Artificial intelligence-estimated biological heart age using a 12-lead electrocardiogram predicts mortality and cardiovascular outcomes

Artificial intelligence age prediction using electrocardiogram data: Exploring biological age differences

BACKGROUND

Biological age can be predicted using artificial intelligence (AI) trained on electrocardiograms (ECGs), which is prognostic for mortality and cardiovascular events.

OBJECTIVE

We developed an AI model to predict age from an ECG and compared baseline characteristics to identify determinants of advanced biological age.

METHODS

An AI model was trained on ECGs from cardiology inpatients aged 20-90 years. AI analysis used a convolutional neural network with data divided in an 80:20 ratio (development/internal validation), with external validation undertaken using data from the UK Biobank. Performance and subgroup comparison measures included correlation, difference, and mean absolute difference.

RESULTS

A total of 63,246 patients with 353,704 total ECGs were included. In internal validation, the correlation coefficient was 0.72, with a mean absolute difference between chronological age and AI-predicted age of 9.1 years. The same model performed similarly in external validation. In patients aged 20-29 years, AI-ECG-predicted biological age was greater than chronological age by a mean of 14.3 ± 0.2 years. In patients aged 80-89 years, biological age was lower by a mean of 10.5 ± 0.1 years. Women were biologically younger than men by a mean of 10.7 months (P = .023), and patients with a single ECG were biologically 1.0 years younger than those with multiple ECGs (P < .0001).

CONCLUSION

There are significant between-group differences in AI-ECG-predicted biological age for patient subgroups. Biological age was greater than chronological age in young hospitalized patients and lower than chronological age in older hospitalized patients. Women and patients with a single ECG recorded were biologically younger than men and patients with multiple recorded ECGs.

Determinants of artificial intelligence electrocardiogram-derived age and its association with cardiovascular events and mortality: a systematic review and meta-analysis

Artificial intelligence (AI)-ECG-derived age (AI-ECG age) and Heart Delta Age (HDA)-the difference between AI-ECG and chronological age-are emerging tools for assessing cardiovascular health. We systematically searched PubMed, Embase, Web of Science, and Scopus from inception through September 2024. Seventeen original studies utilizing AI algorithms to measure HDA and cardiovascular risk factors, outcomes, or mortality were included. Data were pooled using random- and fixed-effects models for meta-analysis. Hypertension and diabetes mellitus emerged as the most prevalent factors contributing to higher HDA, while cardiac diseases including myocardial infarction and heart failure demonstrated the most significant impact. Pooled analysis revealed a significant association between elevated HDA and increased risks of all-cause mortality (hazard ratio [HR] 1.62, 95% confidence interval [CI] 1.49-1.77) and cardiovascular mortality (HR 2.12, 95% CI 1.71-2.63). HDA could enhance existing risk models and play a critical role in primary healthcare prevention.

Artificial intelligence electrocardiogram-predicted biological age gap and mortality: Capturing dynamic risk with multiple electrocardiograms

BACKGROUND

Artificial intelligence (AI) can predict biological age from electrocardiograms (ECGs), which is prognostic for mortality. Widely available and inexpensive, serial ECG measurements may enhance individual risk profiles.

OBJECTIVE

We investigated whether repeated measurement of AI-derived biological age identifies divergent biological and chronological aging and whether it significantly improves all-cause mortality hazard estimates.

METHODS

This single-center, retrospective cohort study included cardiology patients aged 20-90 years with ≥ 2 ECGs recorded. An AI model estimated the biological age from each ECG, and the biological age gap (difference from chronological age) was calculated. Survival was analyzed using Cox proportional-hazards models; a fixed-hazard model with a single ECG per patient and a time-varying hazards model for multiple ECGs. Models were evaluated with the log-likelihood ratio test, and overall mortality risk predictions were compared with the C-index.

RESULTS

Among 46,960 patients (337,415 ECGs; median follow-up, 4.5 years), the mean biological aging rate was 0.7 ± 4.1 years/y. Increasing biological age gap was associated with a nonlinear mortality hazard increase, whereas negative gaps had a small protective effect. The multiple-ECG model outperformed the single-ECG model with a higher log-likelihood ratio test value (6280 vs 5225) and improved C-index estimates (0.763 vs 0.747; P = .002). The improvement in predictive accuracy increased with more ECGs per patient, plateauing at ≥ 10 ECGs.

CONCLUSION

Many patients demonstrate biological aging that diverges from chronological aging. AI-derived biological age from a single ECG predicted all-cause mortality, but multiple ECGs significantly increased predictive accuracy. Serial biological age estimates may enhance risk assessment and inform personalized care.

Deep learning and electrocardiography: systematic review of current techniques in cardiovascular disease diagnosis and management

This paper reviews the recent advancements in the application of deep learning combined with electrocardiography (ECG) within the domain of cardiovascular diseases, systematically examining 198 high-quality publications. Through meticulous categorization and hierarchical segmentation, it provides an exhaustive depiction of the current landscape across various cardiovascular ailments. Our study aspires to furnish interested readers with a comprehensive guide, thereby igniting enthusiasm for further, in-depth exploration and research in this realm.

