Measuring the relative and absolute lean and adipose tissue volumes by MRI based body composition analysis can be a valuable tool to understand the progression of the metabolic syndrome and muscle decline due to ageing. The strength of tomographic body composition analysis is the ability to directly asses compartmental tissue volumes. For instance, visceral adipose tissue is a strong predictor of many metabolic bio-markers, as opposed to subcutaneous adipose tissue [1].

One method for MRI body composition analysis relies on quantitative relative fat concentration (RFC) images, i.e. images were the intensity in-homogeneity is estimated and then removed based on white adipose tissue as an intensity reference [2]. The effect is quantitative RFC images were the voxel values correspond to the concentration of adipose tissue. Thus, adipose tissue volume within a compartment can be assessed by integrating its RFC values and scaling by the voxel volume.

For intramuscular fat it has been shown that a higher flip-angle increases the precision of RFC, without sacrificed accuracy [3]. The SNR gain of higher flip-angle can be beneficial for MRI body-composition analysis as it can be traded for lowered scan time and/or larger FOV.

The purpose of this study was to measure the reproducibility of adipose tissue volumetry by RFC images between two flip-angles. The study was limited to whole-body total adipose tissue (TAT), total lean tissue (TLT) and total soft tissue (TST), to maximize sensitivity towards systematic errors and to limit errors introduced by segmentation.

In-vivo imaging was performed as part of a study of the effects of standardized resistance training in postmenopausal women using a Philips Ingenia 3.0 T MR-scanner (Philips, Best, TheNetherlands). MR-acquisitions were performed with a 4-point Dixon protocol, covering whole-body. The protocol covered a total of 1.76 m, divided over ten overlapping slabs of axial 3D spoiled gradient multi-echo images. At each station flip-angle 5° and 10° slabs were acquired with common parameters; TR=5.8 ms, TE=1.15/2.30/3.45/4.60 ms and voxel size 2.5×2.5x4 mm. The first and last four slabs consisted of 66 slices; slabs 2-6 consisted of 39 slices that were acquired during 17-sec expiration breath-holds. Water-fat images were computed using an in-house IDEAL-type reconstruction. Each station was calibrated to form RFC and complimentary water images and then merged to whole-body volumes. The fat-fraction criteria used to locate reference voxels were adjusted for each flip-angle, by taking the mean fat-fraction of adipose tissue in the first 10 subjects. The regional ethics committee approved the study, and written informed consent was obtained from all subjects prior to study entry.

The whole-body TST volume was computed as the volume of a soft-tissue mask, TAT by integrating RFC values within the mask and TLT as the difference TST-TAT, as described in [4]. Descriptive statistics (mean ± SD) were calculated for all measurements. Reproducibility was calculated using Bland-Altman analysis (mean of difference, and limits of agreement) as well as coefficient of variation (CoV). To test for proportional errors between TAT volumes a regression analysis of TAT differensens against average TAT was performed. Statistical analysis was performed in SPSS 22 (IBM, 2013).

A total of 29 women were included and analyzed, see example images in figure 1. The average of adipose tissue fat-fractions of the first 10 women was 0.94 at flip-angle 10° and 0.92 at flip-angle 5°. The mean relative difference between flip-angles was 0.92 % for TAT, -1.46 % for TLT and -0.25 % for TST. Descriptive statistics and CoV can be found in table 1. The Bland-Altman plot for TAT can be seen in figure 2. The regression analysis showed that the TAT differences were not proportional to the mean volume, slope 0.005, p=0.121 and CI (-0.001, 0.012).

Small CoV for all three tissue types indicates good reproducibility properties of the RFC based body composition analysis. The systematic difference observed for TAT indicates that RFC has a small dependence on flip-angle that was not proportional to the volume. It is reasonable to believe that this difference will have a minimal effect on compartmental analysis, as the difference observed here is the difference accumulated over the whole-body. Based on the gain in precision at higher flip-angles observed by Peterson et al. [3] and the small difference in volume observed in this study a flip-angle of 10° is preferable to 5° for RFC body composition analysis, especially if targeting smaller compartments.