Flow quantification dependency on background phase correction techniques in 4D‐flow MRI

PURPOSE: To analyze the dependence of flow volume measurements on 3D cine phase-contrast MRI (4D-flow MRI) background phase correction. METHODS: In 31 subjects scanned on a 1.5T MRI scanner, flow volume measurements at 4 vessels were made using phantom corrected 2D phase contrast and 4D flow with background phase correction performed by linear, second, third, and fourth-order polynomial fitting to static tissue. Variations in the amount and distribution of static tissue were made to investigate the influence on flow volume measurements. RESULTS: Bland Altman comparison of 2D phase-contrast and 4D-flow measurements showed low bias (2.3%-4.8%) and relatively large limits of agreement (13.5%-17.6%). Approximately half of this was attributable to sequence and physiological differences between the 2 scan sequences, demonstrated by smaller limits of agreement (5.3%-10.0%) when comparing 4D-flow measurements with differing background phase corrections. Using only 20% of available static tissue points for polynomial fitting resulted in only 1% difference in flow volume measurements. Using asymmetrically distributed static tissue or including nonstatic tissue for polynomial fitting yielded highly variable differences in flow volume measurements, which became more variable with increased polynomial order. Completely asymmetric static tissue selection resulted in high deviations in flow volume measurements (mean > 7%, max = 345%). CONCLUSION: Comparisons between 2D phase-contrast and 4D-flow volume measurements should consider influences from sequence and physiological differences. A subset of static tissue points may be used with low impact on flow measurements, but should avoid the inclusion of nonstatic tissue and avoid asymmetric distribution. Higher-order polynomial fits are more susceptible to inaccurate static tissue selection.