Heart Failure With Preserved Ejection Fraction (HFpEF; Diagnosis), Heart Failure With Reduced Ejection Fraction (HFrEF; Diagnosis), Pulmonary Hypertension (Diagnosis), Valvular Heart Disease Stenosis and Regurgitation (Diagnosis)
Conditions
Keywords
Artificial Intelligence, Machine Learning, Wearable Devices, Portable Devices, Heart Failure, Pulmonary Hypertension, Valvular Heart Disease, Echocardiography, Electrocardiography
Brief summary
This study aims to improve the early detection of undiagnosed heart disease, which causes serious health issues, hospital admissions, and high healthcare costs. Researchers are exploring how artificial intelligence (AI) can analyse routine heart tests, called electrocardiograms (ECGs), to detect heart problems. These tests can be done using both traditional ECG machines and portable, wearable devices like smartwatches, making it easier for people to monitor their heart health at home. While AI has shown promise using past data, this study will involve the collection of ECG data and subsequent testing of its accuracy in real-world settings to ensure it works well for both doctors and patients. The goal is to see if AI can identify conditions like heart muscle weakness, valve issues, and high lung pressure from the ECG data of patients. The researchers will also compare AI's detections with other blood tests commonly used to diagnose heart disease. The AI models that will be used are being tested for research and validation purposes only. They will not be used for clinical decision-making or providing information to influence diagnosis, treatment, or patient care during the study. The AI outputs are not shared with clinicians and will have no impact on the care pathway. This research will demonstrate if AI-powered ECG analysis - whether from traditional or portable devices - can provide a low-cost, non-invasive way to detect heart disease early and improve health assessments.
Interventions
OTHERTraditional 12-lead Electrocardiogram
12-lead ECG investigation is a standard, non-invasive diagnostic procedure used as an intervention to assess participants' cardiac electrical activity. For the purposes of this study, 12-lead ECGs will be collected for the application of AI-ECG models for the detection of HF, VHD, and/or PH and will not be used to inform or alter patients' standard NHS care.
OTHERApple Watch Series 4 Single-lead Electrocardiogram
Single-lead ECG taken using an Apple Watch Series 4 is a non-invasive, participant-initiated recording of cardiac electrical activity through a wearable device. While more limited than a 12-lead ECG, it can capture rhythm abnormalities-such as atrial fibrillation-and offers a convenient method for remote or continuous heart monitoring during the study. For the purposes of this study, single-lead ECGs will be collected for the application of AI-ECG models for the detection of HF, VHD, and/or PH and will not be used to inform or alter patients' standard NHS care.
OTHEREko Core 500 Digital Stethoscope 3-lead Electrocardiogram
Three-lead ECG recorded using the Eko CORE 500 digital stethoscope is a non-invasive, clinician-operated cardiac assessment tool that captures real-time electrical activity of the heart during auscultation. It provides enhanced diagnostic information compared to single-lead recordings, allowing detection of arrhythmias and signs of structural heart disease at the point of care, supporting integrated clinical and digital assessment. For the purposes of this study, 3-lead ECGs will be collected for the application of AI-ECG models for the detection of HF, VHD, and/or PH and will not be used to inform or alter patients' standard NHS care.
OTHERAliveCor KardiaMobile Single- and 6-lead Electrocardiogram
A single- or 6-lead ECG recorded using the AliveCor KardiaMobile 6L device which is a portable, non-invasive method for capturing cardiac electrical activity. Operated by the participant or clinician, the device enables rapid rhythm assessment and detection of abnormalities such as atrial fibrillation. The 6-lead configuration offers more comprehensive data than single-lead recordings, supporting enhanced arrhythmia and conduction analysis in both in-clinic and remote settings. For the purposes of this study, single- and 6-lead ECGs will be collected for the application of AI-ECG models for the detection of HF, VHD, and/or PH and will not be used to inform or alter patients' standard NHS care.
OTHERPhlebotomy for N-terminal pro-B-type natriuretic peptide
A minimally invasive biomarker assessment used to evaluate cardiac wall stress and function. Elevated levels can indicate the presence or severity of heart failure and other forms of structural heart disease, making it a valuable tool for diagnosis, risk stratification, and monitoring of cardiac status throughout the study period. For the purposes of this study, NT-proBNP will be collected to assess its accuracy at detecting HF, PH, and VHD with comparison with AI-ECG detections. The investigators will also evaluate the accuracy of AI-ECG detections combined with NT-pro-BNP, for detecting HF, VHD, and PH. The investigators will not be using NT-proBNP results to inform or alter patients' standard NHS care.
Sponsors
Imperial College London
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to 90 Years
Healthy volunteers
Yes
Inclusion criteria
* Patients aged 18-90 years * No prior formal diagnosis of HF (including systolic and diastolic dysfunction), PH, or VHD * Ability to provide informed consent
Exclusion criteria
* Severe arrhythmia or unstable cardiovascular disease * Prior formal diagnosis of HF (including systolic and diastolic dysfunction), PH, or VHD * Cardiac implantable electronic device in-situ, including a permanent pacemaker or implantable cardioverter defibrillator * Involvement in current research or recent involvement in any research prior to recruitment
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| AI-ECG model classification performance for detection of structural heart disease (SHD) | From enrolment to end of patient's study visit (up to 1 hour) | AI-ECG model classification performance for HF, PH, and VHD, will be assessed for all ECG modalities (single-, 3-, 6-, and 12-lead ECGs) using the area under the receiver operating characteristic (AUROC; pre-defined threshold). |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Additional AI-ECG performance metrics for detection of SHD | From enrolment to end of patient's study visit (up to 1 hour) | Secondary performances measures will include sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and F1 score for all ECG modalities. |
| NT-proBNP performance metrics for detection of SHD | From enrolment to end of patient's study visit (up to 1 hour) | performances measures will include area under the receiver operating characteristic (AUROC; pre-defined threshold), sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and F1 score for all ECG modalities. |
| Combined AI-ECG and NT-proBNP performance analysis for detection of SHD | From enrolment to end of patient's study visit (up to 1 hour) | NT-proBNP and AI-ECG predictions will be combined in a logistic regression model to assess the performance of a combined approach. |
Countries
United Kingdom
Contacts
CONTACTAhmed YM El-Medany, MBChB, MRCP, MSc, FHEA
Outcome results
None listed