A5028320574- ECG Monitoring and Analysis
- Cardiovascular Function and Risk Factors
- Blood Pressure and Hypertension Studies
- Cardiovascular Disease and Adiposity
- Heart Rate Variability and Autonomic Control
- Health, Environment, Cognitive Aging
- Machine Learning in Healthcare
- Cardiac electrophysiology and arrhythmias
- Cardiovascular Health and Risk Factors
- Liver Disease Diagnosis and Treatment
- Genetic Associations and Epidemiology
- COVID-19 diagnosis using AI
- Atrial Fibrillation Management and Outcomes
- Cardiovascular Effects of Exercise
- Receptor Mechanisms and Signaling
- Cardiac Health and Mental Health
- Photoreceptor and optogenetics research
- Health Systems, Economic Evaluations, Quality of Life
- Cardiomyopathy and Myosin Studies
- Acute Myocardial Infarction Research
- Cardiac Arrhythmias and Treatments
- Machine Learning in Bioinformatics
- Sex and Gender in Healthcare
- Cardiac pacing and defibrillation studies
- Artificial Intelligence in Healthcare
Imperial College London
2023-2025
Lung Institute
2024-2025
Imperial College Healthcare NHS Trust
2024
Hammersmith Hospital
2023-2024
Chelsea and Westminster Hospital NHS Foundation Trust
2023
Importance Hypertension underpins significant global morbidity and mortality. Early lifestyle intervention treatment are effective in reducing adverse outcomes. Artificial intelligence–enhanced electrocardiography (AI-ECG) has been shown to identify a broad spectrum of subclinical disease may be useful for predicting incident hypertension. Objective To develop an AI-ECG risk estimator (AIRE) predict hypertension (AIRE-HTN) stratify hypertension-associated Design, Setting, Participants This...
Abstract Aims Many research databases with anonymised patient data contain electrocardiograms (ECGs) from which traditional identifiers have been removed. We evaluated the ability of artificial intelligence (AI) methods to determine similarity between ECGs and assessed whether they potential be misused re-identify individuals datasets. Methods results utilised a convolutional Siamese neural network (SNN) architecture, derives Euclidean distance metric two input ECGs. A secondary care dataset...
Abstract The electrocardiogram (ECG) can capture obesity-related cardiac changes. Artificial intelligence-enhanced ECG (AI-ECG) identify subclinical disease. We trained an AI-ECG model to predict body mass index (BMI) from the alone. Developed 512,950 12-lead ECGs Beth Israel Deaconess Medical Center (BIDMC), a secondary care cohort, and validated on UK Biobank (UKB) ( n = 42,386), achieved Pearson correlation coefficient (r) of 0.65 0.62, R 2 0.43 0.39 in BIDMC cohort Biobank, respectively...
BACKGROUND: Subtle, prognostically important ECG features may not be apparent to physicians. In the course of supervised machine learning, thousands are identified. These limited conventional parameters and morphology. We aimed investigate whether neural network–derived could used predict future cardiovascular disease mortality have phenotypic genotypic associations. METHODS: extracted 5120 from an artificial intelligence–enabled model trained for 6 simple diagnoses applied unsupervised...
Abstract Background Subtle prognostically-important ECG features may not be apparent to physicians. In the course of supervised machine learning (ML), many thousands are identified. These limited conventional parameters and morphology. Hypothesis Novel neural network (NN)-derived can predict future cardiovascular disease mortality Methods Results We extracted 5120 NN-derived from an AI-ECG model trained for six simple diagnoses applied unsupervised identify three phenogroups. derivation...
<h3>Background</h3> Subtle, prognostically-meaningful ECG features may not be apparent to physicians. In the course of supervised machine learning training, many thousands are identified. These limited conventional parameters and morphology. novel neural network (NN)-derived have clinical, phenotypic, genotypic associations prognostic significance. <h3>Methods Results</h3> We extracted 5120 NN-derived from an AI-ECG model trained for six simple diagnoses applied unsupervised identify three...
Background and Aims Artificial intelligence-enhanced electrocardiograms (AI-ECG) can be used to predict risk of future disease mortality but has not yet been adopted into clinical practice. Existing model predictions lack actionability at an individual patient level, explainability biological plausibility. We sought address these limitations previous AI-ECG approaches by developing the estimator (AIRE) platform. Methods Results The AIRE platform was developed in a secondary care dataset...
Abstract Background Artificial intelligence-enhanced electrocardiograms (AI-ECG) can be used to predict risk of events but existing models suffer from lack actionability, explainability and biological plausibility, therefore are not clinically. Purpose We sought address these limitations previous AI-ECG approaches by developing the estimator (AIRE). Methods AIRE was developed in a dataset 1,163,401 ECGs 189,539 patients secondary care setting USA. Uniquely, uses deep learning with residual...
