MR neurography is arising as a reliable method in the diagnosis of nerve disease [1]. Quantitative MRN has been shown to be more precise than conventional T2 imaging for detecting nerve injuries [2]. Diffusion tensor imaging (DTI) can provide precise information about the microstructure of nerve fibers [3,4] and has been shown to have the capability to detect disease in human nerves [5]. In high-resolution imaging, conventional 2D single-shot echo planar imaging (SS-EPI) becomes highly susceptible to distortion artifacts from eddy currents and off-resonance effects [6]. Among the existing alternatives to SS-EPI, diffusion-prepared multi-shot turbo spin echo (TSE) can minimize geometric distortions and chemical shift artifacts [7]. Diffusion-prepared multi-shot TSE alone is highly sensitive to phase errors that originate from eddy currents and motion [8]. Phase errors induce a modulation of magnitude after the tip-up pulse [9,10,11]. Therefore, phase errors need to be considered if using multi-shot TSE for quantitative diffusion imaging. The present work introduces a flow-compensated diffusion-prepared 3D TSE sequence and demonstrates high-resolution DTI of the nerves in the lower extremity.

Flow-compensated diffusion-prepared sequence: A sequence combining a flow-compensated diffusion-preparation with a 3D TSE readout was developed by adding dephasing and rephasing gradients during the preparation and in the TSE readout to reduce sensitivity to phase errors [9,10,11] (Fig.1).

SNR efficiency optimization: Refocusing angle schemes and corresponding signal of nerve (T1=1000,T2=75ms) were simulated using extended phase graphs (EPG) for 3D TSE for variable echo train length and TR. T1 and T2 relaxation effects were included during the refocusing pulse train and T1 relaxation was considered between shots. SNR and SNR efficiency values were computed as: $$$\text{SNR}=s_{0}\,\Delta\text{x}\Delta\text{y}\Delta\text{z}\sqrt{\Delta t\,N_{x}N_{y}N_{z}}$$$ and $$$\text{SNR}_{\text{eff}}=\text{SNR}\sqrt{\text{t}_\text{scan}}$$$, where s₀ is the signal at k=0, $$$\Delta\text{x}$$$, $$$\Delta\text{y}$$$ and $$$\Delta\text{z}$$$ are the acquisition voxel sizes, N_x, N_y and N_z are the matrix dimensions, $$$\Delta\text{t}$$$ is the sampling interval and t_scan is the scan time [12].

In vivo measurements: DTI of the knee of two healthy volunteers was conducted using a 16-channel knee coil on a 3T Philips system with the developed sequence. 1) A sagittal acquisition was performed on one subject and 2) an axial acquisition on the second subject. Readout parameters: 1) FOV=160×127×100$$$\text{mm}^{3}$$$, acquisition voxel=1.7×1.7×1.7$$$\text{mm}^{3}$$$, reconstruction voxel=0.5×0.5×0.85$$$\text{mm}^{3}$$$, TR/TE=1700/19ms, TSEfactor=60, averages=2, duration=15m33s. 2) FOV=140×140×52$$$\text{mm}^{3}$$$, acquisition voxel=0.7×0.7×8 $$$\text{mm}^{3}$$$, reconstruction voxel=0.49×0.49×0.4 $$$\text{mm}^{3}$$$, TR/TE=1650/21ms, TSEfactor=40, averages=2, duration=5m28s. DTI: b=0 and b=600 in six directions were acquired with TEprep=60ms. An axial T2-weighted mDIXON TSE scan with an acquisition voxel=0.3×0.38×4$$$\text{mm}^{3}$$$ was acquired as an anatomical reference.

Post-processing: A coronal reformat was obtained from the sagittal dataset. Iso-diffusion-weighted images (iso-DWIs) were generated. Diffusion tensors were computed using linear fitting and DTI parameters were obtained from the derived eigenvalues. Mean values of mean diffusivity (MD) and fractional anisotropy (FA) were measured. Projections of the primary eigenvectors were generated for visualization purposes.

Optimization of SNR_eff yielded a TSE factor of 40 and a TR of 1650ms for the axial protocol (Fig.2) and a TSE factor of 60 and TR of 1700ms for the sagittal protocol (not shown). Iso-DWIs show sagittal and coronal views of the long axis of the tibial nerve (Fig.3, left). Mean values for the tibial nerve in a region above the knee were found to be MD=$$$(1.1\pm0.1)\times10^{-9}$$$,FA=$$$0.57\pm0.06$$$. MD and FA maps of the nerve show little spatial variation along the course of the nerve (Fig.3, center and right). The primary eigenvector projections in Fig.4 follow the course of the nerve along its long axis. Axial iso-DWIs and MD maps with sub-millimeter in-plane resolution (Fig.5) show the sub-fascicular structure within the tibial and peroneal nerves.Simulation results show that a compromise between SNR and scan time can be found with considerations of relaxation and refocusing angle modulation in 3D TSE. SNR efficiency becomes important in quantitative scans with intrinsically low SNR and long acquisition times, since high noise levels can result in the incorrect estimation of quantitative values and long scan times can exceed the clinical limits. In vivo results demonstrate that the developed method can generate high-SNR, high-resolution iso-DWIs in the extremities free of eddy-current, off-resonance and chemical-shift artifacts in acceptable scanning times. Measured MD and FA and eigenvectors for the tibial nerve are in agreement with reported values [12] and therefore indicate that the proposed method minimizes phase errors. Axial acquisitions with sub-millimeter in-plane resolution show artifact-free iso-DWIs and MD maps with resolved nerve fascicles. The present work therefore introduces a multi-shot method that is capable of accurate, high-resolution quantitative diffusion imaging of peripheral nerves and that is robust to many of the problems commonly encountered in other DWI sequences.