Diffusion weighted imaging (DWI) can describe the microstructure of nerve fibers [1,2] and is therefore a very powerful tool in the study of neuropathic changes [3]. The lumbar plexus is a challenging region due to the need for high isotropic resolution to depict the oblique geometry of the lumbosacral nerve roots. T2-weighted imaging with 3D turbo spin echo (TSE) has been shown to meet these resolution requirements [4] but diffusion imaging is challenged by motion and field inhomogeneity present in this region. Diffusion-prepared sequences have been recently applied in a range of organs and tissues in order to achieve high-resolution diffusion-weighted imaging (DWI) or diffusion tensor imaging (DTI), free of the sensitivity to off-resonance effects associated with the use of single-shot EPI [5-7]. Diffusion preparation is usually combined with velocity compensation in order to reduce the sensitivity of the preparation to motion effects [7]. Effects from eddy currents also need to be considered when designing a diffusion preparation [8]. Another important consideration when applying diffusion in the body is the sensitivity of the preparation module to transmit B1 inhomogeneity effects. Recent works focusing on non-diffusion weighted T2 preparation have proposed the use of MLEV or hyperbolic secant pulses in order reduce the sensitivity of the preparation to transmit B1 effects, but without completely removing them [9,10]. An adiabatic T2 preparation using a BIR-4 pulse has been shown to be robust to both B1 and B0 effects when applied for T2 preparation [10,11]. The present work proposes a BIR-4-based diffusion preparation module in combination with a 3D TSE readout to achieve high-resolution DWI of the lumbar plexus.

Sequence development: Two alternative diffusion preparation schemes were used: one employing the adiabatic BIR-4 RF pulse (Fig. 1a) and one employing double-refocused hyperbolic secant pulses (DRHS) (Fig. 1b). Diffusion gradients were added around the refocusing pulses of the preparation in order to achieve velocity compensation (first gradient moment nulling). The diffusion preparation was combined with a 3D TSE readout. A dephasing gradient was placed before signal restoration and later balanced in the 3D TSE readout to reduce the effects of phase from eddy currents and motion.

In vivo measurements: DWI of the lumbar plexus of two healthy volunteers was conducted using a 16-channel torso coil and the posterior coil on a 3T Philips system with the developed sequence. Coronal acquisitions were performed with the sequence parameters: FOV=380×380×80$$$\text{mm}^{3}$$$, acquisition voxel=2×2×2$$$\text{mm}^{3}$$$, reconstruction voxel=0.7×0.7×1$$$\text{mm}^{3}$$$, TR/TE=1500/24ms, TSEfactor=60, averages=2, duration=15m50s. DWI: b=0 and b=400 in six directions were acquired with TEprep=58ms. For one subject, the DWI acquisition was performed with both the DRHS and the BIR-4 diffusion preparation.

Post-processing: Iso-diffusion-weighted images (iso-DWIs) were generated from the b=600 data.

In comparison to DRHS (Fig.1b), the BIR-4 implementation yielded higher signal in the acquisitions without diffusion in regions with high B1 inhomogeneity (Fig.2). In those regions, DWI with DRHS showed severe signal loss in the individual DWI images (Fig.3) that resulted in the loss of small features especially in the upper region of the coronally acquired images. In the same regions, DWI with BIR-4 showed betterimage quality. High-resolution iso-DWIs (Fig.4) showed high signal of small nerve roots and nerve branches with an isotropic resolution of 2 mm.Diffusion imaging in the presence of motion and eddy currents without careful consideration in the design of the acquisition method faces many challenges. Flow compensation is necessary to minimize phase accumulation from first-order motion that leads to signal loss. Eddy currents inevitably occur in the implementation of diffusion gradients and can result large spatially varying signal losses. Even when the effects of motion and of eddy currents are minimized, sensitivity to transmit field inhomogeneity results in the inaccurate restoration of diffusion-prepared signal and consequently in signal loss. DWI of the lumbar plexus is susceptible to motion and to B1 inhomogeneity particularly in the higher regions of the FOV in the vicinity of the abdominal organs. In vivo results demonstrate that the proposed method is robust to motion and eddy current effects. In vivo results also show that the adiabatic BIR-4 pulse design improves the performance of the diffusion preparation in regions with high B1 inhomogeneity in the cases with and without diffusion. In the presence of diffusion, where further signal losses are expected, the complete loss of small nerve structures results in regions with high B1 inhomogeneity when the diffusion preparation uses hyperbolic secant pulses. Therefore, the present work presents a diffusion preparation that can achieve high isotropic-resolution DWI of the lumbar plexus in the presence of motion, eddy currents and transmit B1 effects.