Diffusion-weighted imaging (DWI) has been established as a central part of oncological MR imaging over the past decade (1). In clinical practice, mostly EPI-based sequences are used, mainly due to short acquisition times and insensitivity to motion-induced phase errors. However, a disadvantage of EPI-based DWI is the relatively high sensitivity to B0-field variations resulting in spatial distortion and off-resonance effects. Especially at higher field strengths (3 Tesla) and heterogenious anatomical regions (neck region, tissue-air surfaces), these artifacts lead to very poor image quality. Several approaches have been suggested in order to reduce such artifacts including readout-segmented EPI ², zoomed EPI ³ and recently EPI with integrated 2D shim and frequency adjustment EPI (iShim) ⁴.

The aim of this study was to compare the performance of these available techniques in diffusion-weighed MR imaging of the neck at 3 Tesla.

A cylindrical DWI phantom was equipped with a Maxwell coil allowing for the application of an external magnetic field gradient of different strengths (Figure 1, A). This phantom design was used to produce B0-field-heterogenieties simulating B0-field variations that are typically observed in the anatomical region of the neck. The effect of increasing B0-field variations on image quality was observed using a standard EPI-based DWI sequence, a zoomed EPI sequence, a readout-segmented EPI sequence as well as an EPI sequence with integrated slice-per-slice shim and update of center frequency (“iShim”) on a 3 Tesla system (Magnetom Skyra, Siemens Healthcare, Erlangen, Germany). The extent of spatial image distortions and off-resonance artifacts was quantified and compared for different strengths of the applied field gradient.

In addition, the different EPI DWI sequences were applied on 10 volunteers (head/neck region). Image quality of these measurements was assessed visually.

With increasing strength of the superimposed external gradient field spatial distortions clearly increased in all sequence techniques. The strongest spatial distortions were observed in the conventional EPI sequence and the zoomed EPI Sequence while spatial distortions were moderate in the readout-segmented EPI sequence and only minor using the iShim EPI Sequence. At higher B0-field variations, complete signal loss, ghosting and failed fat suppression were observed in all techniques but the iShim sequence. Phantom results are shown in Figure 2.

Similar results were observed in the volunteer study. Failed fat suppression was observed in the lateral shoulder areas in all sequences. Strong ghosting artifacts were observed in the conventional EPI-sequence and the zoomed EPI sequence in the lower neck regions. In all sequences but the iShim sequence, signal losses of different extents were observed in the middle and upper neck regions. These results are illustrated in Figure 3.

The results observed in the phantom study as well as the volunteer study revealed the best image quality when using the iShim EPI sequence. The readout-segmented EPI sequence showed an intermediate image quality while the lowest image quality was observed using the conventional and the zoomed EPI sequences. The slice-per-slice update of B0-field maps and frequence adjustment of the iShim sequence evidently leads to an efficient correction of artifacts caused by B0-field variations.These results are in concordance with a previous study using DWI for breast imaging ⁴. Whether these results can be reproduced in a clinical setting has to be further evaluated in clinical studies with appropriate patient populations.Our results suggest that EPI with slice-dependent shim and frequency adjustment yields the best image quality for diffusion-weighted imageing of the neck at 3 Tesla. Further studies in a clinical setting are necessary to evaluate the diagnostic impact of this improvement in image quality.