4D flow MRI provides unparalleled clinical information enabling the non-invasive evaluation of hemodynamics. One area of interest has been assessing abnormal blood flow patterns in patients with liver cirrhosis and portal hypertension^1-3. High portal pressure can lead to an array of complications related to abnormal hemodynamics; thus, 4D flow MRI has the potential to characterize these abnormal flow patterns and identify patients at risk for complications before they become clinically apparent². However, assessing abdominal vasculature with 4D flow MRI is challenging due to the wide range of velocities, flow rates, and vessel diameters encountered. We hypothesize that adjusting pre-processing steps required to analyze 4D flow MRI⁴ in light of these limitations can significantly improve flow quantification throughout the abdominal vasculature.The study cohort consisted of 16 prospectively recruited patients (54±9yrs, 4 women) with liver cirrhosis (50%, 44%, and 6% Child-Pugh Grade A, B, and C, respectively) and sequelae of portal hypertension (splenomegaly alone or with portosystemic shunts). All patients fasted prior to undergoing non-contrast 4D flow MRI at 3T (MAGNETOM Skyra, Siemens Medical Systems, Erlangen, Germany) with ECG- and respiratory navigator gating in an oblique axial imaging volume to include the hepatic and splenic vasculature with the navigator positioned at the lung-spleen interface. Pulse sequence parameters: spatial res=2.1-2.4 x 2.1-2.4 x 2.5-2.7mm; temporal res=40.8-44.0ms; flip angle=7°; TE=2.6-3.1ms; k-t GRAPPA R=5. Velocity sensitivities of 50 and 100 cm/s were used for systemic/portal venous and arterial flow quantification, respectively. Scans were pre-processed in MatLab (the MathWorks, Natick, MA, USA) using two distinct workflows: (1) random noise error correction > background phase error correction > anti-aliasing (for all acquisitions); (2) background phase error correction > anti-aliasing (only for high venc acquisitions). PC-MRAs were then generated and served as the basis for subsequent segmentation of the vasculature in Mimics (Materialise, Plymouth, MI) as well as masks for flow visualization and quantification in Ensight (CEI, Apex, NC, USA). Vessel segmentation quality was ranked independently by two reviewers and compared via Wilcoxon signed rank tests: completely missed (0); small remnant captured (1); captured with rough boarders and/or no branches (2); captured with clear borders and/or branches (3). Flow quantification was compared with paired t-tests and relative error in flow input and output in 2 areas of the portal system: proximal v. distal main portal vein ± coronary vein and the portal venous confluence.Inter-observer reliability was modest for qualitative segmentation rankings (κ=0.6), but segmentation was significantly improved with the second workflow (p < 0.05) for both reviewers’ rankings, capturing greater vascular anatomy for subsequent flow quantification (Figure 1). Measured peak velocities were similar between the two methods (p > 0.05), but flow measurements were significantly higher (p < 0.05) for most vessels using the second workflow, particularly for the hepatic veins (Figure 2). The second work flow also significantly reduced relative error in flow quantification in the portal system (12±8% v. 35±34%, p<0.001).Optimal use of the algorithm by Walker et al to correct for background offset error caused by eddy currents requires the fitted plane surface of the static tissue to be calculated before any other data manipulation⁴. Applying noise filters or anti-aliasing first, such as in workflow 1, removes critical information required for accurate background offset estimation leading to inferior segmentation and flow quantification when compared to workflow 2. Additionally, areas of slow flow due to large portosystemic shunts were recovered in workflow 2 by omitting random noise correction (e.g. hepatic veins or atrophic portal systems). This should only be done with care but may be appropriate for low venc acquisitions with inherently low noise. Anti-aliasing should also be performed cautiously and only after noise correction as most algorithms cannot deal with multiple wraps and produce artefacts. By applying background offset error correction first and judiciously using noise filters and anti-aliasing, one can achieve more accurate flow quantification and capture additional vascular anatomy with non-contrast enhanced 4D flow MRI. This is critical for this technology’s clinical utility as its longer acquisition and data processing time must be paired with accurate assessment of unparalleled clinical information to guide patient care.In order to accurately capture the wide range of velocities, flow rates, and vessel diameters in the abdomen with 4D flow MRI, it is critical to correct the data for background phase offset errors before any other manipulations are applied and use noise filters and anti-aliasing cautiously. Doing so ensures preservation of clinically important flow data and provides improved quantification accuracy necessary for the technology to be implemented in clinical settings.