Diagnostic Performance of Breast Cancer Screening Second Reading Process Assisted by AI (IMA-L2)

Therapixel

Status: Enrolling

Conditions: Breast Cancer

Treatments: Device: MammoScreen

Study type: Interventional

Funder types: Industry

Identifiers

NCT05800132

ID-RCB 2022-A01488-35

Details and patient eligibility

About

The goal of this clinical trial is to compare, for the second reading of screening mammograms, the standard of care with an AI assisted scenario.

The main questions it aims to answer are:

• will the experimental pathway be non-inferior to the standard pathway?
• will the experimental pathway be economically superior to the standard pathway?

Participants will receive screening mammograms, as part of the breast cancer screening program. Researchers will compare the interpretation made by the second reader in the standard pathway with the AI assisted interpretation made by a different reader in the experimental pathway.

Full description

All mammograms eligible for a second reading, and included in the study, go through 2 arms:

• Conventional second reading by a radiologist accredited to do second reading in France (L2 control arm),
• Second reading assisted by the AI (L2-AI experimental arm): mammograms will be read by the AI first and only if deemed suspicious, a radiologist accredited to do second reading in France will review them.

The most pejorative assessment among the two arms is used as the final decision: if both arms consider the mammogram as negative, the participant won't be recalled, if one arm consider the mammogram as positive, the participant will be called back for further examinations.

Clinical performances and economical impacts of both scenarii will be compared.

Enrollment: 5,000 estimated patients

Sex: Female

Ages: 50 to 74 years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

• Affiliated to the French social security system
• Whose mammograms meet the following characteristics: 4 (2 per breast, CC and MLO) meeting local regulatory standards and with correct DICOM metadata,
• Whose mammograms have been determined to be normal or to have benign lesions by the radiologist who performed the first reading of the images (L1) in the framework of the organized breast cancer screening program
• Having expressed her willingness to participate in the study, to undergo all the procedures and to make herself available for the expected duration of their participation,
• Having completed and signed the informed consent form.

Exclusion criteria

• Woman with breast implants,
• With clinical symptoms of breast cancer,
• With a history of breast surgery (breast reduction or surgery for benign lesion),
• Pregnant or breastfeeding,
• With medical conditions that may interfere with her ability to understand the requirements of the protocol or give informed consent,
• Deprived of liberty by judicial or administrative order,
• Participant whose mammogram images have at least one of the following defects: poor quality of the images, non-standard mammography (projection, magnification, compression), ML/LM/SIO/XCCL or XCCM view.

Trial design

Primary purpose: Screening

Allocation: Non-Randomized

Interventional model: Crossover Assignment

Masking: None (Open label)

---

5,000 participants in 2 patient groups

L2 - Standard second read

No Intervention group

Description:

Interpretation of mammograms done by a radiologist accredited to do second reading in France.

L2-AI - AI-assisted second read

Experimental group

Description:

Interpretation of mammograms done by the AI-based device. Mammograms deemed suspicious by the AI (i.e., receiving an AI global score higher than a defined threshold) will be interpreted by a radiologist accredited to do second reading in France. Mammograms assessed as negative or low suspicious by the AI won't be further interpreted by a radiologist.

Treatment:

Device: MammoScreen