Magnetic resonance neurography has been traditionally based on T2-weighted imaging for depicting nerve anatomy, diagnosing and localizing nerve lesions and tracking neuropathic changes [1]. Those changes usually appear as a hyper-intense signal on T2-weighted images, but the assessment of signal changes based on qualitative information remains challenging [2]. In contrast to that, quantitative T2 mapping provides a more precise method for monitoring nerve disease progression [3]. However, due to the complexity of nerve anatomy in most regions, T2 mapping of nerves has focused on axial 2D sequences in the extremities [4]. The lumbar plexus remains a challenging area due to the need for high isotropic resolution to depict the complex oblique geometry of the lumbosacral nerve roots. It has been recently shown that 3D turbo spin echo (TSE) imaging can achieve this high isotropic resolution and even resolve small nerve branches [5-7]. Additionally, the T2 quantification can be significantly affected by B0 and B1 inhomogeneities that are common in this region. Therefore, the purpose of the present work is to combine 3D TSE imaging with an adiabatic T2 preparation to achieve B1-insensitive high-resolution isotropic T2 mapping of the lumbar plexus.

Pulse sequence: A sequence was developed composed of an adiabatic T2 preparation, a spoiler gradient and a 3D TSE readout for imaging (Fig. 1). The T2 preparation consists of a modified BIR-4 RF pulse, where two gaps were introduced to achieve a module with variable TE. The T2Prep module was performed with increasing gaps to acquire images with increasing T2-weighting.

Simulations: Bloch simulations (Fig. 2a) were performed to investigate the dependency of the presented method on B1 and B0 variations using the following parameters: BIR-4 pulse duration: 10 ms, B1: 13.5 µT, frequency sweep: 3.7 kHz, T2 preparation duration: 20/40/60/80 ms, T1/T2: 1400/70ms.

In vivo measurements: The lumbar plexus of five healthy volunteers (mean age: 27.2) was scanned on a 3 T system (Ingenia, Philips Healthcare) with a 16-channel torso coil and the built-in-table posterior coil. First, a flow-suppressed (1.7×1.7 mm² in-plane) T2-weighted 3D TSE sequence was performed to depict plexus anatomy [5]. B1 and B0 maps were then acquired to quantify B1 and B0 variations over the FOV, using a dual-TR sequence [8] and a two-point Dixon sequence, respectively. Finally, the developed sequence for T2 mapping was performed with T2 preparation durations of 20/40/60/80 ms and with sequence parameters: FOV = 38×38×8 cm³, acquisition voxel = 1.7×1.7×1.7 mm³, TR/TE = 1.6 s/21 ms, TSE factor = 80, total scan duration = 9m48s.

Postprocessing and data analysis: T2 maps were calculated with an exponential two-parameter fit on a voxel-by-voxel basis. Masks of the S1 and L5 spinal nerves were generated from the T2 maps by means of manual segmentation and superimposed on the corresponding anatomical images.

Fig. 2a shows the dependence of the extracted T2 parameter on B0 and B1 offsets. Fig. 2b shows the histogram of B0 and B1 offsets in a typical subject in vivo. Based on the observed range of B0 and B1 in vivo, the employed T2 preparation should be considered minimally sensitive to B0 and B1 offsets. Fig. 3 shows a MIP of the anatomical 3D TSE. T2-weighted images (T2Prep duration: 20/60 ms) and the corresponding T2 maps are shown for the S1 and L5 spinal nerve (Fig. 3). In the superimposed T2 maps of the L5 and S1 some spatial variation of the T2 value along the nerve is visible. This observation was confirmed with the results of the volunteer study shown in Fig. 4. Here, the highest T2 value was observed at the dorsal root ganglion (DRG), followed by proximal to DRG and distal to DRG (p < 0.01). The T2 values from the S1 were also statistically significantly higher than values from the L5 in all three investigated locations (p < 0.045). The combination of 3D TSE imaging and adiabatic T2 preparation enables B1-insensitive and isotropic high-resolution T2 mapping of the lumbar plexus. The present T2 preparation did not use additional flow suppression gradients and still suffers from venous signal contamination in some regions. However, the present T2 preparation module could be enhanced with gradients to suppress venous signal [5-7]. In addition, the present volunteer study showed that there is a spatial variation of the T2 along the nerve route as well as variations in the mean T2 value between different nerves. We showed that T2 mapping of the lumbar plexus is feasible using a T2-prepared 3D TSE and might be useful for monitoring nerve disease progression in the lumbar spine.