Synopsis

Keywords: Pulse Sequence Design, Perfusion

DiSpect resolves the multi-dimensional displacement spectrum that spins exhibit between tagging and imaging. Previously, we showed that DiSpect can be used to trace blood draining from the capillary bed through the cerebral venous system. This work presents three innovations: 1) Validation of our results with flow phantom experiments and comparison of our in-vivo data with a detailed QSM-venogram. 2) Interleaved imaging to simultaneously probe the deep and superficial veins. 3) Use of a conformal head-cap array for imaging, with higher SNR, allowing dynamic measurement of vein drainage and observation of venous territories across large time scales.

Introduction

Probing the dynamics of cerebral veins can be useful in disease diagnosis and biophysical function studies^1,2. However, current non-contrast enhanced methods are limited in studying venous flow. Arterial Spin Labeling (ASL) is not suitable because it requires a single labeling plane³. Velocity-based techniques (Time of Flight (TOF)/phase-contrast MRI^4,5,6) have limited sensitivity to slow flow. Displacement Spectrum Imaging (DiSpect)⁷ resolves a multi-dimensional displacement spectrum for each image voxel, revealing the blood displacement between tagging and imaging. We have shown that DiSpect can be used to trace blood flow from the capillary bed into the superficial cerebral veins (SCV) and map venous territories⁸, similar to a previously proposed preliminary technique based on remote sensing⁹.

Here, we present several extensions. First, we validate DiSpect in a phantom by computing velocity maps and in-vivo by comparing DiSpect to a QSM-venogram. Second, we perform an interleaved two-slice acquisition to simultaneously image the SCV and deep cerebral veins (DCV). Finally, we image with a conformal head-cap array with higher surface sensitivity. The improved SNR makes it feasible to trace blood draining for a long duration after tagging, allowing us to observe venous territories over large time-scales.

Methods

Flow Phantom:
Scanner experiments were conducted with a 3T GE MR750W system (Waukesha, Wisconsin). DiSpect acquisitions were performed with the sequence shown in Figure 1a. The sequence was implemented on the HeartVista platform (Palo Alto, CA). A flow phantom was constructed to simulate the SCV, with four tubes merging at the top (Figures 2b-c). The phantom was placed inside a water pool and a peristaltic pump was used to draw water out of the tubes (Figure 2a). A 2D-DiSpect acquisition was performed with an axial imaging slice (resolution=4x4mm², FOV=20x20cm²). Displacement encoding (resolution=3x3mm², FOV=4.2x6cm²) was in the LR and SI directions, resulting in a 2D coronal spectrum (projected along AP). Images were acquired from 100ms to 4s after tagging (150ms increments). Image voxels containing the four tubes were selected and their displacement spectra visualized. The average flow velocity of the tubes were estimated based on their displacement spectra.
For validation of estimated velocities, the tubes were imaged with a 4D-flow^10,11 sequence (Resolution=1x1x1mm³/TE=5ms/TR=33ms/FA=5^o/V_enc=5cm/s).

Validation with QSM-Venogram:
We used Quantitative Susceptibility Mapping (QSM) to obtain a detailed map of veins¹². An axial Multi-Echo 3D-GRE sequence was performed in-vivo (resolution=0.7×0.7×2mm³, TE1/ΔTE/TE16=2.3/0.8/14.3ms, TR=50ms, BW=62.5kHz, FA=12^o). STI Suite V3.0¹³ was used to obtain susceptibility maps for each echo with the iLSQR¹⁴ method. The maps from 4-8th echoes were averaged. To visualize a coronal vein structure, a maximum intensity projection was performed on the echo-averaged QSM map along AP direction, with a slab thickness of 5cm.
Next, a 2D-DiSpect acquisition (image resolution=4x4mm², FOV=16x16cm², displacement resolution=6x6mm², FOV=12x7.2cm²) was performed with LR and SI displacement encoding (Figure 3a). Images were acquired from 100ms to 2.2s (150ms increments). For visualization, five vein ROIs were selected and their spectra were overlaid on the QSM-venogram.

Interleaved Two-Slice Acquisition:
An interleaved two slice acquisition (Figure 4a) was performed to simultaneously image SCV and DCV. Two axial slices were chosen: one for SCV (4x4mm², 16x16cm² FOV) and one for DCV (4x4mm², 22x22cm² FOV). Image acquisitions after tagging (100ms to 1.6s, in 100ms increments) were alternated between the two slices. AP and SI displacement encoding was performed (resolution=6x6mm², FOV=12x7.2cm²).

DiSpect with Head-Cap Array:
To improve the SNR, an in-vivo experiment was performed with a custom 15-channel conformal head-cap array¹⁵(Figure 5c). An axial slice (resolution=4x4mm², 16x16cm² FOV) was imaged with LR and SI displacement encoding (resolution=6x6mm², FOV=12x7.2cm²). Images were acquired from 100ms to 3.1s (150ms increments).
A 3D phase-contrast-based venogram was acquired for in-vivo experiments.

Results and Discussion

For the flow phantom experiment, the displacement spectrum of each tube is overlaid on the corresponding schematic of the tubes (Figure 2d). The 4D-flow velocity maps and the estimated velocity maps from DiSpect are shown (Figure 2e). The spectra show water arriving to the imaging slice over 4s from the tubes as well as the pool. The estimated velocity maps from DiSpect closely match 4D-flow results.

The displacement spectra overlaid on QSM-venograms are shown (Figure 3d). QSM recovers smaller veins that are missing in the phase-contrast-venogram (Figure 3e). The displacement spectra show great overlap with the QSM-venogram, also displaying flow from small veins.

For the interleaved acquisition, five vein ROIs were selected from the slices and their spectra were overlaid on the sagittal phase-contrast-venogram displaying the corresponding vein (Figure 4c). The acquisition can simultaneously recover flow from SCV and DCV enabling full brain coverage.

Finally for the dataset with the head-cap, four vein ROIs were selected and their spectra overlaid on the coronal phase-contrast-venogram of the corresponding vein. The dynamic results show that signal can be retained for a long duration after tagging, allowing sufficient time for blood coming from capillaries farther in the surrounding tissue to drain into the imaged vein (Figure 5d).

Conclusions

We validate DiSpect for venous imaging with a flow phantom experiment and a comparison with a QSM-venogram. We demonstrate the possibility to image two slices in an interleaved manner to image draining of multiple vein groups. Finally, we show that using a conformal coil with high surface SNR, we can observe venous territories across large time scales.

Acknowledgements

We thank the following funding sources: R01MH127104, UCB-UCSF VCR Faculty Collaboration Award and GE Healthcare.