Ultra-high field upper extremity peripheral nerve and non-contrast enhanced vascular imaging

Objective The purpose of this study was to explore the efficacy of Ultra-high field [UHF] 7 Tesla [T] MRI as compared to 3T MRI in non-contrast enhanced [nCE] imaging of structural anatomy in the elbow, forearm, and hand [upper extremity]. Materials and method A wide range of sequences including T1 weighted [T1] volumetric interpolate breath-hold exam [VIBE], T2 weighted [T2] double-echo steady state [DESS], susceptibility weighted imaging [SWI], time-of-flight [TOF], diffusion tensor imaging [DTI], and diffusion spectrum imaging [DSI] were optimized and incorporated with a radiofrequency [RF] coil system composed of a transverse electromagnetic [TEM] transmit coil combined with an 8-channel receive-only array for 7T upper extremity [UE] imaging. In addition, Siemens optimized protocol/sequences were used on a 3T scanner and the resulting images from T1 VIBE and T2 DESS were compared to that obtained at 7T qualitatively and quantitatively [SWI was only qualitatively compared]. DSI studio was utilized to identify nerves based on analysis of diffusion weighted derived fractional anisotropy images. Images of forearm vasculature were extracted using a paint grow manual segmentation method based on MIPAV [Medical Image Processing, Analysis, and Visualization]. Results High resolution and high quality signal-to-noise ratio [SNR] and contrast-to-noise ratio [CNR]—images of the hand, forearm, and elbow were acquired with nearly homogeneous 7T excitation. Measured [performed on the T1 VIBE and T2 DESS sequences] SNR and CNR values were almost doubled at 7T vs. 3T. Cartilage, synovial fluid and tendon structures could be seen with higher clarity in the 7T T1 and T2 weighted images. SWI allowed high resolution and better quality imaging of large and medium sized arteries and veins, capillary networks and arteriovenous anastomoses at 7T when compared to 3T. 7T diffusion weighted sequence [not performed at 3T] demonstrates that the forearm nerves are clearly delineated by fiber tractography. The proper digital palmar arteries and superficial palmar arch could also be clearly visualized using TOF nCE 7T MRI. Conclusion Ultra-high resolution neurovascular imaging in upper extremities is possible at 7T without use of renal toxic intravenous contrast. 7T MRI can provide superior peripheral nerve [based on fiber anisotropy and diffusion coefficient parameters derived from diffusion tensor/spectrum imaging] and vascular [nCE MRA and vessel segmentation] imaging.

. Although, commercial [1.5T/3T] MRI is routinely used in upper extremity imaging [shoulder, forearm, hand, and wrist], the resolution achieved could be limited. UHF MRI could significantly improve the resolution capabilities of conventional MRI [3,4]. However, 7T human MRI is fraught with technical challenges mainly related to radiofrequency [RF] field inhomogeneity and safety concerns for neurologic and whole-body imaging applications [1,[5][6][7][8],. The upper extremity offers the advantage of a compact isolated anatomy with relatively smaller electric size [when compared to the head/abdomen]. Therefore it can be relatively easier to generate uniform and safer RF field distributions for 7T in upper extremity imaging. As a result, significantly higher anatomical resolution and further improvements in contrast of tissue interfaces can potentially be achieved at 7T for upper extremity applications without increased scanning times and/or considerable compromise in overall image quality.
Higher SNR/CNR can be useful in providing intricate detail in neural [9], vascular [10], cartilage [11], tendon [12] and joint [13] imaging. As described in [14][15][16], 2D and 3D multi planar high resolution T1 weighted [T1] volumetric interpolated breath-hold exam [VIBE] imaging produces optimal anatomic detail and reveals high resolution structures. Nerves demonstrate a signal similar to adjacent soft tissue [14,17]. High resolution T2 weighted [T2] double-echo steady state [DESS] can achieve excellent SNR/CNR for discriminating nerves from muscle, fascia, cartilage and synovial tissue in forearm and joint structures [18]. According to [19], susceptibility weighted imaging [SWI] is particularly useful for visualization of vasculature since it is highly sensitive to deoxyhemoglobin in venous blood, making it useful in imaging vascular trauma, abnormalities and visualizing neovascularization of tumors. Diffusion based sequences specifically monitor the random movement of water molecules in anisotropic tissue [20,21].
