Pancreatic steatosis refers to fat accumulation in the pancreas and is implicated in the pathogenesis of diabetes mellitus type 2 (DM2)[1,2]. Fat may accumulate in two distinct compartments: (a) the endocrine and exocrine parenchymal tissue in the intra-lobular space, and (b) the adipose tissue in the inter-lobular space. Frequently, large amount of bulk fat can infiltrate the interlobular space, resulting in an irregular inter-digitated appearance of the pancreas. While multiecho gradient-echo MRI with proton-density fat fraction (PDFF) reconstruction may allow quantification of pancreatic fat on a pixel-to-pixel basis [3,4], a region-of-interest (ROI) placed in the pancreas may include heterogeneous population of pixels of the lower-fat intra-lobular tissue, higher-fat inter-lobular tissue, and admixture of the two due to partial-volume effects. We hypothesized that the presence of interlobular fat impacts repeatability of the PDFF measurement. The purpose of the study was to evaluate the effect of interlobular fat on the repeatability of pancreatic PDFF measurement by multiecho gradient-echo MRI.In this IRB-approved, HIPAA-compliant prospective observational study, 21 adult subjects with uncontrolled DM2 underwent 28 quantitative MRI examinations of the pancreas using a multiecho gradient-echo sequence with PDFF map reconstruction (mDixon-Quant) on a Philips Achieva 3T system (Philips Healthcare, Best, the Netherlands) using a Torso XL 16-channel phased array coil and SENSE factor of 2. One subject with large body habitus precluding use of the phased array coil was imaged with the built-in body coil without SENSE. All subjects signed informed consent prior to imaging. Two sets of axial images were acquired in two separate breath-holds, at 3- and 4-mm thick sections, but using otherwise identical imaging parameters, including 1x1 mm in-plane resolution. A 20x10-mm manually drawn rectangular ROI was manually drawn within the pancreatic body and co-localized on the 3- and 4-mm PDFF maps (Fig 1). Attempts were made to avoid splenic artery/vein, main pancreatic duct and obvious areas of inter-lobular fat. On the same section, the outer contour of the pancreas was manually segmented (Fig 1), including inter-lobular fat when present. The pixel PDFF values within the rectangular and segmented ROIs were exported into MATLAB (MathWorks, Natick, MA, USA), and the mean PDFF values were calculated, before and after exclusion of the high-fat voxels having PDFF ≥ 50%, thought to be contaminated by inter-lobular fat.The agreement of PDFF measurement between the 3- and 4-mm acquisitions was assessed by Bland-Altman analysis and intraclass correlation coefficient (ICC), shown in Fig 2. Mean FF values in the rectangular ROI (avg. 112.3 pixels) showed large variability exceeding ±10% PDFF and only moderate agreement by ICC. Exclusion of high-fat pixels had little effect on agreement. This suggests that PDFF in rectangular ROI is not impacted by interlobular fat, but repeatability is limited due to the small size of ROI given the intrinsic PDFF estimation noise. The mean PDFF values of the larger segmented ROIs (avg. 1597.7 pixels) were less variable after exclusion of the high-fat pixels (approx. ±4% FF) with excellent agreement by ICC. This suggests that PDFF in segmented ROI is impacted by interlobular fat, but high repeatability can be achieved after exclusion of the high-fat pixels contaminated by interlobular fat and averaging across a larger segmented area.A segmented analysis including most of the visualized pancreas but excluding intra-lobular fat (i.e. voxels with proton-density fat fraction>50%) provides a repeatable method for assessing pancreatic steatosis.