Deep Learning ECG Evaluation and Clinical Assessment for Competitive Sport Eligibility (VALETUDO)

I.R.C.C.S Ospedale Galeazzi-Sant'Ambrogio

Status: Enrolling

Conditions: Preventive Cardiology; Electrocardiogram; Artificial Intelligence; Sports Cardiology

Study type: Observational

Funder types: Other

Identifiers

NCT06285084

VALETUDO Trial (L4195)

Details and patient eligibility

About

The goal of this observationl study is to evaluate the possibility of building a Deep Learning (DL) model capable of analyzing electrocardiographic traces of athletes and providing information in the form of a probability stratification of cardiovascular disease.

Researchers will enroll a training cohort of 455 participants, evaluated following standard clinical practice for eligibility in competitive sports. The response of the clinical evaluation and ECG traces will be recorded to build a DL model.

Researchers will subsequently enroll a validation cohort of 76 participants. ECG traces will be analyzed to evaluate the accuracy of the model to discriminate participants cleared for sports eligibility versus participants who need further medical tests

Full description

The goal of this observationl study is to evaluate the possibility of building a Deep Learning (DL) model capable of analyzing electrocardiographic traces of athletes and providing information in the form of a probability stratification of cardiovascular disease.

The DL model requires training to be calibrated. The project plans to conduct accuracy evaluations on the validation population (76 athletes) and training trials on a different dataset (455 athletes).

There will be an initial phase of system training. Athletes will be assessed according to current guidelines and the italian cardiological guidelines for competitive sports participation - COCIS, with the required diagnostic tests on a case-by-case basis. At the end of the cardiac evaluation, athletes can be classified as "fit" or "unfit" for competitive activity.

Participants will submit the ECGs of "fit" and "unfit" athletes, categorized into these two groups, to a deep learning algorithm to train the artificial intelligence system.

A population of consecutive athletes will then be recruited to form the validation set for the test. These athletes have indications for evaluation for the granting of competitive fitness, as indicated by the referring sports physicians. In this case as well, athletes in the validation set will be assessed according to guidelines and COCIS with appropriate tests on a case-by-case basis to evaluate fitness for competition.

Participants will subject the ECGs of the validation set athletes to the artificial intelligence model to assess accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and AUC in discriminating athletes judged "fit" from those judged "unfit" for competitive activity after cardiac investigations.

Enrollment: 531 estimated patients

Sex: All

Ages: 18 to 60 years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

• Athletes in need of cardiac or sports medical evaluation for the issuance of competitive eligibility.
• Enlisted athletes involved in sports like soccer or those with mixed or aerobic cardiovascular demands according to the COCIS 2017 classification.
• Aged 18 years or older but not exceeding 60 years.
• No history of cardiovascular disease.
• Signed Informed Consent.

Exclusion criteria

• Athletes engaging in skill-based sports as per the COCIS 2017 classification.
• High clinical probability of cardiovascular disease, such as typical angina or heart failure.
• Pregnancy and/or breastfeeding (confirmed through self-declaration).

Trial design

531 participants in 2 patient groups

Training Cohort

Description:

455 Athletes already evaluated for sports participation clearance, whom ECG and clinical evaluation (cleared - not cleared for competitive sports participation) will be fed into the DL model

Validation Cohort

Description:

76 Athletes evaluated using standard sports eligibility clearance tests and our DL model