Feasibility and potential limitations of abbreviated breast MRI: an observer study using an enriched cohort.

Abstract

Breast cancer screening using magnetic resonance imaging (MRI) has been introduced in Western countries primarily for populations with an elevated risk of breast cancer. We conducted an observer study involving an experimental abbreviated MRI interpretation by Japanese radiologists, using an enriched cohort, to evaluate its feasibility in a screening setting. Eighty-eight breast MRI examinations including 28 cases with breast cancer were enrolled as study subjects. Two radiologists independently reviewed the MR images, first with only two sequences (abbreviated series), and then with all of the images provided for clinical care (full diagnostic series). The difference in sensitivity and specificity was evaluated using McNemar's test. Interobserver agreement was assessed by calculating κ values. A total of 176 breasts including 31 cancers (3 cases with bilateral disease) were included. No significant difference in sensitivity or specificity for either observer was observed between the abbreviated series and the full diagnostic series (observer 1: sensitivity 87.1 vs 87.1%, p=1.00, specificity 91.7 vs 90.3%, p=0.791; observer 2: sensitivity 93.5 vs 96.8%, p=1.00, specificity 83.4 vs 89.7%, p=0.064). Moderate interobserver agreement (κ=0.56) was observed for the abbreviated series, whereas substantial agreement (κ=0.69) was observed for the full diagnostic series. The diagnostic accuracy of the abbreviated breast MRI was not inferior to that of the conventional full diagnostic interpretation, although a slight decline in interobserver agreement was observed.