Introduction:

4D angiography was shown to be feasible at 7T using pulsed arterial spin labeling (PASL)^1,2. With pseudo-continuous arterial spin labeling (pcASL) a higher arterial signal was demonstrated at 1.5T compared to PASL³. However, at 7T, pcASL suffers from specific absorption rate (SAR) constraints, low B₁⁺ efficiency for the labeling and B₁⁺ inhomogeneity in the readout. This might lead to long acquisition times and poor contrast in the angiography images. To counteract high SAR values, Variable-Rate Selective Excitation (VERSE) pulses for the labeling scheme show a significant impact⁴. The efficiency of B₁⁺ in the labeling could be improved by phase-only shimming⁵ as well as B₁⁺ phase/amplitude shimming⁴. To overcome inhomogeneity in slice-selective small flip angle gradient echo sequences, multiple spokes are widely used⁶. In this work, we combine a B₁⁺ phase shim trading B₁⁺ homogeneity and transmit efficiency applied to VERSEd pcASL labeling pulses as well as 2-spoke excitation for the readout in a 4D angiography sequence at 7T. This method was compared to the standard circular polarized (CP) mode for labeling and readout.

Methods:

All measurements were performed on a 7T MR system (MAGNETOM Terra, Siemens Healthcare GmbH, Erlangen, Germany) using a 32-channel Rx/8Tx head-coil (Nova Medical, Wilmington, Massachusetts, USA) and carried out in accordance with the institutional guidelines and with approval of the local Ethics Committee (Friedrich-Alexander University (FAU) Erlangen-Nürnberg, Germany). An unbalanced pCASL labeling scheme (T_Pulse=500μs, T_Gap=1200μs, G_max=4.2mT/m, G_ave=0.6mT/m, labeling duration=300ms) was modified with VERSE pulses⁷ (50% amplitude reduction, G_VERSE,max=10mT/m). For the readout a segmented spoiled turbo flash (TFL) readout was used (isotropic resolution 1mm³, 52 slices, PAT=3, partial fourier=6/8, TE=2.07ms, nominal flip angle=8°, echo spacing=4.7ms, TR=1600ms). The post labeling delay (PLD) was varied from 100:150:550ms. After each labeling 31 segments of the readout train were applied, resulting in a temporal resolution of 145.7ms and a total acquisition time of 4:40min per time-point. For the parallel transmission (pTx) pulse calculations, relative B₁⁺ maps⁸ as well as B₀ maps were acquired. Additional time-of-flight (TOF) images were achieved at the labeling plane for the hybrid B₁⁺ shim and the shim values were calculated by minimizing the cost function $$$cost=0.3\cdot CoV^{2} + 0.7\cdot\eta^{2}$$$, with $$$CoV=\frac{std\left(\left|\sum_{c=1}^C B_{1,c}^{+}(r)\right|\right)}{mean\left(\left|\sum_{c=1}^C B_{1,c}^{+}(r)\right|\right)}$$$ and $$$\eta = mean\left(\frac{\left|\sum_{c=1}^C B_{1,c}^{+}(r)\right|}{\sum_{c=1}^C \left|B_{1,c}^{+}(r)\right|}\right)$$$ (compare Fig.1a). The readout pTx pulses consists of two 0.9ms long Sinc-shaped spokes placed by an iterative fourier spoke placement⁹. The complex weights for each spoke pulse were calculated by minimizing $$$cost=\parallel\left|Ab\right| - \left|m\right|\parallel^{2} + \lambda\parallel b \parallel^{2}$$$ with the desired magnetization pattern m, b the complex weights of each RF channel and spoke in Volt, A the system matrix containing the excitation trajectory as well as B₁⁺, B₀ maps, and the regularization factor for the RF energy $$$\parallel b \parallel^{2}$$$. A magnitude-least-squares (MLS) algorithm was used for the optimization¹⁰ (compare Fig.1b). Additionally, a 1mm³ isotropic TOF sequence was obtained at the same position (4 slabs à 20 slices, PAT=2, TE=3.57ms, flip angle=17°, TR=19ms). In total 3 subjects (mean age: 26.7±0.5years) were measured with the hybrid B₁⁺ shim for the labeling and the 2-spokes excitation for the readout (referred as 'full pTx') compared to the sequence in the CP-mode. The 4D-pcASL images underwent postprocessing using SPM12 (Wellcome Trust Centre for Neuroimaging)¹¹ and the FMRIB Software Library (FSL)¹². To evaluate the data transversal maximum intensity projections (MIPs) of the 4D-pcASL as well as of the TOF images were created and filtered by a 2D-Frangi filter (Gaussian kernel σ=1.0, thresholds to control sensitivity α=0.4, β=0.3 and c=40). The 4D-pcASL filtered MIPs were co-registered to the TOF and masked by the filtered MIPs of the TOF. Those masked 4D-pcASL images were used for signal evaluation in the vessels and vessel length determination by processing via Fiji ImageJ as described in a previous work¹³ (compare Fig.2).

Results:

Fig.3a presents the signal intensity of the vessels at the different PLD time-points. Compared to the CP-mode, the full pTx sequence that includes the hybrid B₁⁺ shim for the labeling and 2-spoke pTx for readout shows higher signal in the vessels at all time-points. The measured vessel length (Fig.3b) remains constant for the first three time-points (100-400ms), however with a PLD=550ms the measured vessel length for the full pTx mode increases. An example for all time-points of one subject is given in Fig.4. There, the higher signal intensity of the full pTx sequence compared to the CP-mode can also be observed as well as an increasing number of small vessels can be depicted at later time-points (red arrows).

Discussion and Conclusion:

The combination of B₁⁺-shimming for the labeling and 2-spoke dynamic pTx for the readout in 4D-pcASL MRI sequences promises great advantage over conventional CP-mode at 7T. While the higher and more homogenous distribution of the flip angles increases the vessels’ signal intensities, the B₁⁺-shimmed labeling pulse strengthens this effect and yields even more vessel conspicuity, especially at later time-points.

Acknowledgements

No acknowledgement found.