Clinical Validation of an Artificial Intelligence Algorithm to Help Interpret Mammograms

University Hospital, Strasbourg, France1,000 enrolled

Overview

This aims to clinically validate, on a large population, a tumor detection aid software which has already been trained on a representative French population (from several hospital centers and liberals from several departments in the west and east of France). This population consists of 1000 patients who have been treated for breast cancer (histologically proven by breast biopsy) and whose investigators have mammograms performed at the time of diagnosis. The control population consists of the unaffected breast of each patient (with the exception of the rare cases of bilateral cancers). This innovative software has the main feature of recognizing healthy breast tissue, allowing the radiologist to focus on breast tissue at risk, improving the management of medical time and the management of "difficult" files.

Study Type

OBSERVATIONAL

Enrollment

1,000

Conditions

Breast Cancer

Eligibility

Sex: FEMALEMin age: 40 YearsMax age: 75 Years

Medical Language ↔ Plain English

Inclusion Criteria: * Adult woman (40 to 75 years old) * Treatment at the University Hospitals of Strasbourg between 2010 and 2020 for breast cancer * including mammography and histological evidence available * Patient who has already given her consent for the reuse of her anonymous data for research purposes Exclusion Criteria: * Woman who expressed her opposition to participating in the study

Locations (1)

Outcomes

Primary Outcomes

Evaluate the diagnostic performance (sensitivity, specificity) for operator tumor detection assisted by mammographic diagnostic aid software, and compare it to the diagnostic performance of an unassisted operator.

This assessment is based on contouring the tumor area on mammograms.

Time frame: Files analysed retrospectively from January 01, 2010 to January 01, 2020 will be examined