Comparison of Volumetric Measurement Method with Region of Interest Drawing Method for Liver Fat Quantification

Purpose: This study aimed to evaluate the feasibility of measuring liver fat using the volumetric measurement method (Fat(vol)) by comparing it with the conventional 27-regions of interest drawing method (Fat(roi)). Materials and Methods: This retrospective study included 67 patients who underwent liver magnetic resonance imaging with fat quantification in August or September 2020. Two experienced abdominal radiologists measured the proton density fat fraction (PDFF) of the liver using the mDIXON-Quant sequence for each of two methods. The PDFF was measured twice with each method at intervals of at least 4 weeks to avoid recall bias. Measurement times were recorded. The intra-class correlation coefficient (ICC) was calculated for intra-exam repeatability, inter-reviewer reproducibility, and inter-exam agreement. Results: Measurement times for Fat(vol) were significantly shorter than for Fat(roi). Measurement times for Fat(roi) and Fat(vol), respectively, for reviewer A were 209.4 +/- 55.1 s and 137.2 +/- 51.5 s in session 1, and 180.9 +/- 37.3 s and 127.0 +/- 46.1 s in session 2. For reviewer B, the times were 190.7 +/- 30.1 s and 74.8 +/- 27.4 s in session 1, and 174.6 +/- 21.8 s and 64.1 +/- 17.5 s in session 2. In all cases, p < 0.001. The mean PDFF values were 7.2% +/- 6.4% and 7.2% +/- 6.5% (sessions 1 and 2, respectively) for Fat(roi) and 7.4% +/- 6.0% and 7.3% +/- 6.1% for Fat(vol) for reviewer A. For reviewer B, they were 7.1% +/- 6.6% and 7.1% +/- 6.6% for Fat(roi) and 7.4% +/- 5.8% and 7.4% +/- 5.8% for Fat(vol). The ICCs between measurement methods (0.998 and 0.995 for reviewers A and B, respectively), for Fat(vol) within each reviewer (0.999 and 1.000 in sessions 1 and 2, respectively), and between reviewers (0.999) were excellent. Conclusion: The measurement time could be significantly reduced using Fat(vol) compared to Fat(roi) while maintaining the consistency of the liver fat measurement values.