One of the major obstacles in UHF MRI remains the limited availability of customdesigned coils [22] optimized for musculoskeletal applications. Some prior studies have demonstrated the utility and feasibility of customized coil designs for high-resolution wrist imaging at 7T [4,[23][24][25]. Our group has recently developed a custom-designed forearm/hand MRI RF coil system [26][27][28] for 7T UHF imaging. The combination of such a coil system in conjunction with 7T UHF MRI and many of the aforementioned sequences could bear significant potential in upper extremity imaging applications. Some of these applications include 1) sequential monitoring of regeneration after peripheral nerve [PN] repair, which is critical for evaluation of re-innervation and in planning treatment strategies [29,30] and 2) monitoring vascular pathology or outcomes after vascular interventions without need for potentially nephrotoxic or anaphylotoxic contrast agents [31,32].
In this study, utilizing a custom-designed RF coil system, we present results and findings  from various MR sequences [T1 VIBE, T2 DESS, SWI, time-of-flight [TOF], diffusion tensor imaging [DTI], and diffusion spectrum imaging [DSI]] that were optimized for 7T upper extremity imaging while keeping scanning times similar to or lesser than those used at 3T. We show detailed SNR and CNR comparisons of 3T vs. 7T for T1 VIBE and T2 DESS sequences. Furthermore, we show nerve tractography [diffusion imaging], vessel segmentation [nCE magnetic resonance angiography], and imaging of cartilage, synovial fluid, bone marrow and joint anatomy in forearm, hand, and elbow at 7T.

MR hardware and study participants
Briefly, a shielded design of an actively detuned transverse electromagnetic TEM resonator [26,27] with two ports driven by a quadrature hybrid was used in this study. An in-house finite difference time domain [FDTD] package was utilized [6,33] in order to calculate the B 1 + field [transmit field responsible for excitation] and specific absorption rate [34] parameters. An eight channel receive-only insert array was designed and utilized with the TEM coil [26]. The array consisted of eight inductively decoupled surface loops [35] [each loop is 18x8 cm 2 in size] distributed evenly to fit inside the structure of the transmit coil and cover the region of interest.

MR examination
Imaging was performed using a 7T whole-body MRI system [Magnetom, Siemens Healthcare, Erlangen, Germany] and 3T whole-body MRI system [Tim Trio, Siemens Healthcare, Erlangen, Germany] at University of Pittsburgh. The above mentioned RF coil system was utilized at 7T. [26][27][28]. A dedicated eight-channel extremity coil [Siemens, Cleveland, OH, USA] was used at 3T. Coils were positioned in the center of the magnet bore during all imaging sessions. Subjects were positioned prone within the coil with the hands [forearm, and elbow] placed over the head and immobilized with cushions, pads and sandbags to avoid discomfort.

MR imaging
Five imaging sequences [T1 VIBE, T2 DESS, SWI, DTI, DSI, and TOF] were optimized by a highly experienced MR application scientist on the basis of a clinical MRI protocol for upper extremity pathologies. 3T and 7T protocols were optimized for same scanning time comparing a i] Higher Resolution [HR] protocol and ii] Lower Resolution [LR] protocol. We optimized the protocol on normal volunteers at 7T with representative varied resolutions using visual assessment for best image quality. Then for 3T, the protocol was adapted from the 7T protocol while using proper 3T imaging parameters. The protocol was again tested and optimized on normal volunteers and visual inspection was used to pick the best image quality. This study compares T1 VIBE, and T2 DESS qualitatively and quantitatively [SWI was only qualitatively compared.] 3T vs. 7T diffusion and TOF comparisons were not performed for this study. All the sequence parameters are documented in Tables 1 and 2. In the case of the 7T diffusion sequences, the specific b-values were optimized by balancing b-field directions-64-with higher b-values [DTI: 0 to 1300; DSI: 0 to 2000] while obtaining 1] smaller voxel resolution, and 2] acceptable time of acquisition [20] and SAR. 7T with parallel imaging factor of 2 and 3 were utilized to achieve the results shown [Tables 1 and 2]. . Two musculoskeletal radiologists with over thirty-five years of combined clinical experience blindly evaluated images and sequences. All measurements were performed by the same researcher and verified by the MR application scientist and both musculoskeletal radiologists.
