a deep learning model to predict serum potassium levels from both 12-lead ECGs and single-lead smartwatch data, demonstrating its utility for remote monitoring of hyperkalemia.

Rhythm-Ready: Harnessing Smart Devices to Detect and Manage Arrhythmias

PURPOSE OF REVIEW

To survey recent progress in the application of implantable and wearable sensors to detection and management of cardiac arrhythmias.

RECENT FINDINGS

Sensor-enabled strategies are critical for the detection, prediction and management of arrhythmias. In the last several years, great innovation has occurred in the types of devices (implanted and wearable) that are available and the data they collect. The integration of artificial intelligence solutions into sensor-enabled strategies has set the stage for a new generation of smart devices that augment the human clinician. Smart devices enhanced by new sensor technologies and Artificial Intelligence (AI) algorithms promise to reshape the care of cardiac arrhythmias.

Left atrial function during exercise stress echocardiography as a sign of paroxysmal/persistent atrial fibrillation

OBJECTIVE

Atrial cardiomyopathy is closely associated with atrial fibrillation (AF), and some patients exhibit no dysfunction at rest but demonstrate evident changes in left atrial (LA) function and LA volume during exercise. This study aimed to identify distinguishing signs during exercise stress echocardiography (ESE) among patients in sinus rhythm (SR), with and without history of paroxysmal/persistent AF (PAF).

METHODS

A prospective cohort of 1055 patients in SR was enrolled across 12 centers. The main study cohort was divided into two groups: the modeling group (n = 513) and the verification group (n = 542). All patients underwent ESE, which included B-lines, LA volume index (LAVi), and LA strain of the reservoir phase (LASr).

RESULTS

Age, resting and stress LAVi and LASr, and B-lines were identified as a combination of detectors for PAF in both groups. In the entire cohort, aside from resting and stress LAVi and LASr, additional parameters differentiating PAF and non-PAF patients were the presence of systemic hypertension, exercise E/e' > 7, worse right ventricle (RV) contraction during exercise (∆ tricuspid annular plane systolic excursion < 5 mm), a lower left ventricular contractile reserve (< 1.6), and a reduced chronotropic reserve (heart rate reserve < 1.64). The composite score, summing all 9 items, yielded a score of > 4 as the best sensitivity (79%) and specificity (65%).

CONCLUSION

ESE can complement rest echocardiography in the identification of previous PAF in patients with SR through the evaluation of LA functional reservoir and volume reserve, LV chronotropic, diastolic, and systolic reserve, and RV contractile reserve.

Artificial intelligence estimated electrocardiographic age as a recurrence predictor after atrial fibrillation catheter ablation

The application of artificial intelligence (AI) algorithms to 12-lead electrocardiogram (ECG) provides promising age prediction models. We explored whether the gap between the pre-procedural AI-ECG age and chronological age can predict atrial fibrillation (AF) recurrence after catheter ablation. We validated a pre-trained residual network-based model for age prediction on four multinational datasets. Then we estimated AI-ECG age using a pre-procedural sinus rhythm ECG among individuals on anti-arrhythmic drugs who underwent de-novo AF catheter ablation from two independent AF ablation cohorts. We categorized the AI-ECG age gap based on the mean absolute error of the AI-ECG age gap obtained from four model validation datasets; aged-ECG (≥10 years) and normal ECG age (<10 years) groups. In the two AF ablation cohorts, aged-ECG was associated with a significantly increased risk of AF recurrence compared to the normal ECG age group. These associations were independent of chronological age or left atrial diameter. In summary, a pre-procedural AI-ECG age has a prognostic value for AF recurrence after catheter ablation.

Innovative approaches to atrial fibrillation prediction: should polygenic scores and machine learning be implemented in clinical practice?

