Reproducibility of automatic adipose tissue segmentation using PDFF images between 1.5T and 3.0T MR

Motivation: There is a growing need for standardized and reliable methods for adipose tissue quantification in clinical practice. Goal(s): To evaluate the reproducibility of the deep-learning based adipose tissue distribution analysis method across different MR filed strengths, and so establish a reproducible and objective method for adipose tissue quantification. Approach: The whole-body PDFF images of 24 volunteers were acquired at both 1.5 T and 3.0 T scanners. The volumes and PDFF values of the segmented adipose tissue of whole body and subparts were compared. Results: The results demonstrated good reproducibility of volume and PDFF values between 1.5T and 3.0T scanners with the p-value>0.05. Impact: These findings could improve adipose tissue assessment in diverse clinical MR settings. This enhancement would enable large cohort studies to better identify obesity-related health risks using multicenter datasets, thus facilitating a more effective approach to obesity management.