Development of an Artificial Intelligence Algorithm to Detect Pathological Repolarization Disorders on the ECG and the Risk of Ventricular Arrhythmias (DEEPECG4U)

The objective of this study is to prospectively validate in real life cohorts from various departments of the APHP our artificial intelligence (deep-learning) models allowing for :

1. automatic measurement of various ECG quantitative features,
2. identification and typing of LQT and risk of TdP.

Study Overview

• Status: Recruiting
• Conditions: Ventricular Arrythmia; Cardiac Disease

Detailed Description

Torsade-de-Pointes (TdP) are potentially fatal ventricular arrhythmias favored by a prolongation of ventricular repolarization (Long QT, LQT). The different types of existing LQT derive from the inhibition of cardiac potassium currents (IKr ; IKs) or the activation of a late sodium current (INaL). These alterations can be of congenital origin (3 types=>cLQT-1:IKs, cLQT-2:IKr, cLQT-3: INaL) or drug-induced (diLQT, via inhibition of IKr). More than 100 drugs have marketing authorization despite a risk of TdP because they have a favorable benefit/risk ratio (e.g. hydroxychloroquine).

QTc, which represents the duration of ventricular repolarization (msec) and corresponds to the time between the beginning of the QRS and the end of the T-wave, corrected by heart rate, is prolonged in all LQT. Specific T-wave abnormalities as a function of the altered currents have been described and helps to discriminate cLQT/diLQT types. Thus, limiting the analysis of the ECG to that of the QTc is not very predictive because the information contained in an ECG is much richer and is not limited to the simple measurement of an interval.

We have recently shown that analysis of ECGs using artificial intelligence (convolutional neural network, deep-learning) identifies elements of the ECG that are more discriminating in the prediction of the type of LQT and the risk of TdP, beyond of QTc. With these techniques, we have developed a model with probabilistic modules that predict the risk of TdP, identify the type of LQT (score ranging from 0 to 100%) and allow for the quantitative measurements of various common ECG parameters (including QTc, heart rate, PR and QRS).

The objective of the project is to prospectively validate in real life cohorts from various departments of the APHP our model allowing for :

1. automatic QTc measurement,
2. identification and typing of LQT and risk of TdP.

• Study Type: Observational
• Enrollment (Estimated): 5000

Contacts and Locations

• Study Contact: Name: Joe-Elie SALEM, PU-PH; Phone Number: 0033 1 42 17 85 35; Email: joe-elie.salem@aphp.fr

Study Locations

• France
  • Paris, France, 75013
    • Recruiting
    • Centre d'Investigation Clinique Paris-Est/Hôpital Pitié-Salpêtrière
    • Contact: Joe-Elie SALEM, PU-PH; Phone Number: 00 33 1 42 17 85 35; Email: joe-elie.salem@aphp.fr

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

Hospitalized patients from various centres within the APHP (cardiology, internal medicine, rhythmology, clinical pharmacology, oncology, dermatology).

Description

Inclusion Criteria:

• Age ≥ 18
• Patients or subjects taken care in recruiting centres for which an ECG is indicated
• No opposition to participation in the study

Exclusion Criteria:

• Medical contraindication for ECG
• Subjects with pacemaker-driven QRS

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort
Cohort; patients with a clinical indication to perform an ECG

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Diagnostic property of an AI- deep learning model	Evaluate the diagnostic properties (specificity, sensitivity, positive predictive value, negative predictive value) of a deep-learning quantitative QTc measurement model with a standardized and validated expert measurement to identify patients with very pathological QTc (≥500msec) within a population of hospitalized patients from various centres.	Day 0

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Identification of patients with congenital long QT	Evaluate an AI-model for identification of patients with congenital long QT, and discriminate the type within a population of hospitalized patients	Day 0
Identification of patients with drug-induced acquired long QT	Evaluate an AI-model for identification of patients with drug-induced acquired long QT	Day 0
Measurement of ECG quantitative features	Evaluate an AI-model for measurements of QT, PR, QRS, heart rate and QTc.	Day 0

Collaborators and Investigators

• Sponsor: Assistance Publique - Hôpitaux de Paris
• Collaborators: UMMISCO - Institute of Research for Development (IRD); CoreLab Banook
• Investigators: Principal Investigator: Joe-Elie SALEM, PU-PH, Assistance Publique - Hôpitaux de Paris

Study record dates

Study Major Dates

• Study Start (Actual): November 28, 2023
• Primary Completion (Estimated): May 1, 2024
• Study Completion (Estimated): June 1, 2024

Study Registration Dates

• First Submitted: April 13, 2023
• First Submitted That Met QC Criteria: April 13, 2023
• First Posted (Actual): April 26, 2023

Study Record Updates

• Last Update Posted (Estimated): December 21, 2023
• Last Update Submitted That Met QC Criteria: December 15, 2023
• Last Verified: December 1, 2023

More Information

Terms related to this study

• Keywords: Artificial intelligence; Ventricular arrhythmias; ECG; QT; Repolarization
• Additional Relevant MeSH Terms: Pathologic Processes; Cardiovascular Diseases; Heart Diseases; Arrhythmias, Cardiac
• Other Study ID Numbers: APHP211441; 2022-A01502-41 (Other Identifier: IDRCB ANSM)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No