AI-powered ECG Analysis Using Willem™ Software in High-risk Cardiac Patients (WILLEM)

Idoven

Status: Enrolling

Conditions: Sudden Cardiac Death; Cardiomyopathies; Cardiac Arrhythmias; Cardiac Arrest

Treatments: Diagnostic Test: AI-powered ECG analysis to detect cardiac arrhythmic episodes

Study type: Observational

Funder types: Other; Industry

Identifiers

NCT05890716

1903/21

Details and patient eligibility

About

WILLEM is a multi-center, prospective and retrospective cohort study.

The study will assess the performance of a cloud-based and AI-powered ECG analysis platform, named Willem™, developed to detect arrhythmias and other abnormal cardiac patterns. The main questions it aims to answer are:

1. A new AI-powered ECG analysis platform can automatice the classification and prediction of cardiac arrhythmic episodes at a cardiologist level.
2. This AI-powered ECG analysis can delay or even avoid harmful therapies and severe cardiac adverse events such as sudden death.

The prerequisites for inclusion of patients will be the availability of at least one ECG record in raw data, along with patient clinical data and evolution data after more than 1-year follow-up.

Cardiac electrical signals from multiple medical devices will be collected by cardiology experts after obtaining the informed consent. Every cardiac electrical signal from every subject will be reviewed by a board-certified cardiologist to label the arrhythmias and patterns recorded in those tracings. In order to obtain tracings of relevant information, >95% of the subjects enrolled will have rhythm disorders or abnormal ECG's patterns at the time of enrollment.

Full description

The WILLEM study is an investigator-initiated, multicenter, observational trial aiming to validate a cloud-based AI-powered ECG analysis platform to early diagnose and predict the behavior of cardiac abnormalities and cardiac diseases from patients admitted to cardiovascular units. Model-derived diagnosis will be compared with cardiology expert's diagnosis in a test dataset. Clinical outcomes will be included to assess model prediction capabilities: sensitivity, specificity and accuracy. In this observational study, patients will be randomly divided into two groups: (1) a training group to design new methodologies and algorithms; and (2) a test group to evaluate performance of methodologies aiming to avoid overfitting.

Willem™ AI-powered ECG analysis platform supports the analysis of cardiac electrical signals ≥ 10 seconds onwards obtained from devices in-clinic (E.g., 12-lead ECG devices at hospitals or primary care, telemetries, monitors) and at-home or telemedicine interfaces (E.g., Holter devices, event recorders, 6, 3, 2, 1-lead ECG wearables, textile electrodes and patches for mobile cardiac telemetry).

Enrollment: 5,342 estimated patients

Sex: All

Ages: 4+ years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

• Patient presenting relevant cardiac arrhythmias and cardiac patterns (including supraventricular tachycardias, abnormal ECG patterns, ventricular tachycardias, ventricular fibrillation, pulseless electrical activity or asystole among others) that have been recorded with at least one short-term ECG medical device according to guidelines with ≥1 signal-channel.
• Patient with suspected or diagnosed acute/chronic cardiac diseases (including patients with heart failure, patients with history of cardiac arrhythmias, patients with probable coronary artery diseases, patients with cardiomyopathies, patients with pacemakers or implantable cardioverter-defibrillators (ICD), patients with indication of pacemaker or ICD in current or short-term phase, patients participating in other interventional clinical investigation, patients with hemodynamic instability or acute coronary syndromes, pregnant patients, patients with cancer and chemotherapy, patients with life-expectancy lower than 24 months, patients with in or out-of-hospital cardiac arrest with ventricular fibrillation as first documented rhythm).
• At least one ECG tracing that can be exported in raw data.
• Signed informed consent. Patients unable to consent, it will be requested to an authorized relative.

Exclusion criteria

• Unwillingness or inability to sign study written informed consent.
• Unavailable or suboptimal quality of the electrocardiographic signal in raw data.

Trial design

5,342 participants in 2 patient groups

Train group

Description:

Consecutive patients admitted to the hospital due to cardiac disorders (retrospective and prospective) with at least one relevant ECG record \>10 sec in raw data will be used to design new methodologies and algorithms for cardiac patterns recognition.

Treatment:

Diagnostic Test: AI-powered ECG analysis to detect cardiac arrhythmic episodes

Test group

Description:

Consecutive patients admitted to the hospital due to cardiac disorders (retrospective and prospective) with at least one relevant ECG record \>10 sec in raw data will be used to evaluate performance of methodologies aiming to avoid overfitting. Every 10 patients included in Train group; a new patient is included in the test group.

Treatment:

Diagnostic Test: AI-powered ECG analysis to detect cardiac arrhythmic episodes