Qualitative analysis. Criteria for Evaluation of Image Quality: Contrast resolution, sharpness and clarity were the criteria considered in evaluation of image quality, using a scale as follows: 1-nondiagnostic, 2-poor, 3-fair, 4-good and 5-excellent [ Table 3].
Evaluation of Imaging Artifacts: Presence of chemical shift, susceptibility artifact and motion artifact were considered in evaluation of artifacts using a scale as follows: 1-absent, 2-present but not affecting anatomic detail and 3-present and severely affecting image interpretation.
Evaluation of image quality and artifacts were both carried out for 3T and 7T for T1 VIBE, T2 DESS, and T2 Ã SWI images [Figs 1 [left], 2 [left], [3][4][5][6]. The sequences were evaluated retrospectively and independently by two musculoskeletal radiologists who were blinded to the imaging parameters. MR image processing. For post-processing, diffusion-weighted images were used to perform fiber tractography of peripheral nerves in forearm, and the principal fiber directions were used to conduct streamline fiber tracking. After DTI data were transferred to a local computer, FSL [FMRIB, Oxford, UK] and DSI studio [CMU, Pittsburgh, PA] were used for distortion correction and fiber tracking, respectively. Eddy current induces geometric distortion in different acquisition directions. This eddy current is produced in conductive materials/parts of RF coil by MRI gradient coils [20,36]. Combined with subject/patient motion, it can severely damage the diffusion estimation due to pixel mismatch in the temporal image series [20]. Using eddy-correct library available in FSL [37], eddy current and motion correction were implemented where input reference image was non-diffusion weighted image [b = 0]. Then, T1 and T2 weighted images with DTI derived maps [ Fig 7C-7F as later discussed in the Results Section] were utilized to locate the forearm nerves. Circular regions of interest [ROIs] were placed in the anatomic location of the nerves from the FA [ Fig 7D] and color-coded maps [ Fig 7F]. The fiber trajectories generated by an orientation distribution function-streamlined version of the Fiber Assignment by Continuous Tracking algorithm [38][39][40] were used to sample the FA and apparent diffusion coefficient [ADC] maps. Using a random seeding approach, we initiated tracking, from each random position within the seed mask, in the direction of the most prominent fiber. The Following parameters were optimized in order to distinguish nerves from surrounding various anatomical structures: an anisotropy threshold of 0.25, a step size of 0.35 to 0.5 mm, minimum fiber length of 0 to 20 mm, and a turning angle threshold of 60 degrees. The tracking was terminated when the relative fractional anisotropy for the incoming direction dropped below a preset threshold~0.2 [heavily dependent on SNR of the specific subject's scan data] or the turning angle exceeded 60 degrees. Each trajectory generates one profile, and all profiles were averaged to obtain the trend of the index along the fiber orientation. After anatomic confirmation of the nerve tract, mean FA and ADC values were calculated as discussed later in results section.
We used non-contrast enhanced [nCE] angiographic imaging and vessel segmentation techniques to image the digital proper palmar arteries at 7T. In order to extract the three dimensional structure of vasculature in forearm, the T1 VIBE DICOM images were exported in DICOM format to MIPAV [41]. A mask was created manually using paint grow segmentation method in MIPAV, which utilizes the concept of voxel aggregation by grouping the seed points within the volume of interest [selected manually by minimum intensity projection [MIP]]. The vasculature structures were manually traced in order to avoid errors in identifying the arteries and venous structure. After masking each segment [without skeletonizing or dilating to preserve structural vessel information], a surface and texture volume-rendering module was used to enable 3D visualization of the anatomy and segmented structures.

3T vs 7T
Quantitative analysis. Figs 1 and 2   times that of 3T. As slices close to dorsal part of forearm do not have anatomic structures suitable for measuring CNR, some slices were excluded.
