Agreement between manual relaxometry and semi-automated scanner-based multi-echo Dixon technique for measuring liver T2* in a pediatric and young adult population

Commercially available 3D multi-echo Dixon (mDixon) sequences provide parametric maps of liver T2*, obviating manual curve fitting that is often required with conventional gradient recalled echo (GRE)-based multi-echo relaxometry, potentially simplifying clinical work flow. The purpose of our study was to compare T2* values generated by a 3D mDixon sequence to values generated by GRE-based T2* relaxometry with manual curve fitting in a pediatric and young adult population. We reviewed clinical MRI exams performed at 1.5T for liver iron content estimation between February 2015 and June 2016 that included both mDixon and multi-echo GRE pulse sequences. We obtained mean T2* measurements based on each sequence by drawing regions of interest on each of four axial slices through the mid-liver. We compared mDixon-based and GRE-based T2* measurements using paired t-tests and assessed agreement using single-measure intra-class correlation coefficients and Bland-Altman difference plots. One hundred nine patients met inclusion criteria (site 1=82; site 2=27). Mean age was 12.4±5.8 years, and 42 subjects (39%) were female. There was no statistically significant difference in mean T2* values for the two sequences (pooled means: 11.7±11.0 [GRE] vs. 11.7±10.9 ms [mDixon]; P=0.93). There was excellent absolute agreement between sequences (intraclass correlation coefficient [ICC]=0.98 for patients at both sites, confidence interval [CI]: 0.97-0.98 with mean bias of 0.0 ms [-4.2 ms to +4.2 ms]). 3D mDixon is accurate for measuring liver T2* and can likely replace 2D GRE-based relaxometry. (orig.)