Reproducibility of Breast Density and Background Parenchymal Enhancement Grading on Contrast-Enhanced Mammography With AI Support: A Retrospective Single-Centre Study (BD/BPE_CEM)
Beyond Cancer Detection: An AI Framework for Multidimensional Risk Profiling on Contrast-Enhanced Mammography Acronym: AI-CEM-RISK
Retrospective single-centre observational study in women who underwent contrast-enhanced mammography (CEM). It assesses the reproducibility of breast density (BD) and background parenchymal enhancement (BPE) grading by five radiologists with and without AI support, develops linear regression and deep neural network (DNN) models to predict BD and BPE, tests BD prediction on an external public dataset, and explores a multi-output DNN as a proof of concept for proxy estimates of bone mineral density and systolic blood pressure. This is a secondary analysis of the cohort also described in NCT06838117 and NCT06838130.
Study Overview
Status
Intervention / Treatment
Detailed Description
Women referred for diagnostic CEM at the Interventional Breast Unit of P.O. "A. Perrino", Brindisi, were analyzed. Five breast radiologists independently graded BD (BI-RADS A-D) and BPE (minimal to marked) and consensus served as ground truth. Linear regression and a DNN were compared with a simple linear baseline. Inter-reader agreement was measured with Fleiss' kappa. The external validation used 500 BI-RADS C/D cases from VinDr-Mammo and was restricted to density endpoints. The multi-output analysis was hypothesis-generating; bone density and blood pressure outputs are proxy estimates, not measured values.
Study Type
Enrollment (Actual)
Contacts and Locations
Study Locations
-
-
-
Brindisi, Italy, 72100
- Perrino Hospital
-
-
Participation Criteria
Eligibility Criteria
Ages Eligible for Study
Accepts Healthy Volunteers
Genders Eligible for Study
Sampling Method
Study Population
From 314 women who underwent CEM between January 2022 and December 2023 at P.O. "A. Perrino", Brindisi, 213 met the eligibility criteria.
Description
Inclusion Criteria: women aged 28-80 years referred for CEM for diagnostic work-up; histopathology available for BI-RADS 4/5 findings; inclusion not restricted to malignant cases.
Exclusion Criteria: previous history of malignancy; breast biopsy within three weeks before imaging; contraindications to iodinated contrast medium; incomplete imaging records.
Study Plan
How is the study designed?
Design Details
- Observational Models: COHORT
- Time Perspectives: RETROSPECTIVE
Cohorts and Interventions
Group / Cohort |
Intervention / Treatment |
|---|---|
|
Women undergoing CEM for diagnostic work-up
Women aged 28-80 years referred for CEM at the Interventional Breast Unit, P.O. "A. Perrino", Brindisi. Histopathology was available for BI-RADS 4/5 findings. |
Diagnostic Test: contrast-enhanced mammography (Senographe Pristina, GE Healthcare), performed as routine care. Other: AI-assisted grading of breast density and BPE (linear regression and deep neural network). |
What is the study measuring?
Primary Outcome Measures
Outcome Measure |
Measure Description |
Time Frame |
|---|---|---|
|
Inter-reader agreement for breast density and BPE grading, without and with AI support
Time Frame: Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Fleiss' kappa among five radiologists for BD and for BPE, baseline versus AI-assisted reading. |
Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Secondary Outcome Measures
Outcome Measure |
Time Frame |
|---|---|
|
Prediction error (mean squared error) of linear regression and DNN versus a simple linear baseline, for BD and BPE
Time Frame: Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
|
Classification performance (AUC, precision, recall, F1-score) of the models
Time Frame: Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
|
External validation on 500 VinDr-Mammo BI-RADS C/D cases: MSE, R² and AUC for breast density
Time Frame: Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Retrospective reading sessions on examinations performed between January 2022 and December 2023
|
Collaborators and Investigators
Sponsor
Investigators
- Study Chair: Gianluca Gatta, medicine and surgery, University of Campania Luigi Vanvitelli
- Principal Investigator: Graziella Di Grezia, medicine and surgery, Link Campus University
Publications and helpful links
Study record dates
Study Major Dates
Study Start (Actual)
Primary Completion (Actual)
Study Completion (Actual)
Study Registration Dates
First Submitted
First Submitted That Met QC Criteria
First Posted (Actual)
Study Record Updates
Last Update Posted (Actual)
Last Update Submitted That Met QC Criteria
Last Verified
More Information
Terms related to this study
Other Study ID Numbers
- T_3_2025
Plan for Individual participant data (IPD)
Plan to Share Individual Participant Data (IPD)?
IPD Plan Description
Data are available upon request from the corresponding author
Drug and device information, study documents
Studies a U.S. FDA-regulated drug product
Studies a U.S. FDA-regulated device product
This information was retrieved directly from the website clinicaltrials.gov without any changes. If you have any requests to change, remove or update your study details, please contact register@clinicaltrials.gov. As soon as a change is implemented on clinicaltrials.gov, this will be updated automatically on our website as well.