Background parenchymal enhancement assessment: Inter- and intra-rater reliability across breast MRI sequences

Abstract Objective To evaluate inter- and intra-rater reliability of background parenchymal enhancement (BPE) assessment across breast MRI sequences. Materials and methods Institutional review board approval was obtained and the requirement for consent was waived. Three radiologists qualitatively categorized BPE on 150 breast MRI using a four-point scale (minimal, mild, moderate or marked) according to BI-RADS category system. According to MR-sequence used for the assessment of BPE, inter-rater and intra-rater reliability across a simulated reading strategy with four options was performed: (1) initial contrast-enhanced (CE) fat-suppressed T1-weighted images (2) initial CE subtracted images (3) maximum-intensity-projection (MIP) of the first CE subtracted images (4) combination of initial CE fat-suppressed T1-weighted, initial CE subtracted and MIP images. Raters repeated BPE assessment of 45 breast MRI four weeks after the initial assessment. Gwet’s AC1 index with ordinal weights was used to assess reliabilities. Results Gwet’s index for the reliability among the three raters was 0.68 (0.63-0.74) using initial contrast-enhanced fat-suppressed T1 weighted images, 0.74 (0.69-0.80) using subtracted images, 0.80 (0.76-0.83) using MIP, 0.80 (0.77-0.84) using a combination of the initial contrast-enhanced fat-suppressed T1 weighted, initial contrast-enhanced subtracted and MIP images. Test-retest reliability was 0.81 (0.60–1.00) for rater 1, 0.77 (0.55-0.98) for rater 2, 0.79 (0.59-0.99) for rater 3 using the combination of initial contrast-enhanced fat-suppressed T1 weighted, initial contrast-enhanced subtracted and MIP images. Conclusions Overall, the combination of all CE MRI images showed the highest reliability of BPE assessment. However, MIP showed a high reliability with lower reading time compared to the combination of all CE MRI images.