A quantitative method to compare regional tumor contrast between prone and supine breast MRI

Abstract

Abstract Purpose For surgical guidance applications, supine breast MRI tumor contrast should be non-inferior to prone MRI, currently considered the standard-of-care. However, comparing image contrast quantitatively between different MRI sequences and breast orientations presents a significant challenge. Herein, we present a method for quantitatively comparing regional tumor contrast in the prone and supine breast MRI orientations for the purpose of tumor localization, and we apply this framework to assess the performance of two investigational supine scans (i.e. independent and prone-to-supine, P2S) compared to diagnostic prone MRI. Methods Patient tumors from two studies (NCT03573804, NCT03573661) were outlined slice-by-slice by a breast radiologist using Gd-enhanced, T1-weighted MRI. Image data were derived from subjects undergoing standard-of-care prone imaging (n = 78), independent supine imaging (n = 17), and P2S supine imaging (n = 61). Normalized tumor contrast was computed between the segmented tumor and neighboring normal tissue regions and compared for statistical differences amongst cohorts and non-inferiority to prone MRI. Results The independent supine cohort possessed non-inferior tumor-to-fibroglandular contrast compared to prone (p = 0.002), while tumor-to-fibroglandular contrast from the P2S supine cohort was found inferior to the prone cohort. However, both investigational supine scans produced non-inferior tumor-to-adipose contrast when compared to prone MRI (p < 0.001 in both cases). Conclusions Regional contrast between tumor and surrounding fibroglandular tissue suffered at later timepoints observed in the P2S supine study, resulting in inferior tumor contrast. However, when contrast-enhanced supine breast MRI is acquired independently, ratiometric comparisons indicate that tumor contrast is non-inferior to prone MRI.