Agreement and Reproducibility of Proton Density Fat Fraction Measurements Using Commercial MR Sequences Across Different Platforms: A Multivendor, Multi-Institutional Phantom Experiment.

OBJECTIVES

The aim of this study was to evaluate the agreement and reproducibility of proton density fat fraction (PDFF) measurements using commercial magnetic resonance (MR) sequences across different imagers, vendors, and field strengths via a phantom experiment.

MATERIALS AND METHODS

Eleven fat-water emulsion phantoms of varying fat proportions (ie, 0-50 weight%) were constructed. Phantom PDFFs were estimated using commercial chemical shift-based MR imaging sequences with Siemens 1.5 T and 3.0 T, Philips 3.0 T, and GE 1.5 T and 3.0 T imagers, and MR spectroscopic sequences (HISTO) with Siemens 1.5 T and 3.0 T imagers. Agreement among the estimated PDFF values between commercial sequences was evaluated using Bland-Altman analysis. Reproducibility of the PDFF measurements across commercial sequences was evaluated using the reproducibility coefficient. The test-retest repeatability of the PDFF measurements was evaluated using the repeatability coefficient.

RESULTS

The repeatability coefficient of the PDFF measurements was 0.31% to 1.58% for the absolute PDFF value for commercial sequences. Statistically significant biases in the estimated PDFF were noted in 19 of 21 pairwise comparisons of commercial sequences (range of mean biases, -4.48% to 8.15% for the absolute PDFF value). The reproducibility coefficient of PDFF measurements was 9.0% for the absolute PDFF value over all commercial sequences and 10.6% for the absolute PDFF value over all chemical shift-based MR imaging sequences.

CONCLUSIONS

The measurement of the PDFF is highly repeatable with commercial MR sequences but is not reproducible across different sequences, imager vendors, and field strengths. The use of the same sequence and imager is therefore recommended for the longitudinal follow-up of hepatic steatosis using commercial MR sequences for PDFF measurements.