In terms of RF homogeneity [a significant issue [6] at 7T], T1 VIBE and T2 DESS images [Fig 3] demonstrate uniform excitation in axial, coronal and sagittal planes covering the complete volume of forearm. On the 1 st subject, the uniformity was optimized by iteratively fine tuning the RF Tx coil utilizing the experimentally measured B 1 + field inside the coil's volume of interest. That tune was then utilized for all the remaining subjects. The B 1 + field homogeneity demonstrated by COV [Coefficient of variation] was calculated to be 21% in the forearm [encapsulated within the coil's volume] [26]. Also, IEC guideline was followed and SAR was numerically calculated using in-house FDTD method [6,33]. Based on a continuous 1.97μT Resolution of small forearm vessels and nerves in both the superficial and deep soft tissues at HR and LR 7T imaging is shown and compared to 3T. Delineation of muscle fibers/fascial plane interfaces as well as cortical bone/soft tissue interfaces is enhanced at 7T illustrating the benefit of higher CNR.
Image Quality and artifacts evaluation: Table 3 lists all the result of qualitative analysis for image quality and artifacts. Overall image quality of T1, T2 and T2 Ã images at 7T is higher when compared to 3T. Overall artifacts were increased from 3T to 7T but did not impact the image quality.   [42] images which show macro and micro vessels in addition to vessel wall without the use of any kind of Contrast dye/liquid. In comparison to the T1 VIBE slices in the top row, SWI in bottom row clearly depicts arterial vascular pattern and muscular perforators highlighting the importance of SWI at ultra-high fields.  Table 4. The hand and forearm nerves correspond to known anatomic distribution of the radial, ulnar and median nerves.

Discussion
While there are no upper extremity commercial coils available at 7T [except [22]], a customized 7T RF coil was built and compared with available extremity coils at 3T with similar receive elements and comparable dimensions. The use of a combined transmit and receive coil with a small filling factor would be favorable to detect receive signals from a homogeneously excited Ultra-high field upper extremity MR imaging zone of anatomy of interest. A few studies [43][44][45][46] have shown that the SNR obtained from a combined transmit and receive only system could be higher even with a limited FOV. In parallel imaging, the noise distribution is heterogeneous throughout the images. So SNR calculation is approximate but won't be exactly accurate using ROI method [signal mean in anatomy of interest and noise standard deviation in background]. The detailed coil analysis study [47] covers the noise pre-scan method [pseudo-replica method]. When comparing a birdcage coil to a TEM coil, the current produced in end rings of the birdcage coil can generate a considerably high magnetic field component which is coaxially aligned with B 0 static field, and therefore does not result in spin excitation. Such a component can increase Ohmic loss by inducing conduction currents in tissue [48]. Other groups have developed birdcage-based commercial coils and HEM-mode-based resonators demonstrating promising imaging capabilities limited to the wrist [22,24,49]. In contrast, and to our knowledge our current study is the first to explore the utility and relevance of 7T UHF MRI for the upper extremity encompassing the hand, wrist, forearm and elbow.
The quantitative analysis performed on T1 VIBE and T2 DESS images [Figs 1 and 2] demonstrates that higher SNR/CNR were achieved at 7T, almost twice when compared to 3T [most especially in the HR protocol]. As in parallel imaging, the noise distribution is heterogeneous throughout the images, therefore SNR calculation could be a crude approximation using ROI method: signal mean in anatomy of interest over standard deviation of the noise in background. Additionally, Fig 3 shows that homogenous excitation [a typical hurdle for 7T imaging] is quite possible for the forearm with coverage that extends from wrist to elbow. As a part of the qualitative analysis seen in Diffusion-based MRI is utilized as a non-invasive, non-disruptive strategy for sequential assessment of forearm nerves. It is quantified by water diffusion parameters [FA and ADC] as indirect correlates or surrogates of neuroregeneration after transection repair or transplant related nerve outcomes. In order to increase the reliability and interpretation of results, DTI derived maps [FA, ADC, and color-coded] were defined in conjunction with T1 and T2 anatomical validation. High resolution imaging of peripheral nerves may have application in the detection of neuropathy and monitoring to assess disease progression or response to appropriate therapy [15,30] [50]. Moreover, volume rendering and 3D depiction of the course of the forearm nerves [as shown Fig 7] may prove useful in the broader realm of reconstructive surgical or vascularized composite allotransplantation [VCA] applications. While comparison to 3T [in terms of DTI] was outside of scope for this study, the presented 7T images show very good delineation of the forearm nerves, without the use of contrast agents.
