Comparison of simultaneous multi-slice readout-segmented EPI and conventional single-shot EPI for diffusion tensor imaging of the ulnar nerve

Purpose To compare conventional single-shot echo planar imaging (ss-EPI) and simultaneous multi-slice (SMS) readout-segmented EPI (rs-EPI) for magnetic resonance diffusion tensor imaging (DTI) of the ulnar nerve. Materials and methods This study was approved by the local ethics committee. Ten healthy volunteers (mean age 30.4 ± 4.01 years; range 25–36 years) underwent 3T DTI of the ulnar nerve at the level of the cubital tunnel. DTI was performed based on ss-EPI as well as SMS rs-EPI sequences. Signal-to-noise ratio (SNR), image quality, and DTI parameters in the ulnar nerve (fractional anisotropy, FA; mean diffusivity, MD) were compared between the two sequences by two independent radiologists. Results Acquisition time was 5:12 min for ss-EPI and 5:18 min for SMS rs-EPI. Between the two sequences, no significant differences were found for derived DTI measures FA (p = 0.11) and MD values (p = 0.93). Compared to conventional ss-EPI, SMS rs-EPI yielded significantly less ghosting artifacts (p = 0.04) but inferior nerve depiction (p = 0.001) and worse overall image quality (p = 0.008). Conclusion SMS rs-EPI is not advantageous over ss-EPI in DTI of the ulnar nerve at the level of the cubital tunnel.


Introduction
Ulnar neuropathy at the elbow (UNE) is amongst the most common peripheral neuropathies with an annual incidence of about 21/100.000 [1]. UNE is clinically and socioeconomically important since it can lead to continuing atrophy and persistent impairment of intrinsic musculature [2]. Traditional diagnostic testing for UNE mainly consists of physical examination, nerve conduction studies and ancillary neurosonography, all of which are highly operator dependent. Recently, T2-weighted MR neurography has added diagnostic value in precise localization of lesion of UNE [3,4]. Still, the etiology of ulnar neuropathy at the elbow remains unclear in many cases [1,5].
In search of additional tools for characterizing peripheral nerves, diffusion tensor imaging (DTI) has been increasingly used in the last years for imaging of forearm nerves [6,7]. DTI allows voxel-wise measurement of the intensity, main direction, and anisotropy of diffusion of water molecules in body tissues [8]. DTI parameters could have a role as a surrogate marker of peripheral nerve integrity [9,10]. Furthermore, DTI is projected to play a role in monitoring regeneration and degeneration in motor neuron disease in the future [11]. In particular, the median nerve has been extensively studied [12,13]. The ulnar nerve, in contrast, has only been subject of investigation in few studies [3,14,15].
Conventional DTI is technically based on single-shot echo planar imaging (ss-EPI), where the entire k-space is filled during a single T2* decay. Since ss-EPI is prone to motion artifacts, susceptibility artifacts and image blurring, DTI of peripheral nerves is not yet commonly used in clinical routine. Readout-segmented EPI (rs-EPI) is a modern technique, which may overcome the limitations of ss-EPI, because it achieves shorter echo-spacing by dividing the k-space into separately acquired segments [16]. The main disadvantage of rs-EPI is its longer acquisition time. However, the acquisition time could now be reduced by using the simultaneous multi-slice (SMS) technique with blipped 'Controlled Aliasing In Parallel Imaging Results In Higher Acceleration' (blipped CAIPIRINHA). This technique was initially developed for ss-EPI and recently introduced for rs-EPI [17,18].
In the present work, we tested the feasibility of DTI of the ulnar nerve using SMS rs-EPI. Our hypothesis was that SMS rs-EPI would yield significantly higher image quality compared to conventional ss-EPI at comparable acquisition time.

Study population
The ulnar nerve was scanned in ten healthy volunteers (5 males, 5 females; mean age 30.4; age range 25e36; side, 7 right, 3 left). No subject reported symptoms of peripheral neuropathy or previous cervical spine, upper arm or elbow surgery. Approval by the local ethics committee and written informed consent from all volunteers were obtained.

