Subjects
The study protocol was approved by the local ethics commission (registration number: 408/15S) and was carried out in accordance with the Declaration of Helsinki. Written informed consent was obtained from all subjects prior to imaging.
A total of 15 healthy subjects (10 male and 5 female volunteers, 28.5 ± 3.8 years) without any history of neurological or lumbar spinal diseases underwent MRN on a 3 T whole-body magnetic resonance scanner (Ingenia, Philips Healthcare, Best, The Netherlands). The MRN was performed three times in total in each volunteer on the same day with a short break between the single scans, including repositioning of the subject.
Magnetic Resonance Imaging
The LSP of the subjects was scanned with a 16-channel torso coil array and the built-in-table posterior 12-channel coil array. First, a flow-suppressed (2 mm isotropic voxel size) T2-weighted 3D TSE sequence was performed to depict plexus anatomy [4]. Second, an adiabatic T2-prepared 3D TSE sequence with variable duration of the T2 preparation was applied for T2 mapping [8]. A modified B1-insensitive rotation (BIR-4) pulse was used for the T2 preparation to minimize the sensitivity to B0 and B1 inhomogeneities (supplementary material) [8]. The parameters for the T2 mapping sequence were as follows: field of view (FOV) 38 × 38 × 8 cm3, acquisition voxel 2 × 2 × 2 mm3, echo train length 80, T2 preparation durations of 20/40/60/80 ms, fat suppression spectral attenuated inversion recovery (SPAIR) and a scan duration of 6:48 min with a repetition time (TR) of 1.6 s and an effective echo time (TE) of the TSE shot of 15 ms. The flip angle train of the TSE readout was designed in order to achieve a constant signal plateau throughout 80% of the TSE shot duration for white matter [9].
Post-Processing and Data Analysis
The T2 values were calculated offline on a voxel-by-voxel basis with a combination of variable projection (VAPRO) and golden section search [10, 11]. The resulting T2 maps were then analyzed using Horos (version 1.1.7; https://www.horosproject.org). In short, the S1, L5, and L4 nerves of both sides were identified, and a standardized color scheme was applied to the T2 maps (Fig. 1). The 3D imaging data were partly averaged (isotropic voxel size of 3 mm) and then manually reformatted to display a maximum length of the nerve course from the spinal cord to the periphery within the FOV (Fig. 2). The reformatting was done separately for S1, L5, and L4 nerves, and T2 values were measured by manually placing regions of interest (ROIs) in axial slices of the T2 maps (isotropic voxel size of 3 mm; Fig. 2). In each subject, preganglionic (~1 cm before the ganglion), ganglionic (in the middle of the ganglion), and postganglionic (~1 cm after the ganglion) T2 values were measured for each of these nerves on both sides, with uncolored anatomical images serving as reference to support precise ROI positioning based on anatomical identification of these structures (Figs. 1 and 2). To place a ROI, the respective nerve and segment were visually identified and manually surrounded using the ROI generation tool. The software then automatically displayed the mean T2 value of the area enclosed by the ROI.
The described approach of ROI positioning and measurement of T2 values was performed independently by two investigators. The first investigator (MD with experience in neuroradiological imaging since 2012: observer 1) evaluated all three scans of the subjects to assess reproducibility, whereas the second investigator (novice with experience in neuroradiological imaging since 2016: observer 2) evaluated the first scan of each subject to explore interobserver reliability. The two observers were strictly blinded to the measurements of the other, and the first observer was furthermore blinded to the T2 values of the first or first and second scan when performing measurements in the second or third scan of a respective subject. Observer 1 evaluated the first scan in each subject first, followed by a second and third round of analyses using the subjects’ second and third scans. The interval between the three analysis sessions was at least 2 weeks. Mean T2 values >100 ms were removed from the datasets prior to statistical analyses because high T2 values most likely stem from measurements in areas with high fractions of cerebrospinal fluid (CSF) or surrounding vessels.
Statistics
For statistical data analyses, GraphPad Prism (version 6.0, GraphPad Software, La Jolla, CA) and SPSS (version 23.0, IBM, Chicago, IL) were used. The level of significance was set at p < 0.05 for all statistical testing.
The T2 measurements for S1, L5, and L4 of both sides extracted from the ROIs were collapsed by calculating the average of the corresponding right-sided and left-sided values (left and right S1 T2 values together, left and right L5 T2 values together, and left and right L4 T2 values together) for preganglionic, ganglionic, and postganglionic measurements, respectively. This was done separately for the values of observers 1 and 2, and descriptive statistics were calculated based on the collapsed values obtained. During this step, the portion of removed measurements due to T2 values >100 ms was documented.
Based on the mean T2 values (right-sided and left-sided T2 measurements collapsed), intraclass correlation coefficients (ICCs) were calculated to assess reproducibility and interobserver reliability (two-way mixed model, unadjusted) for the different nerves and their segments. For reproducibility, three scans in each subject, all evaluated by observer 1, were considered, whereas the first scan of each subject was evaluated independently by both observers to evaluate interobserver reliability. Furthermore, coefficients of variation (CVs) were calculated as measures for dispersion regarding the T2 values. The mean T2 values (right-sided and left-sided T2 measurements collapsed) for preganglionic, ganglionic, and postganglionic measurements of the three scans in each subject were averaged to generate one T2 value for each of these segments for S1, L5, and L4, respectively. These segment-specific values were compared by Kruskal-Wallis test followed by Dunn’s multiple comparisons test for S1, L5, and L4.