Parallel Transmit Vessel Selective Arterial Spin Labelling: A proof of concept simulation

Background Vessel Selective arterial spin labelling (ASL) provides visualisation and quantification of the perfusion territory from a feeding set of arteries. Methods to selectively label blood include using pseudo-Continuous ASL with additional transverse gradients and phase cycling [1], CASL [4] with obliquely defined labelling planes by rotating the field of view [2] and the use of surface coils [3] to spatially confine the B1 field to one side of the neck. The latter is advantageous as the confined field does not produce any magnetisation transfer (MT) [5] effects within the brain. However, the spatial profile of the surface coil extends across the neck, partially labelling the contralateral arteries. This work aims to reduce the amount of contralateral labelling by employing parallel transmission [6] techniques to vessel selective arterial spin labelling Theory Parallel transmission independently modulates the amplitude and phase of the driving current in multiple transmit antenna elements to spatially tailor the excitation pattern. Vessel selective CASL places only two requirements on the labelling field: a B1 sufficiently high for efficient labelling of flowing spins over the arteries to be labelled, and a low B1 over other arteries so that flowing spins are unperturbed. These two conditions can be mathematically written as a minimisation problem, where the B1 field is minimised in a no label region, and constrained to a target value in the label region, and then solved as a constrained least squares optimisation problem (see equation 1). Methods A computer model of the carotid and vertebral arteries in the neck was constructed in Matlab (The Mathworks Inc., MA) using anatomic and physiological data obtained from time of flight and phase contrast MRI angiograms of a healthy 24-yearold volunteer (see figure 1). Three dimensional B1 profiles of surface coils positioned around the neck were computed by integrating the Biot-Savart equation. Elliptical regions of interest (ROIs) were drawn around the right carotid (RCA) and vertebral (RVA) arteries to indicate a ‘label’ region, and around the left carotid (LCA) and vertebral (LVA) arteries to indicate a ‘no label’ region. Using these ROIs equation 1 was solved using CVX [8]. Complex coefficients and resultant B1 fields were calculated for one, two and four circular surface coils (loop radius 22.5mm, field profiles normalised at B1=3.5μT at a depth of 5cm on axis), positioned on the neck at angles of 30, 150, 240, and 300 to the x-axis respectively. Calculated fields were then used in a numerical Bloch equation simulation [7] of CASL. Flow velocity within each artery was dictated by a pulsatile temporal waveform, and parabolic cross-sectional velocity profile, as measured from a phase contrast angiogram of the volunteer. Typical human CASL parameters were chosen: Glabel=3mT/m, and the labelling plane, zlabel was positioned 19cm proximal to isocentre (the centre of the brain). Labelling was performed for 2 seconds and relaxation times were that of blood water at 3T: T1=2000ms and T2=300ms. Results Figures 2a-c show the total B1 field from one coil, two coils and four coils, all with the same colour scale (0 is dark blue and 3.5μT is dark red). Figures 3a-c are plots of z position and longitudinal magnetisation of a single spin isochromat as it travels through each artery. All three combinations of coils are able to produce a sufficient B1 field at the RCA and RVA, reflected by a similar inversion efficiency in figures 3a-c. With one coil the B1 magnitude at the centre of each artery and inversion efficiencies respectively are 0.39μT and 5.7% for the LCA and 0.52μT and 13.35% for the LVA. Using two coils to create a null over these arteries reduces it to 0.11μT and 0.4% for the LCA, and 0.33μT and 5.55% for the LVA. With four coils a very well defined null region is created over both the LCA and LVA, resulting in negligible B1 fields and inversion efficiencies of 0.034μT and 0.05% for the LCA and 0.044μT and 0.15% for the LVA. Discussion and Conclusion Results show that selective labelling of the RCA and RVA with minimal labelling of the LCA and LVA is possible using parallel transmit methods to spatially localise the B1 field. As the number of coils increases there is an improvement in the ability to tailor the shape and depth of the null region. Using the calculated fields in a simulation of the labelling procedure in the human carotid and vertebral arteries shows that the labelling efficiency on the RCA and RVA can be maintained when compared to the single coil case, whilst contralateral labelling is reduced by up to two orders of magnitude. Parallel transmit ASL could also be used to improve the repeatability of separate coil CASL as the B1 fields produced at each artery no longer depend solely on coil positioning. Further work will be to extend the mathematical framework to include constraints for limiting power deposition, and to implement parallel transmit vessel selective arterial spin labelling in-vivo using a pair of independent transmit coils and parallel transmit system. References [1] Wong, E. C. MRM 58:1086-1091 (2007). [2] Werner, R et al. MRM 53:1006-1012(2005). [3] Mildner et al. MRM 49:791-795 (2003). [4] Williams et al. PNAS 89:212-216 (1992). [5] Hernandez-Garcia et al. NMR in Biomedicine 20:733-742 (2007). [6] Ibrahim et al. MRI 18:733-42, (2000). [7] Bittoun et al. MRM 2:113-120 (1984). [8] CVX: Matlab software for disciplined convex programming, http://cvxr.com/cvx. Equation 1: minimise ||Si . c||, i ∈ No Label Region subject to Si . c = Ti, i = Centre of Label Region |c| ≤ 1