Patient selection
Between September 2017 and November 2023, we conducted a retrospective and multitechnique approach, focused on confirmed breast cancer cases in patients who underwent both MRI and CEM. A comprehensive dataset was compiled, including patient age, breast density pattern, CEM and MRI results, histological characterization and size of each lesion, and any additional lesions discovered during the examinations.
Image evaluation
Two radiologists retrospectively reevaluated all images in both techniques.
Breast density was assessed using ACR criteria and background parenchymal enhancement (BPE) was established by the BI-RADS statements. Suspicious lesions detected in CEM and MRI were classified according to the Breast Imaging Reporting and Data System (BI-RADS) scale, with BI-RADS 1-2 as benign and BI-RADS 3-5 as malignant. Malignant lesions were categorized as unifocal, multifocal, multicentric, or bilateral. Also, the maximum diameter of each lesion was collected in both techniques.
Image Acquisition
CEM was used as a problem-solving technique, complementing DM, digital breast tomosynthesis (DBT), and US. Breast MRI was performed after histological malignancy confirmation. In cases of neoadjuvant chemotherapy, another preoperative MRI was performed to evaluate the tumor response.
MRI studies were performed on a 1.5-Tesla MR system in the prone position, with no breast compression, and using a dedicated four-channel breast coil. Our MRI protocol included the following sequences and parameters: an axial STIR-T2 weighted sequence, an axial diffusion-weighted imaging (DWI) sequence and a dynamic acquisition three-dimensional (3D) axial sequence after the administration of 0.1 mmol/kg of Gadolinium.
All CEM examinations started with an intravenous injection of 1.5 mL/kg of iodinated contrast agent, using an automated rate of 3mL/s followed by a saline flush of 25 mL. At the time of injection, the breast was not compressed to allow for normal tissue perfusion. After a waiting time of 2 min, the breast was compressed and low (28 and 32kV) and high energy (49kV) images were acquired. Recombined subtraction images were formed, highlighting areas of iodine uptake and canceling signal from normal background parenchymal enhancement.
Statistical Analysis
We regarded the pathological anatomy results (both biopsy and surgical specimens) as the gold standard, inclusive of benign findings.
Statistical comparisons were performed using the Wilcoxon test, the McNemar test, the paired T-test, and the Mann-Whitney test. Reliability was evaluated using the intraclass correlation coefficient (ICC). Receiver operating characteristic (ROC) curves were used to quantify the ability to distinguish tumor malignancy. Box plots were used to study the discrepancies between the different lesion size measurements. We considered two-sided p values <0.05 as significant.