Short- and midterm reproducibility of marrow fat measurements using mDixon imaging in healthy postmenopausal women.

Abstract

We tested the short- and midterm reproducibility of vertebral marrow fat fraction (FF) measurements using mDixon imaging. Thirty postmenopausal women underwent mDixon scans to obtain L1-4 FF from three slices per vertebra by two independent observers (session 1). Measurements were repeated after 6weeks (session 2) and 6months (session 3). The mean FF for three regions of interest per vertebra was calculated. The coefficients of variation (CVs) were calculated for each participant and imaging session, and the intraclass correlation coefficients (ICCs) were calculated to assess interobserver and intersession agreements. There were no significant differences in FF measurements among the three slices, imaging sessions or observers. The mean intrasubject CV for FF measurement reproducibility was 1.94%. The interobserver agreement for the average FF value was excellent (ICC ≥0.945 for each session). The ICC for intersession agreement was excellent (ICC ≥0.955 between sessions). The mean intersession CV was lower within a short-term interval (2.97%) than within sessions 1 and 3 (4.80%) or sessions 3 and 2 (4.44%). The overall mean CV for the reproducibility of FF measured with mDixon imaging over the short- and midterm was 4.09% (95% CI, 3.79-4.40%). mDixon is a reproducible method for FF quantification over short- and midterm intervals up to 6months in healthy postmenopausal women. Our results also provide data by which a power analysis can be optimized when designing studies involving the use of FF derived from similar mDixon sequences.