Atrial fibrillation (AF) prediction and screening are of important clinical interest because of the potential to prevent serious adverse events. Devices capable of detecting short episodes of arrhythmia are now widely available. Although it has recently been suggested that some high-risk patients with AF detected on implantable devices may benefit from anticoagulation, long-term management remains challenging in lower-risk patients and in those with AF detected on monitors or wearable devices as the development of clinically meaningful arrhythmia burden in this group remains unknown. Identification and prediction of clinically relevant AF is therefore of unprecedented importance to the cardiologic community. Family history and underlying genetic markers are important risk factors for AF. Recent studies suggest a good predictive ability of polygenic risk scores, with a possible additive value to clinical AF prediction scores. Artificial intelligence, enabled by the exponentially increasing computing power and digital data sets, has gained traction in the past decade and is of increasing interest in AF prediction using a single or multiple lead sinus rhythm electrocardiogram. Integrating these novel approaches could help predict AF substrate severity, thereby potentially improving the effectiveness of AF screening and personalizing the management of patients presenting with conditions such as embolic stroke of undetermined source or subclinical AF. This review presents current evidence surrounding deep learning and polygenic risk scores in the prediction of incident AF and provides a futuristic outlook on possible ways of implementing these modalities into clinical practice, while considering current limitations and required areas of improvement.

Unlocking Hidden Risks: Harnessing Artificial Intelligence (AI) to Detect Subclinical Conditions from an Electrocardiogram (ECG)

Recent artificial intelligence (AI) advancements in cardiovascular medicine offer potential enhancements in diagnosis, prediction, treatment, and outcomes. This article aims to provide a basic understanding of AI enabled ECG technology. Specific conditions and findings will be discussed, followed by reviewing associated terminology and methodology. In the appendix, definitions of AUC versus accuracy are explained. The application of deep learning models enables detecting diseases from normal electrocardiograms at accuracy not previously achieved by technology or human experts. Results with AI enabled ECG are encouraging as they considerably exceeded current screening models for specific conditions (i.e., atrial fibrillation, left ventricular dysfunction, aortic stenosis, and hypertrophic cardiomyopathy). This could potentially lead to a revitalization of the utilization of the ECG in the insurance domain. While we are embracing the findings with this rapidly evolving technology, but cautious optimism is still necessary at this point.

Consider this a WARNing

Atrial fibrillation (AF) prediction can be valuable at many timescales and in many populations. In this issue of Patterns, Gavidia et al. train a model called WARN for short-term prediction of AF in the timescale of minutes in patients wearing 24-h continuous Holter electrocardiograms. The ability to predict near-term (e.g., 30 min) AF has the potential to enable preventive therapies with rapid mechanisms of action (e.g., oral anticoagulation, anti-arrhythmic drugs). In this way, efficient, continuous, and algorithmic monitoring of AF risk could reduce burden on healthcare workers and represents a valuable clinical pursuit.

Early warning of atrial fibrillation using deep learning

Atrial fibrillation (AF), the most prevalent cardiac rhythm disorder, significantly increases hospitalization and health risks. Reverting from AF to sinus rhythm (SR) often requires intensive interventions. This study presents a deep-learning model capable of predicting the transition from SR to AF on average 30.8 min before the onset appears, with an accuracy of 83% and an F1 score of 85% on the test data. This performance was obtained from R-to-R interval signals, which can be accessible from wearable technology. Our model, entitled Warning of Atrial Fibrillation (WARN), consists of a deep convolutional neural network trained and validated on 24-h Holter electrocardiogram data from 280 patients, with 70 additional patients used for testing and further evaluation on 33 patients from two external centers. The low computational cost of WARN makes it ideal for integration into wearable technology, allowing for continuous heart monitoring and early AF detection, which can potentially reduce emergency interventions and improve patient outcomes.

Opportunistic Screening of Chronic Liver Disease with Deep Learning Enhanced Echocardiography

Importance

Chronic liver disease affects more than 1.5 billion adults worldwide, however the majority of cases are asymptomatic and undiagnosed. Echocardiography is broadly performed and visualizes the liver; but this information is not leveraged.

Objective

To develop and evaluate a deep learning algorithm on echocardiography videos to enable opportunistic screening for chronic liver disease.

Design

Retrospective observational cohorts.

Setting

Two large urban academic medical centers.

Participants

Adult patients who received echocardiography and abdominal imaging (either abdominal ultrasound or abdominal magnetic resonance imaging) with ≤30 days between tests, between July 4, 2012, to June 4, 2022.

Exposure

Deep learning model predictions from a deep-learning computer vision pipeline that identifies subcostal view echocardiogram videos and detects the presence of cirrhosis or steatotic liver disease (SLD).