Abstract Advanced data-driven methods can outperform conventional features in electrocardiogram (ECG) analysis, but often lack interpretability. The variational autoencoder (VAE), a form of unsupervised machine learning, address this shortcoming by extracting comprehensive and interpretable new ECG features. Our novel VAE model, trained on dataset comprising over one million secondary care median beat ECGs, validated using the UK Biobank, reveals 20 independent that capture information...
Abstract Background Obesity, a key risk factor for cardiometabolic diseases, is poorly evaluated by body mass index (BMI), which doesn't account visceral fat and distribution. Given that obesity-related cardiac changes can manifest in surface ECG changes, we hypothesised an AI-ECG model could predict BMI, the discrepancy between AI-predicted actual BMI (delta-BMI) indicate health. We also sought to understand biological mechanisms contributing AI-ECG-derived delta-BMI. Methods The was...
Abstract Background Conduction disease can manifest as a broad QRS complex on the ECG. Broad is conventionally categorised left bundle branch block (LBBB) and right (RBBB) or non-specific intraventricular conduction delay (IVCD), based morphology. These subgroups were coined over century ago have been relevant for heart failure with reduced ejection fraction cardiac resynchronization therapy (CRT). However, there may be more precise phenogroups underlying complexes. Purpose Using...
Abstract Background Females are typically underserved in cardiovascular medicine and often considered lower risk of disease. The use sex as a dichotomous variable for stratification fails to capture the heterogeneity within each sex. Purpose We aimed develop an artificial intelligence-enhanced ECG (AI-ECG)-derived continuous score that can continuum phenotypes sex, with particular goal addressing female disadvantage. Methods trained convolution neural network identify using 12-lead ECG....
Abstract Background The ACC/AHA pooled cohort equations (PCE) calculates the 10-year primary risk of atherosclerotic cardiovascular disease (ASCVD), an indication to initiate prevention statin therapy in international guidelines. validity PCE is limited by its overestimation high-risk cohorts and ethnic heterogeneity. Electrocardiography (ECG) has not been incorporated current ASCVD estimation tools. Artificial Intelligence (AI)-based models are capable capture patterns that informative for...
Abstract Background Dilated cardiomyopathy (DCM) is a primary heart muscle disease characterised by left-ventricular dilatation and reduced contractility, often associated with arrhythmia conduction disease. DCM has genetic component, putting family members of diagnosed patients at risk for the Stratification tools to provide personalized screening advice are currently lacking. Artificial intelligence (AI) based analysis electrocardiogram (ECG) offers potential solution this challenge....
Abstract Background Conventional analysis of the electrocardiogram (ECG) is based on extraction human-defined, visually recognisable features. However, these are not optimised to capture all information content in ECG. The variational autoencoder (VAE), a form unsupervised machine learning, can address this shortcoming by computationally extracting comprehensive and interpretable new ECG features, called latent factors. These factors provide low-dimensional representation that maximises data...
Abstract Background Artificial intelligence-enhanced electrocardiography (AI-ECG) can be used to identify existing disease, but could additionally predict occurrence of future disease and death. Purpose We developed an AI-ECG platform that predicts mortality a wide spectrum arrhythmia cardiovascular disease. Methods The risk estimation (AIRE) was in dataset 1,163,401 ECGs from 189,539 patients secondary care setting the USA. Uniquely, AIRE uses deep learning with residual convolutional...
Abstract Background/Introduction Many research databases contain anonymised electrocardiograms (ECGs) linked to other sensitive information. ECGs hold features unique individuals, potentially enabling subject identification from ECGs. Purpose We assessed if artificial intelligence approaches output ECG pair similarity can re-identify individuals Additionally, we aimed explore clinical risk prediction using over time. Methods used a convolutional Siamese neural network (SNN), with triplet...
Most artificial intelligence-enhanced electrocardiogram (AI-ECG) models used to predict adverse events including death require that the ECGs be stored digitally. However, majority of clinical facilities worldwide store as images. A total 1 163 401 (189 539 patients) from a secondary care data set were available both natively digital traces and PDF digitization pipeline extracted signals PDFs. Separate 1D convolutional neural network (CNN) trained on or digitized ECGs, with discrete-time...
BACKGROUND: Myocardial infarction (MI) is a complex disease caused by both lifestyle and genetic factors. This study aims to investigate the predictive value of risk, in addition traditional cardiovascular risk factors, for recurrent events following early-onset MI. METHODS: The Italian Genetic Study Early-Onset Infarction cohort enrolling patients with MI before 45 years. Monogenic variants causing familial hypercholesterolemia were identified, coronary artery polygenic score (PGS) was...