RETURN TO ABSTRACT LISTING


Objective:
To determine the optimal projection order and timing after contrast administration for visualization of cancer on contrast-enhanced mammography (CEM).

Materials and Methods:
This institutional review board-approved reader study consists of two groups of consecutive CEM cases performed on patients prior to cancer diagnosis. Forty cases were acquired from 2016 to 2018 with craniocaudal (CC) views performed first (group 1), and 38 cases were acquired from 2019 to 2020 with mediolateral oblique (MLO) views performed first (group 2). The side of pathology was imaged first in either the CC or MLO projection (earlier-obtained imaging), followed by the contralateral side in the same projection. Imaging was then repeated for the other projection (later-obtained imaging). Five readers evaluated cases for cancer visibility, confidence in margins, and cancer conspicuity against background parenchymal enhancement (BPE) using a 5-point Likert scale. Analysis was performed comparing earlier- and later-obtained images as well as CC compared to MLO projection. Objective contrast-to-noise measurements (CNR) were performed. Fleiss’ kappa was used to determine agreement. P-values for reader preferences of lesion evaluation were determined using 3-way ANOVA.

Results:
Seventy-eight female patients were included. Mean age was 58 years. There was no significant difference in menopausal status, breast density, or BPE between group 1 and 2. Mean acquisition time of the earlier- and later-obtained imaging was 2.5 minutes and 4.75 minutes, respectively. Inter-rater agreement for lesion type ranged between k = 0.59 and k = 0.63. In 35/390 (9%) of instances, an individual reader changed the reported lesion type between earlier and later imaging, with most instances (n = 28/35, 80%) reflecting a situation where a finding was downgraded to a less conspicuous finding on later-obtained imaging. There was overall reader preference for earlier-obtained images for evaluating cancer visibility, confidence in margins, and conspicuity of lesion against BPE (p < 0.001). Among all readers, there was preference for CC views over MLO projection for the evaluation of lesion conspicuity against BPE (p = 0.045) and no significant preference for the evaluation of lesion visibility (p = 0.078) or confidence in margins (p = 0.35). There was no difference in CNR based on timing or projection.

Conclusion:
There is institutional variability in both projection order and image acquisition timing for CEM protocol, with a previous systematic review revealing at least 7 different combinations in projection order. Our study demonstrates that earlier-obtained recombined imaging is significantly preferred in cancer lesion characterization, with a few instances demonstrating that biopsy-proven lesions may appear more conspicuously on earlier-obtained imaging (e.g., mass versus non-mass enhancement). There was also a trend of preference of CC over MLO projection. Given better characterization on earlier imaging, CEM protocols should consider prioritizing imaging the side of pathology.