Diffusion Tensor Imaging for Peripheral Nerves in the Upper Extremities using Realtime B0 Correction & Image based Distortion Correction: A feasibility study
Maggie Mei Kei Fung1, Ek Tsoon Tan2, David Soon Yiew Sia3, and Darryl Sneag3

1MR Apps & Workflow, GE Healthcare, New York, NY, United States, 2MR, GE Global Research Center, Niskayuna, NY, United States, 3MRI Research Lab, Hospital of Special Surgery, New York, NY, United States

Synopsis

Diffusion tensor imaging (DTI) can potentially be helpful in visualizing peripheral nerves and assessing nerve damages. However, upper extremity DTIs (wrist, elbows & arm) are susceptible to distortion and fat suppression failure, especially in arms-down position where the area of interest is far from iso-center and can have more B0 inhomogeneity. In this study, we aim to investigate whether a combination of B0 correction methods can help reduce fat suppression failure, improve spatial misalignment and thus improve nerve tracking. We observed consistent fat suppression improvement at the wrist, but no significant improvement in spatial accuracy.

Purpose:

DTI has been utilized in imaging nerve injuries and peripheral nerve tumors[1].However, upper extremity(UE) DTI is susceptible to distortion and fat suppression failure, especially in the patient-friendly arms-down position where the area of interest is far from iso-center and has increased B0 inhomogeneity. Spatial mismatch makes it difficult to do proper seeds placement and tracks the nerve in its entirety in fiber tracking software. In this study, we aimed to investigate the utility of a realtime B0 correction(RTB0) method[2], and an image registration-based distortion correction (ADDC) method [3,4] in UE-DTI.

Methods:

To reduce B0 inhomogeneity & fat suppression failure, we applied the RTB0 technique, which examined the FID for each slice to determine the optimal center frequency and apply it to each slice in realtime (Fig 1). To further reduce misalignment due to tissue susceptibility and eddy-current-induced distortion, the ADDC post-processing technique is used , where a rigid registration of the DTI T2 image to the PD image was performed. To account for susceptibility & eddy current effects, an additional refinement of alignment was performed using 20-parameter cubic-polynomial and 4-parameter linear basis-function respectively. Peripheral nerve DTIs with and without RTB0 were performed on 3 consented healthy volunteers on a GE 3T 60cm bore scanner (MR750) using a 16channel Flexible extremity array in arms down position. Optimized DWI parameters were: FOV:12cm(LR)x9.6cm(AP), Matrix: 80(freq) x 64(phase), TR/TE:5000ms/60.5ms (non-RTB0), 66.5ms(RTB0), single spin echo, BW:250kHz, Slice thickness:3mm, # slices:25/station (i.e. 9cm per station), Adiabatic IR fat suppression , TI=79ms, b-value=0s/mm2(4 NEX), 600s/mm2 (4 NEX), diffusion direction: 15, scan time: 5:25min. Since the most common injury sites of the upper extremity are wrist, elbow and upper arm, we performed DTI at these 3 sites for each volunteer (adjusting the coil placement for each region of interest). Both datasets were then processed through ADDC.

To assess the spatial accuracy, an experience radiologist identified the median, ulnar and radial nerve positions on the proximal, mid, distal locations of each imaging site. The nerve positions were identified on the following datasets: 1. Anatomical PD images, 2. DTI without RTB0, 3. DTI with RTB0, 4. DTI without RTB0 & ADDC, 5. DTI with RTB0 & ADDC. Using the PD image as gold standard, the absolute displacement errors were computed for datasets 2 to 5 in AP and LR direction, and compared using Wilcoxon rank sum test. Fat suppression quality was also noted in the comparison.

Results & Discussion:

Figure 2 shows the center frequency offset calculated from RTB0 at awrist, elbow and arm respectively on a volunteer. For elbow & arm, since they are relatively cylindrical in shape, the CF offset is mainly driven by magnet B0 inhomogenity, which is minimal in the 9cm SI coverage range. While in wrist, where there is shape changes at the wrist joint, there is a higher CF offset due to changes in tissue susceptibility. This large CF offset contributes to fat suppression failure at the distal wrist with conventional DTI in all 3 volunteers. RTB0 eliminates this fat suppression failure since it enables the spectrally selective water excitation pulse to operate at the optimal water peak even at off-isocenter location (Figure 3). Figure 4 shows the absolute displacement error in AP & LR for datasets 2 to 5. We did not observed improvement in spatial accuracy, and upon further inspection, it could be due to the differences in X/Y/Z shim values & CF in DTI vs RTB0 series, which contributed to stretching, skewing and shifting of the image. In future experiement, we should fix the CF & shimming parameters to eliminate this confounding factor. Figure 5 shows a 3D rendering of the peripheral nerve around a nerve sheath tumor at the wrist in a patient. Note that nerve bundles were best tracked in the dataset with both RTB0 & ADDC applied in this case, showing that ADDC & RTB0 could potentially help with visualization of nerve bundles.

Conclusion:

In this work, we demonstrated fat ghosting reduction using the RTB0 acquisition method. And in our initial investigation, the combination of RTB0 & ADDC method does not provide significant difference or improvement in alignment in this small SI coverage scenario. Future investigation can look at larger SI coverage & comparison with fixed shimming parameters to better assess the spatial alignment differences.

Acknowledgements

No acknowledgement found.

References

[1] Naraghi el al, Semin Musculoskelet Radiol 2015,19:191–200. [2] Fung et al, ISMRM 2015 Proceeding, 1606, [3] Jenkinson et al, NeuroImage 2002, 17(2), 825-841, [4]Klein et al, IEEE Transactions on Medical Imaging 2010, 29(1):196 - 205.

Figures

PSD diagram of DWI with FID acquisition after the 90º excitation for real time B0 correction. Diffusion gradient (dotted) is off during dummy acquisitions .

Slice-wise center frequency offsets obtained from RTB0 in 1 volunteer. CF offset variations are minimal for elbow & wrist, but much higher in wrist where there is shape variation and therefore higher tissue susceptibility.

Image comparison between A)Axial PD, B)conventional DTI & C)DTI with RTB0 at the wrist. Note the fail fat suppression (arrow) in conventional DTI, which is corrected by RTB0. The fat signal also overlap the ulnar nerve (double arrow) in B), which make the nerve assessment impossible, while RTB0 nicely depict the nerve location.

Absolute displacement errors of the 4 DTI datasets in AP (4A), and LR (4B) directions. Mean & standard deviation are listed in Table 4C.

Visualization of ulnar nerve bundles around a nerve sheath tumor (arrow) at the wrist. The nerves were better visualized around the tumor with a combination of RTB0 & ADDC.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2287