Imaging protocol
All subjects were scanned in a clinical 3T scanner (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany). They were positioned in the prone position with the arm extended overhead. A 4-channel flexible coil (Siemens Healthcare) was placed around the elbow (Fig. 1).
All sequences were planned in the axial plane at the level of the cubital tunnel. First, a T2-weighted fat-suppressed sequence was acquired in order to assess the anatomy and to exclude structural pathologies (TR, 5800 ms; TE, 40 ms; field-of-view, 272 Â  Next, DTI was performed using both a conventional ss-EPI sequence and an SMS rs-EPI sequence with two-fold slice acceleration (imaging parameters see Table 1).
Acquisition times of both diffusion sequences were almost equaled for reasons of comparison and were kept reasonably short for potential future application in clinical routine.

Assessment of image quality
Two independent readers (***initials blinded***; reader 1: clinical fellow with 5 years training in musculoskeletal radiology; reader 2: research fellow in musculoskeletal radiology with 2 years training in neuroradiology) rated on a 5-point-Likert-like scale (1, excellent; 2, good; 3, moderate; 4, poor; 5, non-diagnostic) the following qualitative imaging parameters: ghosting artifacts, anatomical distinction of the ulnar nerve and overall image quality.

Post-processing and measurements
All images were post-processed using the syngo.via platform (Version VB10A, Siemens Healthcare, Erlangen, Germany). For the DTI sequences, parametrical voxel-wise maps of the fractional anisotropy (FA) and mean diffusivity (MD) were automatically generated by the scanner's console ( Table 2). For FA and MD

SNR estimation
The SNR was estimated on the b0 images by dividing the signal of the ulnar nerve by the background noise (standard deviation of the signal) measured in an ROI at the edge of the field-of-view.

Image acquisition
All morphological and diffusion images were successfully acquired (for example FA and MD maps see Fig. 2). Specific absorption rate (SAR) limits were respected in all sequences without the need for switching to first-level mode. map for the ss-DTI at the same level, where anatomical depiction of the ulnar nerve was graded "good" and ghosting artefacts were rated "moderate" for both readers for both maps. Panel (C) and (E) demonstrates the corresponding rs-DTI maps. Here the MD map was graded "moderate" for anatomical depiction of the ulnar nerve by both readers and "moderate" for ghosting artifacts by reader 1 and "low" by reader 2. The FA map for rs-DTI was rated "poor" by both readers for anatomical depiction and "moderate" for ghosting artifacts for both readers.

Qualitative ratings
Anatomical depiction of the ulnar nerve and overall image quality were both perceived significantly better in the ss-EPI sequence (p ¼ 0.002 and 0.008, respectively

The role of DTI for imaging of peripheral nerves
Until now diagnostic imaging of peripheral neuropathies with MR neurography rests upon prolonged T2-relaxation times in fat saturated pulse sequences to localize a intraneural edema within the course of the fascicles. However, T2 hyperintense signal alteration is not specific to any defined pathophysiological condition since they merely echo the grade of density and compartmentalization of water-hydrogen protons [21]. Another problem linked to imaging of peripheral nerves is the magic angle effect due to an artificial increase in T2-relaxation time [22]. This phenomenon is caused by interaction of water-hydrogen protons with the surrounding collagen matrix of the endoneurium and interfascicular compartments when nerve fibers run at an angle of 55 relative to the main static magnetic field [23]. Especially at the elbow the magic angle effect can limit selectivity between physiological pathologic conditions. Husarik et al. found that the T2w-signal of the ulnar nerve at the elbow is increased in about 60% of individuals without any symptoms of ulnar neuropathy [24].
Due to these limitations of T1 and T2 weighted anatomical MR imaging, DTI is a promising new approach to deliver more accurate diagnostic information about the nerve [25]. DTI relies on Brownian motion of waterehydrogen protons along different gradient directions. Changes in tissues self-diffusivity lead to alterations in derived DTI parametric maps of fractional anisotropy (FA) and mean diffusivity (MD). In the peripheral nervous system it is assumed that FA reflects the microstructural nerve integrity and mean diffusivity (MD) measures the density of the axonal membrane [26]. In general, peripheral nerves with their strong directionality seem to be predestinated for DTI. However, recent studies (mainly using ss-EPI) have indicated the technique may be limited in small nerves due to low in-plane resolution, wide slice thickness and partial volume contamination effects. Thus, in search of alternative sequences for DTI of peripheral nerves, in the present study the feasibility of SMS rs-EPI for DTI of the ulnar nerve at the cubital tunnel was tested and compared with conventional ss-EPI. We assessed quantitative data by means of FA, MD and SNR and qualitative data by means of ghosting and motion artifacts, anatomical distinction and overall image quality in axial DWI images in 10 healthy volunteers.

