Introduction

Measuring spinal cord blood flow (SCBF) plays an important role in studying normal spinal cord function and diagnosing spinal cord injuries. Perfusion imaging including Arterial Spinal Labeling (ASL) has been successfully applied to measuring cerebral blood flow (CBF) of human brains. A few preclinical studies^1,2 have investigated spinal cord blood flow with ultra-high field MRI in animals. However, few studies have applied ASL to studying spinal cord perfusion in humans. In this study, we optimized multi-delay pCASL to measure cervical spinal cord blood flow on healthy humans.

Methods

1. Labeling position selection
Figure 1a shows the anatomical structure and blood supply on each level of the spinal cord. In this study, we focused on cervical levels, especially C1 to C3 segments, where the main blood supply is from vertebral arteries, which branch into anterior and posterior spinal arteries and radiculo-medullary arteries. In other words, there are 2 blood supplies to the spinal cord, one vertical from anterior and posterior spinal arteries and another horizontal from radiculo-medullary arteries. Figure 1b shows the imaging slab and the labeling plane that was placed 50-60mm below the center of imaging slab. Figure 2 shows the phase-contrast image in the labeling plane and the flow velocity both in the vertebral arteries (VA) and carotid arteries (CA). The mean flow velocities in VA were lower than in CA (31.1±4.9cm/sec in CA and 20.3±1.4cm/sec in VA), resulting in slightly reduced labeling efficiency (0.90±0.01 in CA and 0.85±0.01 in VA).

2. Modification of pCASL protocol
Since the spinal cord is surrounded by cerebrospinal fluid (CSF) with pulsatile flow that may influence perfusion signals, the background suppression module of the sequence was optimized³ to suppress CSF signals rather than tissues. The T1 range of the CSF was assumed to be 4200-4600ms according to literature^4,5.

3. In-vivo experiments
Data was collected on 5 healthy subjects (age=26±1.9years, 5 males) on a Siemens 3T Prisma scanner with 32-channel head coil and spinal array. The parameters for multi-delay pCASL included in-plane resolution of 2×2mm2, FOV=96×96, TR=4000ms, TE=20ms, 24 slices of 2mm thickness. The labeling duration was 1500ms and 10 post labeling delays (PLDs) were evenly spaced between 500ms and 2500ms with 2 measurements at each PLD. The total perfusion scan time was about 10 minutes. M0 image was acquired with no background suppression. RESOLVE diffusion MRI (1×1×2 mm³) was used to visualize the gray and white matter contrast for segmentation in the spinal cord. A single-slice pulse-triggered phase-contrast MRI was acquired on the labeling plane to measure the flow velocity of arteries.

4. Data processing and quantification
The SCBF and arterial transit time (ATT) maps were fitted pixel by pixel using the dynamic model proposed by⁶. The spinal cord ROI was selected manually on the diffusion image using ITK-SNAP software. The diffusion image and ROI mask were then coregistered to the M0 image using SPM12.

Results

Figure 3 shows the comparison of regular ASL sequence with background suppression of brain tissue and the modified sequence to suppression CSF. With modified background suppression, reduced CSF signal and less distortion can be observed in spinal cord regions. Figure 4a shows the structure image, apparent diffusion coefficient (ADC) image, M0 image and the fitted SCBF map within the ROI. The structural image shows a clear contour of the spinal cord, and the ADC image shows gray and white matter contrast. Table 1 shows the quantification of SCBF and ATT of the 5 subjects within the spinal cord ROI. The fitted SCBF and ATT values are stable and are close to the CBF and ATT values of the white matter in the brain.

Discussion

In this study, multi-delay pCASL was optimized to measure perfusion of the cervical spinal cord (C1-3). The measured SCBF and ATT values are comparable to those of the white matter in human brain. However, there are a few limitations. Firstly, since the voxel size is still coarse given the small size of the spinal cord (~1cm). This will lead to partial volume of gray and white matter, and signal contamination from surround CSF even with improved background suppression. Secondly, we found two peaks in the perfusion time series and this pattern is stable across subjects (Fig. 4b). It may be that the first peak is from blood supply through radiculo-medullary arteries, while the second peak reflects the blood supply from anterior and posterior spinal arteries, which takes longer to reach the spinal cord. This requires further investigation and may require a quantification model that considers dual blood supplies.

Conclusion

In this study, we used multi-delay pCASL to measure SCBF at cervical levels and achieved reasonable SCBF values. Higher resolution is desired to get less partial volume effects and better gray and white matter contrast. The initial results suggest that ASL measurement of spinal cord perfusion at cervical level is feasible, and further developments of imaging protocol and quantification model are required for accurate quantification of SCBF, as well as to expand the technique to thoracic and lumbar segments.

Acknowledgements

No acknowledgement found.