Subjects
Thirty healthy volunteers were recruited for this retrospective study. The same cohort has been under investigation in a previous study, but with different purposes [24]. The inclusion criteria were (i) age between 20 and 40 years and (ii) BMI between 20 and 33 kg/m2. Exclusion criteria were (i) prevalent or history of metabolic disorders, neuromuscular diseases, spine or thigh trauma; (ii) body conditions related to disbalance and/or morphological asymmetry at the level of the hip (such as scoliosis, advanced hip arthrosis); and (iii) general MRI contraindications. No subject reported any major physical conditions limiting mobility, and all subjects were considered having a normally active lifestyle. All subjects were right-footed.
Written informed consent was obtained from all subjects enrolled in this study. The study protocol was in accordance with the Declaration of Helsinki and its later amendments and was approved by the local institutional review board. The time between data acquisition of the first and last subject of the study was 4.5 months, and the interval for study inclusion was from April 2019 to August 2019.
Magnetic resonance imaging
All subjects underwent 3-T MRI (Ingenia, Philips Healthcare, Best, The Netherlands; bore diameter 70 cm, maximum field of view 55 cm) in a supine position using the built-in-the-table posterior and an anterior coil (32 channels, 60 cm coverage in feet-head direction, multi-array surface coils).
The imaging protocol, a standard protocol for quantitative MRI of the hip and thigh region at our institution, comprised an axial six-echo three-dimensional spoiled gradient-echo sequence for chemical shift encoding-based water-fat separation at the bilateral thigh and hip region. Sequence parameters were set as follows: repetition time 6.4 ms, echo time 1.1 ms, Δ echo time 0.8 ms, field of view 220 × 401 × 252 mm3, acquisition matrix 68 × 150, voxel size 3.2 × 2.0 × 4.0 mm3, frequency encoding direction left-to-right, no parallel imaging, and scan time 1:25 min:s per stack. Images were acquired in two stacks to cover the volume of the upper endplate of L4 down to the mid-thigh region. The six echoes were acquired in a single repetition time using non-flyback (bipolar) read-out gradients. A flip angle of 3° was used to minimise T1 bias effects [25, 26].
Muscle and femur compartment segmentation and PDFF extraction
The gradient-echo imaging data were processed online using the fat quantification routine of the MRI vendor (Philips Healthcare, Best, The Netherlands). PDFF maps were generated using a complex-based water-fat separation algorithm that accounts for known confounding factors including a single T2* correction, phase error correction, and consideration of the spectral complexity of lipids using the multi-peak fat spectrum model of Ren et al. [27]. Segmentation was performed by a medical doctor (F.Z.), supervised by a radiologist (T.B.) with 9 years of experience, using the free open-source software Medical Imaging Interaction Toolkit (MITK; developed by the Division of Medical and Biological Informatics, German Cancer Research Center, Heidelberg, Germany; www.mitk.org; Fig. 1).
The gluteus and quadriceps femoris muscle groups as well as three subregions of the femur (head, neck, and greater trochanter) were manually segmented on both sides in the PDFF maps (Figs. 1 and 2). No supportive semiautomatic or automatic segmentation techniques (such as thresholding or region growing) were used. Segmentation of the quadriceps muscle included all four compartments (rectus femoris, vastus lateralis, vastus medialis, and vastus intermedius) and was performed in five consecutive axial slices, starting ten axial slices caudal of the small trochanter. Segmentation of the gluteal muscle included all three compartments (gluteus maximus, gluteus medius, and gluteus minimus) and was performed in five consecutive axial slices, with the centre slice being located at the level of the maximum diameter of the piriformis muscle. Segmentation of the femur subregions was performed with sufficient distance to the cortical bone in order to avoid the inclusion of tissue other than bone marrow. The PDFF [%] was extracted and averaged over both sides, which was achieved separately for the muscle (PDFFgluteus and PDFFquadriceps) and femoral bone marrow compartments (PDFFfemoral head, PDFFfemoral neck, and PDFFfemoral greater trochanter).
Reproducibility of PDFF measurements
Four randomly selected subjects (two males and two females) from the study population were used to determine the inter-reader reproducibility of the PDFF measurements. All muscle and bone marrow segmentations as described above were performed independently in those four subjects by a second reader (L.G., 4 years of experience in image segmentation and analysis).
Statistical analysis
Statistical analyses were performed with SPSS (SPSS Inc., Chicago, IL, USA). All statistical tests were conducted using a two-sided level of significance equal to α = 0.05.
The Kolmogorov-Smirnov test was used to test for normal distribution of the measured parameters. Mean and standard deviation (SD) were calculated for age and PDFF of the three bone marrow and two muscle compartments (parametric data distribution for age and PDFF). Median and interquartile range (IQR) were calculated for BMI (non-parametric data distribution for BMI). Values were compared between males and females using Mann-Whitney U tests or unpaired t tests, depending on non-parametric versus parametric data distribution. Correlations of PDFF, age, and BMI were analysed using Spearman’s Rho. Furthermore, partial correlation analyses between PDFF values of different compartments were performed adjusting for BMI and age.
The inter-reader reproducibility error for the four subjects analysed by the two readers was expressed as the root mean square of the absolute precision error (absolute units) and root mean square of the relative precision error (expressed as coefficient of variation, relative units) according to Gluer et al. [28].