Technical considerations
We found that SMS rs-EPI yielded similar FA and MD as well as similar ghosting artifacts compared to ss-EPI, but significantly worse anatomical depiction of the ulnar nerve and poorer overall image quality, likely related to the slightly lower SNR.
Still, obtained SNR values for the ulnar nerve in this study were in line with current literature for normal values in peripheral nerves [7,9]. Sufficient SNR is important 1) in order to clearly identify the relatively small ulnar nerve at the elbow; 2) to minimize uncertainty of derived MD and FA values; and 3) to minimize uncertainty of the voxel-wise eigenvector of the diffusion tensor [27,28].
At the same time, spatial resolution is a crucial factor for accurate DTI measurements. Within the cubital tunnel, the ulnar nerve normally is surrounded by fat, which poses an additional challenge to DTI because potential partial volume effects may significantly influence FA and MD measurements. To avoid this potential source of bias, we applied a relatively high spatial resolution, allowing us to draw ROIs small enough to avoid inclusion of extraneural fat, yet at the cost of lower SNR. In order to still maintain sufficient SNR, we chose the echo time as short as possible, used 6/8 partial Fourier sampling, applied 20 gradient directions and used a relatively low upper b-value of 1,000 s/mm 2 .
A study by Zhou et at. noted that SNR and spatial resolution had a strong impact on FA values in forearm nerves. They found FA values to be higher with low SNR and to be lower with higher voxel size [29]. Whilst keeping voxel size and ROIs as small as possible, both diffusion sequences seem to be appropriate for quantitative DTI. A reported phenomenon of SMS acquisition, which is not yet fully understood, is a potential increase of measured FA values at higher slice acceleration; however, in the present study we did not make such observations [30,31]. In theory, given its shorter echo time and echo spacing, readout-segmented EPI inherently reduces susceptibility artifacts and geometric distortions compared to single-shot measurements. However, our visual image assessment did not reveal a significant advantage of SMS rs-EPI over ss-EPI. Recently another promising multi-shot DTI technique, termed multiplex sensitivity-encoding (MUSE) has been developed. This technique inherently corrects nonlinear shot-to-shot phase variations without using navigator echoes [32].

Limitations
We acknowledge several limitations of the current study: Our sample size with only ten healthy volunteers is relatively small, but in line with previous studies on peripheral nerves, and typical for pilot studies investigating novel techniques [9,29].
Furthermore, we suppose that a larger sample size would not have had a relevant impact on the results of this study, because the inter-individual variability of DTI measurements was relatively low. Last, the phased-array coil used in our study comprised only four independent channels; hence, we could not take unlimited advantage of the multi-slice technique. However, previous studies have shown that a slice acceleration factor of 2 is the most reasonable compromise between image quality and acquisition time [7,31].

Conclusion
In our study, we found that SMS rs-EPI was not superior to ss-EPI for DTI of the ulnar nerve at the elbow. In order to increase clinical acceptance and impact of diffusion-weighted imaging of peripheral nerves, further attempts are necessary to find the best technique.