AI-Enhanced ECG Applications in Cardiology: Comprehensive Insights from the Current Literature with a Focus on COVID-19 and Multiple Cardiovascular Conditions

The application of artificial intelligence (AI) in electrocardiography is revolutionizing cardiology and providing essential insights into the consequences of the COVID-19 pandemic. This comprehensive review explores AI-enhanced ECG (AI-ECG) applications in risk prediction and diagnosis of heart diseases, with a dedicated chapter on COVID-19-related complications. Introductory concepts on AI and machine learning (ML) are explained to provide a foundational understanding for those seeking knowledge, supported by examples from the literature and current practices. We analyze AI and ML methods for arrhythmias, heart failure, pulmonary hypertension, mortality prediction, cardiomyopathy, mitral regurgitation, hypertension, pulmonary embolism, and myocardial infarction, comparing their effectiveness from both medical and AI perspectives. Special emphasis is placed on AI applications in COVID-19 and cardiology, including detailed comparisons of different methods, identifying the most suitable AI approaches for specific medical applications and analyzing their strengths, weaknesses, accuracy, clinical relevance, and key findings. Additionally, we explore AI's role in the emerging field of cardio-oncology, particularly in managing chemotherapy-induced cardiotoxicity and detecting cardiac masses. This comprehensive review serves as both an insightful guide and a call to action for further research and collaboration in the integration of AI in cardiology, aiming to enhance precision medicine and optimize clinical decision-making.

Age prediction from 12-lead electrocardiograms using deep learning: a comparison of four models on a contemporary, freely available dataset

Objective.The 12-lead electrocardiogram (ECG) is routine in clinical use and deep learning approaches have been shown to have the identify features not immediately apparent to human interpreters including age and sex. Several models have been published but no direct comparisons exist.Approach.We implemented three previously published models and one unpublished model to predict age and sex from a 12-lead ECG and then compared their performance on an open-access data set.Main results.All models converged and were evaluated on the holdout set. The best preforming age prediction model had a hold-out set mean absolute error of 8.06 years. The best preforming sex prediction model had a hold-out set area under the receiver operating curve of 0.92.Significance.We compared performance of four models on an open-access dataset.

Age Prediction From 12-lead Electrocardiograms Using Deep Learning: A Comparison of Four Models on a Contemporary, Freely Available Dataset

Objective

The 12-lead electrocardiogram (ECG) is routine in clinical use and deep learning approaches have been shown to have the identify features not immediately apparent to human interpreters including age and sex. Several models have been published but no direct comparisons exist.

Approach

We implemented three previously published models and one unpublished model to predict age and sex from a 12-lead ECG and then compared their performance on an open-access data set.

Main results

All models converged and were evaluated on the holdout set. The best preforming age prediction model had a hold-out set mean absolute error of 8.06 years. The best preforming sex prediction model had a hold-out set area under the receiver operating curve of 0.92.

Significance

We compared performance of four models on an open-access dataset.

BioAgeNet: An Age-Informed Convolutional Autoencoder for ECG Clustering Indicating Health

Biological Age (BA) indicates the authentic ageing progression of an individual in relation to their quality of life. The noninvasive identification of BA is crucial in predicting longevity and early age-related diseases and enabling personalized healthcare. Potential biomarkers of BA are vague and need attention. The ageing process stands out as a prominent risk factor for cardiovascular diseases. Consequently, an Electrocardiogram (ECG), the most popular and easily accessible signal, is explored to analyze the effect of age. Numerous studies have delved into supervised deep-learning approaches for ECG analysis, particularly in predicting age. These studies rely on regression-based methods and necessitate additional analysis for extracting health-related insights, such as the correlation of error between Chronological Age and AI-predicted Age with mortality. Moreover, as the shortage of cardiologists' annotated data is apparent, we propose an Age-Informed Convolutional Autoencoder that clusters ECG deep features associated with age to assess the quality of life possessed at the current age. We also proposed a three-step training strategy combining model training and deep ECG features clustering with a controlled initialization. We find that a combination of age and ECG reveals the heart's BA and is a contributing biomarker for estimating the overall BA of the body. This approach marks substantial progress in analyzing age-related impacts on ECG. It provides new perspectives on different cardiovascular disorders and can potentially transform personalized healthcare in the future.

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.