Main Outcome and Measures

Clinical diagnosis by paired abdominal ultrasound or magnetic resonance imaging (MRI).

Results

A total of 1,596,640 echocardiogram videos (66,922 studies from 24,276 patients) from Cedars-Sinai Medical Center (CSMC) were used to develop EchoNet-Liver, an automated pipeline that identifies high quality subcostal images from echocardiogram studies and detects the presence of cirrhosis or SLD. In the held-out CSMC test cohort, EchoNet-Liver was able to detect the presence of cirrhosis with an AUC of 0.837 (0.789 - 0.880) and SLD with an AUC of 0.799 (0.758 - 0.837). In a separate test cohort with paired abdominal MRIs, cirrhosis was detected with an AUC of 0.704 (0.689-0.718) and SLD was detected with an AUC of 0.726 (0.659-0.790). In an external test cohort of 106 patients (n = 5,280 videos), the model detected cirrhosis with an AUC of 0.830 (0.738 - 0.909) and SLD with an AUC of 0.768 (0.652 - 0.875).

Conclusions and Relevance

Deep learning assessment of clinical echocardiography enables opportunistic screening of SLD and cirrhosis. Application of this algorithm may identify patients who may benefit from further diagnostic testing and treatment for chronic liver disease.

The potential of artificial intelligence to revolutionize health care delivery, research, and education in cardiac electrophysiology

The field of electrophysiology (EP) has benefited from numerous seminal innovations and discoveries that have enabled clinicians to deliver therapies and interventions that save lives and promote quality of life. The rapid pace of innovation in EP may be hindered by several challenges including the aging population with increasing morbidity, the availability of multiple costly therapies that, in many instances, confer minor incremental benefit, the limitations of healthcare reimbursement, the lack of response to therapies by some patients, and the complications of the invasive procedures performed. To overcome these challenges and continue on a steadfast path of transformative innovation, the EP community must comprehensively explore how artificial intelligence (AI) can be applied to healthcare delivery, research, and education and consider all opportunities in which AI can catalyze innovation; create workflow, research, and education efficiencies; and improve patient outcomes at a lower cost. In this white paper, we define AI and discuss the potential of AI to revolutionize the EP field. We also address the requirements for implementing, maintaining, and enhancing quality when using AI and consider ethical, operational, and regulatory aspects of AI implementation. This manuscript will be followed by several perspective papers that will expand on some of these topics.

Deep learning evaluation of echocardiograms to identify occult atrial fibrillation

Atrial fibrillation (AF) often escapes detection, given its frequent paroxysmal and asymptomatic presentation. Deep learning of transthoracic echocardiograms (TTEs), which have structural information, could help identify occult AF. We created a two-stage deep learning algorithm using a video-based convolutional neural network model that (1) distinguished whether TTEs were in sinus rhythm or AF and then (2) predicted which of the TTEs in sinus rhythm were in patients who had experienced AF within 90 days. Our model, trained on 111,319 TTE videos, distinguished TTEs in AF from those in sinus rhythm with high accuracy in a held-out test cohort (AUC 0.96 (0.95-0.96), AUPRC 0.91 (0.90-0.92)). Among TTEs in sinus rhythm, the model predicted the presence of concurrent paroxysmal AF (AUC 0.74 (0.71-0.77), AUPRC 0.19 (0.16-0.23)). Model discrimination remained similar in an external cohort of 10,203 TTEs (AUC of 0.69 (0.67-0.70), AUPRC 0.34 (0.31-0.36)). Performance held across patients who were women (AUC 0.76 (0.72-0.81)), older than 65 years (0.73 (0.69-0.76)), or had a CHA2DS2VASc ≥2 (0.73 (0.79-0.77)). The model performed better than using clinical risk factors (AUC 0.64 (0.62-0.67)), TTE measurements (0.64 (0.62-0.67)), left atrial size (0.63 (0.62-0.64)), or CHA2DS2VASc (0.61 (0.60-0.62)). An ensemble model in a cohort subset combining the TTE model with an electrocardiogram (ECGs) deep learning model performed better than using the ECG model alone (AUC 0.81 vs. 0.79, p = 0.01). Deep learning using TTEs can predict patients with active or occult AF and could be used for opportunistic AF screening that could lead to earlier treatment.