7T TOF imaging [shown Fig 8] and corresponding vascular segmentation [shown in Fig 9] allow depiction of macro and microvascular anatomy of the hand, forearm, and hand [wrist] in 2D monitoring of microvascular integrity as it relates to certain vascular disease processes [32,51]. In addition, SWI [ Fig 6] could be especially useful in visualizing various macro or micro-vascular pathologies secondary to vascular trauma, thrombotic/embolic occlusion, and neovascularization of tumors, and in high-resolution MR venography [19,32,52]. While we optimized the SWI sequence for image quality, this specific sequence with the current standard clinical practices, may not be appropriate for studying the anatomic structures in question. While contrast enhanced imaging is utilized extensively [53], the fact that 7T high quality vascular imaging can be accomplished without the use of intravenous contrast is invaluable in patients with organ transplantation, autoimmune vasculitis or diabetic peripheral vascular disease as these conditions often have concomitant renal vasculitis and/or renal insufficiency.
MR angiography using 1.5T imaging has been previously described in evaluating neuropathic leg pain [54] and other diabetic vasculopathy [55]. 3T imaging has been reported in evaluation of the brachial plexus [56,57]. In addition, 3T MR angiography has also been used in imaging of intracranial vessel disease/vasculitis [58,59]. 7T MR angiography has also been explored [60] specifically in the upper extremity focusing on wrist [61] or hand [49] but limited to the palmar vasculature. The work presented here expands higher resolution MRA to other regions including the digital vasculature and capillary networks in the pulps of fingers.
The choice of CT Angiography [CTA] vs MRA or vice versa for specific indications is still a debatable issue [62][63][64]. The choice of CTA, usually depends upon the region of interest involved in either neuro [65][66][67][68], coronary [34,[69][70][71][72] [19,60,[73][74][75][76][77]. While CTA faces contrast-agent related nephrotoxic or anaphylotoxic risks, it has the advantage of a faster scan time that could be important in emergent clinical indications such as pulmonary embolism or acute cerebrovascular disease [65]. Earlier work however has shown that the diagnostic performance of 3T CE whole-heart coronary MR angiography approaches the diagnostic performance of 64-section CT [78]. In addition, 7T UHF MRI, and parallel imaging with higher number of receive channels could produce highly improved signal/contrast to noise ratio, faster scanning time, and significant advantage for repeated or longitudinal/sequential imaging applications [78][79][80]. Furthermore, the avoidance of radiation by MR offers critical advantage in pediatric imaging especially of the craniofacial skeleton [in concussion or traumatic brain injury] [81]. When compared to MR related research studies, a significant growth in research utilization of CT has been observed [82]. However, there is need for more MRA studies [especially utilizing UHF MRI] [19,60,[73][74][75][76][77]. Non-invasive methods like nCE MRA, TOF, SWI, and diffusion related methods have very high potential in safe, non-radiation, nCE detection of high-signal-intensity plaque, coronary arterial wall [69], and atherosclerotic vascular disease [83,84] as well as in a host of other indications [19, 60, 65-70, 73, 74, 85]. UHF 7T imaging can alleviate and overcome several imaging challenges encountered in neurovascular disease at lower field strengths [1.5T and 3T] like venous contamination obscuring underlying arterial architecture or delineation of contextual anatomical structures [65].

Conclusion
High quality non-contrast enhanced ultra-high resolution neuro-vascular upper extremity imaging is possible at 7T. High quality 7T images were obtained using T1 VIBE, T2 DESS and T2 Ã SWI sequences as well as TOF, DTI, and DSI sequences in conjunction with a customdesigned RF coil system. Analysis demonstrate that 1] CNR/SNR at 7T are almost two folds [for T1 VIBE and T2 DESS sequences] of that achieved at 3T for upper extremity applications and 2] broader clinical potential is present for 7T imaging in musculoskeletal applications.