Heart Failure with Preserved Ejection Fraction: The Pathophysiological Mechanisms behind the Clinical Phenotypes and the Therapeutic Approach

Heart failure (HF) with preserved ejection fraction (HFpEF) is an increasingly frequent form and is estimated to be the dominant form of HF. On the other hand, HFpEF is a syndrome with systemic involvement, and it is characterized by multiple cardiac and extracardiac pathophysiological alterations. The increasing prevalence is currently reaching epidemic levels, thereby making HFpEF one of the greatest challenges facing cardiovascular medicine today. Compared to HF with reduced ejection fraction (HFrEF), the medical attitude in the case of HFpEF was a relaxed one towards the disease, despite the fact that it is much more complex, with many problems related to the identification of physiopathogenetic mechanisms and optimal methods of treatment. The current medical challenge is to develop effective therapeutic strategies, because patients suffering from HFpEF have symptoms and quality of life comparable to those with reduced ejection fraction, but the specific medication for HFrEF is ineffective in this situation; for this, we must first understand the pathological mechanisms in detail and correlate them with the clinical presentation. Another important aspect of HFpEF is the diversity of patients that can be identified under the umbrella of this syndrome. Thus, before being able to test and develop effective therapies, we must succeed in grouping patients into several categories, called phenotypes, depending on the pathological pathways and clinical features. This narrative review critiques issues related to the definition, etiology, clinical features, and pathophysiology of HFpEF. We tried to describe in as much detail as possible the clinical and biological phenotypes recognized in the literature in order to better understand the current therapeutic approach and the reason for the limited effectiveness. We have also highlighted possible pathological pathways that can be targeted by the latest research in this field.

Artificial intelligence-enhanced 12-lead electrocardiography for identifying atrial fibrillation during sinus rhythm (AIAFib) trial: protocol for a multicenter retrospective study

Introduction

Atrial fibrillation (AF) is the most common arrhythmia, contributing significantly to morbidity and mortality. In a previous study, we developed a deep neural network for predicting paroxysmal atrial fibrillation (PAF) during sinus rhythm (SR) using digital data from standard 12-lead electrocardiography (ECG). The primary aim of this study is to validate an existing artificial intelligence (AI)-enhanced ECG algorithm for predicting PAF in a multicenter tertiary hospital. The secondary objective is to investigate whether the AI-enhanced ECG is associated with AF-related clinical outcomes.

Methods and analysis

We will conduct a retrospective cohort study of more than 50,000 12-lead ECGs from November 1, 2012, to December 31, 2021, at 10 Korean University Hospitals. Data will be collected from patient records, including baseline demographics, comorbidities, laboratory findings, echocardiographic findings, hospitalizations, and related procedural outcomes, such as AF ablation and mortality. De-identification of ECG data through data encryption and anonymization will be conducted and the data will be analyzed using the AI algorithm previously developed for AF prediction. An area under the receiver operating characteristic curve will be created to test and validate the datasets and assess the AI-enabled ECGs acquired during the sinus rhythm to determine whether AF is present. Kaplan-Meier survival functions will be used to estimate the time to hospitalization, AF-related procedure outcomes, and mortality, with log-rank tests to compare patients with low and high risk of AF by AI. Multivariate Cox proportional hazards regression will estimate the effect of AI-enhanced ECG multimorbidity on clinical outcomes after stratifying patients by AF probability by AI.

Discussion

This study will advance PAF prediction based on AI-enhanced ECGs. This approach is a novel method for risk stratification and emphasizes shared decision-making for early detection and management of patients with newly diagnosed AF. The results may revolutionize PAF management and unveil the wider potential of AI in predicting and managing cardiovascular diseases.

Ethics and dissemination

The study findings will be published in peer-reviewed publications and disseminated at national and international conferences and through social media. This study was approved by the institutional review boards of all participating university hospitals. Data extraction, storage, and management were approved by the data review committees of all institutions.

Clinical Trial Registration

[cris.nih.go.kr], identifier (KCT0007881).

Aging Biomarker Consortium, Zhang W, Che Y, Tang X, Chen S, Song M, Wang L, Sun AJ, Chen HZ, Xu M, Wang M, Pu J, Li Z, Xiao J, Cao CM, Zhang Y, Lu Y, Zhao Y, Wang YJ, Zhang C, Shen T, Zhang W, Tao L, Qu J, Tang YD, Liu GH, Pei G, Li J, Cao F. A biomarker framework for cardiac aging: the Aging Biomarker Consortium consensus statement

Cardiac aging constitutes a significant risk factor for cardiovascular diseases prevalent among the elderly population. Urgent attention is required to prioritize preventive and management strategies for age-related cardiovascular conditions to safeguard the well-being of elderly individuals. In response to this critical challenge, the Aging Biomarker Consortium (ABC) of China has formulated an expert consensus on cardiac aging biomarkers. This consensus draws upon the latest scientific literature and clinical expertise to provide a comprehensive assessment of biomarkers associated with cardiac aging. Furthermore, it presents a standardized methodology for characterizing biomarkers across three dimensions: functional, structural, and humoral. The functional dimension encompasses a broad spectrum of markers that reflect diastolic and systolic functions, sinus node pacing, neuroendocrine secretion, coronary microcirculation, and cardiac metabolism. The structural domain emphasizes imaging markers relevant to concentric cardiac remodeling, coronary artery calcification, and epicardial fat deposition. The humoral aspect underscores various systemic (N) and heart-specific (X) markers, including endocrine hormones, cytokines, and other plasma metabolites. The ABC's primary objective is to establish a robust foundation for assessing cardiac aging, thereby furnishing a dependable reference for clinical applications and future research endeavors. This aims to contribute significantly to the enhancement of cardiovascular health and overall well-being among elderly individuals.