Following approval from the Ethics Committee No. 118/20, this study was conducted at the Radiodiagnostics Department of the Maggiore della Carità Hospital in Novara and the University of Eastern Piedmont, Italy, from May 1, 2021, to October 17, 2022. A total of 100 women, with an average age of 57.5 years (range 33–82 years) ± 12 ds, all of whom had histologically confirmed breast cancer and were scheduled for surgical intervention, were consecutively and prospectively enrolled.

Inclusion criteria were as follows:

Women scheduled for surgery with a histologically confirmed invasive breast carcinoma (T1–2);

Age > 30 years;

Provision of written informed consent;

Negative medical history for adverse events related to the use of iodinated contrast agents;

Normal renal function, verified by creatinine and glomerular filtration rate measurements.

Exclusion criteria were as follows:

CEM technique

CEM was performed using the Selenia Dimensions Mammography System® (Hologic, USA). Before conducting CEM, venous access was established in the forearm, through which a low-osmolarity iodinated contrast agent (Iomeron 350) was administered in a single dose using an automatic injector (Bracco Injeneering Empower CTA+), at an injection rate of 2–3 ml/sec, followed by a saline flush of 20 ml. Intravenous contrast agent injection was performed with the patient in a seated position.

The administered contrast volume was calculated at 1.5 ml/kg of body weight, up to a maximum of 110 ml. Two min post-injection, dual-energy mammographic projections in cranio-caudal (CC) and medio-lateral oblique (MLO) views were acquired for each breast. The examination was completed with delayed acquisitions in CC and MLO projections for each breast, starting from the 7th min following contrast agent administration.

Radiological image analysis

To assess the contribution of the delayed-phase acquisition in identifying suspicious lesions, three experienced breast radiologists (R1, R2, R3) in service near the Breast Unit section for a minimum of ten years, and for a minimum of two years in the CEM section (since May, 1, 2021) conducted a blind review the CEM images. Although aware of the existence of a histologically confirmed primary tumor, these radiologists were charged with identifying areas of contrast enhancement that suggested a unifocal lesion or raised suspicion of multifocality, multicentricity, and/or bilaterality.

For precise topographical localization of lesions, the breast gland was segmented into five quadrants: upper outer quadrant (UOQ), lower inner quadrant (LIQ), upper inner quadrant (UIQ), lower outer quadrant (LOQ), and retro-areolar area.

In assessing CEM enhancement, the readers employed the ACR BI-RADS® 2022 scale (1 = negative; 2 = benign; 3 = probably benign; 4 = probably malignant; 5 = highly suspicious). For analysis purposes, this scale was simplified into a binary measure, where 0 denoted no lesion or benign lesion (BI-RADS 1/2) and 1 indicated a suspicious or highly suspicious lesion (BI-RADS 3/4/5). This dichotomous approach was adopted because the radiologists, already aware of the presence of the malignant lesion, seldom assigned intermediate values (BI-RADS 3 and 4).

In the evaluation process, two breast radiologists blindly reviewed the entire set of examinations, including low-energy (LE) images and the recombined images (RI) acquired in both the early and delayed phases (protocol A) (Fig. 2).

Fig. 2

Protocol A: Low-energy (LE) image and recombined image in the early phase, as well as subtracted image from the delayed-phase acquisition in a left mediolateral oblique (LMLO) or b left craniocaudal (LCC) view

The third breast radiologist blind-reviewed only the LE images and contrast-enhanced acquisitions from the early phase, excluding the delayed-phase acquisitions (protocol B) (Fig. 3).

Fig. 3

Protocol B: early-phase acquisition followed by subsequent subtracted image in a left mediolateral oblique (LMLO) or b left craniocaudal (LCC) view

The assessments provided by the three readers were then statistically compared to evaluate the reproducibility between the first two radiologists, who assessed the full examination under protocol A, and between these radiologists and the third, who only reviewed early-phase acquisitions under protocol B.

Dosimetry

The Medical Physics Department analyzed 60 examinations comprising 4 tomosyntheses (TOMO) (2 for each breast in CC and MLO projections) and 4 dual-energy mammograms (MLO and CC) for each breast, acquired after contrast medium injection in both “early” and “delayed” modes, resulting in 16 2D acquisitions and four 3D acquisitions per patient. The overall mean glandular dose (AGD) of the protocol, as well as individual AGD values related TOMO and CEM projections, was evaluated.

The dosimetric information (AGD) was retrieved by the local Dose Monitoring System (Graydetector, Elco). The accuracy of the displayed AGD is routinely checked in the quality control program and maintains an accuracy within 10%.

Statistical analysis

The reproducibility of lesion scoring between two readers was evaluated using the Cohen's κ-coefficient, which is expressed as a number between 0 and 1, where 0 indicates no reproducibility, and 1 represents perfect agreement, with values > 0.75 generally indicative of good reproducibility. To ease the interpretation of κ in terms of disagreement between the two observers, a 95% confidence interval (CI) was adopted.

The distribution of κ values was assumed to be approximately normal. To test the significance of the differences between two independents κ values, a normal curve test was applied using the following formula:

$$\text=\frac_- _}_^-_^}}$$

A p-value < 0.05 was considered statistically significant.