Preoperative assessment of breast cancer: Multireader comparison of contrast-enhanced MRI versus the combination of unenhanced MRI and digital breast tomosynthesis

Abstract

PurposeTo compare the sensitivity for breast cancer (BC) and BC size estimation of preoperative contrast-enhanced magnetic resonance imaging (CEMRI) versus combined unenhanced magnetic resonance imaging (UMRI) and digital breast tomosynthesis (DBT). Patients and methodsWe retrospectively included 56 women who underwent DBT and preoperative 1.5 T CEMRI between January 2016–February 2017. Three readers with 2–10 years of experience in CEMRI and DBT, blinded to pathology, independently reviewed CEMRI (diffusion-weighted imaging [DWI], T2-weighted imaging, pre- and post-contrast T1-weighted imaging) and a combination of UMRI (DWI and pre-contrast T1-weighted imaging) and DBT. We calculated per-lesion sensitivity of CEMRI and UMRI + DBT, and the agreement between CEMRI, UMRI and DBT versus pathology in assessing cancer size (Bland-Altman analysis). Logistic regression was performed to assess features predictive of cancer missing. ResultsWe included 70 lesions (64% invasive BC, 36% ductal carcinoma in situ or invasive BC with in situ component). UMRI + DBT showed lower sensitivity (86–89%) than CEMRI (94–100%), with a significant difference for the most experienced reader only (p = 0.008). False-positives were fewer with UMRI + DBT (4–5) than with CEMRI (18–25), regardless of the reader (p = 0.001–0.005). For lesion size, UMRI showed closer limits of agreement with pathology than CEMRI or DBT. Cancer size ≤1 cm was the only independent predictor for cancer missing for both imaging strategies (Odds ratio 8.62 for CEMRI and 19.16 for UMRI + DBT). ConclusionsUMRI + DBT showed comparable sensitivity and less false-positives than CEMRI in the preoperative assessment of BC. UMRI was the most accurate tool to assess cancer size.