In this pilot study, the feasibility of using advanced qMRI methods in conjunction with MRN in the PNS was investigated, with application to the sciatic nerve in healthy volunteers. The rationale for this study was based on the lack of prospective studies employing advanced qMRI methods to study the PNS, despite their demonstrated ability to provide biophysically meaningful information pertinent to the underlying pathophysiological mechanisms involved in neurological disease5. The main reasons behind the lack of implementation of qMRI methods in this context are likely to be related to the technical challenges associated with imaging the peripheral nerves, which include among others the small structure of the peripheral nerves, the surrounding tissue types with different magnetic susceptibility properties, the blood vessel distribution and flow effects, and RF coil and pulse sequence designs.
In this study, the reproducibility and reliability of a multi-parametric qMRI protocol was investigated by taking into account the main technical challenges involved to perform multi-shell DWI (for the estimation of DTI and DKI metrics), qMT (for the estimation of BPF and T2B) and IR (for the estimation of qT1) on a commercial 3T MRI system. A key feature of the qMRI protocol in this study included the use of ZOOM EPI, which benefits from the combined use of fat saturation with inner volume excitation, allowing alias free images with reduced sensitivity to the susceptibility artifacts that commonly affect long EPI readout acquisitions42,43. In addition, the qMRI protocol was acquired with a unified MRI signal readout, which enabled the acquisition of a uniquely rich set of image contrasts with matched resolution, distortion and intrinsic geometric alignment, all important aspects for successful multimodal characterisation of neural tissue microstructure44. Indeed, our unified-ZOOM EPI acquisition strategy is one of the main innovations of this work. The strategy makes high-quality multi-parametric qMRI feasible in the PNS, and has the potential of bringing advanced MRI methods for quantitative microstructural assessment one step closer to the clinic. Without the use of ZOOM-EPI for all our qMRI metrics, taking advantage of acceleration methods such as multi-slice excitation and parallel imaging reconstruction, this protocol would be much longer, and as a consequence a reduced choice of metrics would be sampled.
Previous studies have used DWI to examine the median, ulnar, radial, tibial and sciatic nerves, demonstrating the reliability of these measurements8, and how these maybe influenced by the anatomical location, age, sex, height, weight, body mass index (BMI)9,10, and by the nature of the pathological conditions implicating the PNS11–15. These studies have focussed on conventional DTI metrics (i.e., MD, RD, AD and FA), thus providing more specific information related to axon and myelin sheath integrity, typically invisible with conventional structural imaging.
This study sought to examine the feasibility of extending such previous approaches focussing on conventional DTI metrics by accounting for non-Gaussian diffusion through DKI17, in order to obtain additional information related to neural tissue microstructure. In terms of the standard DTI metrics, the mean values obtained in this study seem to follow a similar trend and appear to be in agreement with the results of previous studies examining the sciatic nerve in healthy controls8,10, although differences in technical and demographic factors do not permit a direct comparison. The potential value and feasibility of DKI measurements in the PNS has only recently been addressed53, but more studies will be required in the future to understand the additional information related to tissue composition and microstructure that can potentially be obtained in pathological conditions affecting the PNS.
The role of qMT in the study of the PNS currently remains unexplored, despite the potential benefits previously demonstrated31–36, with most of the studies in the PNS currently relying on semi-quantitative assessments of myelin content through the use of MTR23–27. Future studies will therefore aim to clarify the potential benefits of using qMT in the PNS over semi-quantitative approaches like MTR. However, similar to MTR measurements, it is important to recognise the differences in tissue structure and composition of the nerves in the PNS as compared to the CNS, in order to interpret the origin of the qMT contrast in future investigations of pathological conditions affecting the PNS. In particular, assuming a two-pool description of biological tissues, the bound pool fraction in the PNS is represented by various tissue types, such as collagen, myelin and the proteins contained in the axons and Schwan cells23,25, thus dissimilar to the CNS tissue composition. The relative contribution of each tissue type to the qMT measurements remains unknown, and future research will be directed at addressing this gap.
The measurement of T1 relaxation time has provided invaluable information in a variety of clinical applications over the years37, although the lack of implementation of T1 relaxometry in the study of the PNS is likely explained by the technical challenges associated with obtaining accurate T1 measurements40. Despite the availability of a variety of time-efficient T1 mapping methods to study the CNS54–57, their translation to the PNS may not be straightforward, and could be hampered by a number of site-specific factors such as radiofrequency pulse imperfections and incomplete magnetisation spoiling40. In this study, an IR-based T1 mapping method was used, benefiting from the aforementioned inherent qualities of ZOOM EPI acquisitions42,43, while making use of a slice shuffling scheme to allow T1 mapping of a large section of the sciatic nerve in a clinically acceptable scan time46. The mean T1 relaxation time in the healthy sciatic nerve was found to be longer (1635 ms) than previously measured in the healthy cervical spinal cord (1142 ms) using the same approach46, and also slightly longer than the previously reported T1 relaxation time in the healthy median nerve (1410 ms) at 3T, using a different T1 mapping approach58. As previously mentioned, these variations might be explained by technical factors, anatomical location i.e., differences in tissue organisation and properties, and also demographic factors. More research will be required in the future to determine the role of T1 relaxometry in the study of the PNS.
In this study, reproducibility of the qMRI measurements was assessed by means of calculating the scan-rescan, intra-rater and inter-rater %COV from repeated measurements in a subset of study participants. Furthermore, intra-rater and inter-rater reproducibility was assessed by means of calculating the ICC. In order to assess the intra-rater and inter-rater quality of the image segmentations, the DSC was also calculated. Similar studies in the literature utilising a multi-parametric qMRI protocol with which to directly compare the reproducibility results of this study are not available. When comparing the %COV values obtained in this study with previous similar investigations in the brain and spinal cord, however, one must take into consideration the smaller size of the structure evaluated in this study. In particular, partial volume averaging is expected to have higher influence when assessing smaller structures, which is also supported by the moderate intra-rater and inter-rater DSC results in this study. Nevertheless, the ICC results of this study for the standard DTI metrics have demonstrated good to excellent agreement59, and are in line with previous similar investigations in the sciatic nerve8.
Finally, we acknowledge a number of limitations of our approach. One of main limitations of this study is that with 12 subjects the effect of age, gender, BMI, height and weight on the multi-parametric qMRI measurements obtained in this pilot study was not examined specifically, even though some of the demographic factors have been shown to influence various qMRI measurements significantly10. In addition, some of the unique technical features employed in this study, for example the use of ZOOM EPI, together with modifications to the sequence design in order to allow time-efficient acquisition of the qMRI metrics reported herein, may not be readily available in non-specialist centres, thus limiting the widespread implementation of the proposed qMRI protocol. Also, while DTI fitting is now part of most scanner software packages, analysis of advanced features such as DKI, qMT and qT1 are bespoke for our protocol.
In conclusion, this pilot study demonstrates the feasibility of combining MRN with a rich multi-parametric qMRI protocol based on a unified ZOOM-EPI readout, which enables the measurement of diffusion, quantitative magnetisation transfer and T1 relaxation properties of the healthy sciatic nerve in vivo. The reproducibility of the qMRI methods employed were found to be consistent with previous studies, demonstrated by the comparably low %COV, high ICC and DSC values obtained from scan-rescan sessions, intra-rater and inter-rater assessments. Future investigations involving a larger sample population will be required to confirm the findings of this study, to explore the demographic determinants of the qMRI measurements investigated, and to determine their potential role in pathological conditions implicating the PNS.