Comparison of whole-body and abdominal MRI for the estimation of renal sinus fat in healthy volunteers and people with type 1 and type 2 diabetes

Motivation: Whole-body MRI performed in large cohorts with interslice-gaps is not ideal to determine small organ-substructures such as the renal sinus fat (RSF). Thus, the accuracy of such estimates remains unclear. Goal(s): To compare the estimates of RSF obtained by automatic segmentation of whole-body TSE images with interslice-gaps and continuous abdominal dual-echo GRE images. Approach: Test the performance of U-Net models for RSF segmentation in both the external and internal sets and compare outcomes. Results: U-Net models showed high accuracy in the detection of RSF volume. The estimates of RSF from whole-body TSE images were ~30% lower as compared to abdominal dual-echo GRE images. Impact: Whole-body MRI with interslice-gaps can be used to accurately estimate renal sinus fat in people with type 2 diabetes, however, corrections of values based on more detailed, high-resolution MRI measurements are necessary.