While previous studies have reported on supine MRI feasibility [17, 19, 20, 33], prone vs. supine MRI characteristics [11, 33, 34], and prone-to-supine (P2S) MRI study results [10, 12], tumor contrast has yet to be systematically compared between prone and supine breast MRI scans. To address this gap, we utilized ratiometric contrast measurements to compare prone and supine breast MRI acquisitions and assess acceptability of investigational supine MRI (as non-inferior to prone breast MRI tumor contrast) for the first time. Our results showed no statistical difference in tumor contrast between prone and supine breast MRI when acquired separately (independent supine MRI) and suggest that breast orientation change from prone (for SoC imaging) to supine (for surgical planning) is acceptable, clinically, for determining MRI-defined tumor extent in the supine position.
While supine MRI has been demonstrated to produce images of excellent quality using clinically available MRI machines, coils, and sequences;[17, 24, 34, 35] supine MRI is not a replacement for diagnostic prone MRI. From a diagnostic perspective, the prone pendant breast position is advantageous for examining disease extension with little to no respiratory motion artifacts. On the contrary, respiratory motion during ungated supine MRI scans can cause streaking artifacts in the phase encode direction, directly resulting in displaced tissue signal and degraded image quality.[17]
Although motivation to perform prone-to-supine MRI within a single imaging session is strong, the sufficiency of tumor contrast in supplemental supine scans for margin delineation has not been quantified previously. Amongst the two published P2S studies [10, 12], both reported on observed lesion displacement from prone-to-supine patient positioning, but did not otherwise examine tumor contrast. Aribal et al. (2019) did investigate the radiologists’ ability to detect lesions in supine MRI [12]. However, detecting tumor vs. determining its extent are different tasks, the latter requiring high fidelity in tumor margin delineation, which can be evaluated quantitatively by assessing tumor contrast. Our approach was designed to assess image contrast as it relates to segmenting tumor extent in a 3D volume – an important step in surgical planning for optimal (margin negative) breast conserving surgery [22].
P2S supine MRI (without additional contrast injection after the prone MRI) demonstrated inferior tumor contrast relative to SoC prone tumor contrast, with lower tumor-to-fibroglandular contrast found in the P2S supine MRI data. Loss of contrast at later time points (average delay time = 23 min.) may be explained by known Gd-kinetics in which fibroglandular tissue often shows persistent Gd-uptake while tumorous regions washout the Gd-based contrast in the later phases of image acquisition [36, 37].
Unlike the P2S supine image data, independent supine (separate imaging session) MRI exhibited non-inferior contrast compared to prone, indicating that independent supine imaging can produce non-inferior tumor-to-fibroglandular contrast when compared to SoC prone MRI. While supine MRI achieves tumor contrast comparable to prone MRI when acquired separately, patient and clinical care demands associated with multiple visits suggest a single imaging session solution is needed. Given these preliminary prone-to-supine results, a secondary prone-to-supine study is underway which utilizes two boluses, one before prone and one before supine acquisitions, in a single imaging exam session (ClinicalTrials.gov, # NCT00159939). Moving forward, the ratiometric contrast approach used here is a powerful tool for examining and comparing image contrast amongst multiple MRI sequences and breast exam protocols independent of patient orientation or image sequence.
By partitioning regional tumor contrast into tumor-to-fibroglandular and tumor-to-adipose categories, differences associated with specific types of surrounding tissue were identified; however, patient-specific factors such as disease type, lesion type, and receptor status were not examined. In addition, the cohort sizes of each imaging procedure compared were different (independent supine n = 17, P2S supine n = 61, and prone n = 78) due to the differing sizes of two clinical trials. Additional statistical measures were taken to ensure a fair comparison amongst cohorts including Kolmogorov-Smirnov tests for sample normalcy, post-hoc power analysis, and assumptions of non-equal variance amongst the three cohorts.