Theoretical investigation and optimization of the labeling process in continuous artery-selective spin labeling (CASSL)

Introduction Continuous artery-selective spin labeling (CASSL) [1] is one technique to image perfusion territories of cerebral arteries. To date, CASSL was only applied to major brain feeding arteries, but, recently, empirical findings have demonstrated that this method is even capable of selectively label the blood of individual branching intracranial arteries [2]. However, the exact behavior of spins moving through the labeling plane multiple times in a non-selected vessel may only be determined by computer simulations. In this study we performed extensive simulations to better understand the underlying labeling process to optimize key labeling parameters and demonstrate prospects and constraints of this method to selectively label the blood of individual branching intracranial arteries in-vivo. Material and Methods In CASSL selective labeling is obtained by a rotating labeling plane in conjunction with a frequency modulation of the labeling RF pulses that fixes the position of the labeling plane to the selected artery. A saturation of the magnetization in non-selected vessels depends basically on the angle θ between selected vessel and rotating labeling plane, its rotation frequency frot and the distance d from the labeling focus, respectively. The locus of resonance at a distance d to the labeling spot will vary in time and is given by [1] as S(t) = d· tanθ·sin(2π frot t+φ) with φ as an arbitrary phase. It is hypothesized that this will lead to a pseudo-random behavior when a spin is moving through the labeling plane multiple times and that this will on average cause a saturation of the magnetization in a given voxel. The ratio of the maximum velocity of the labeling plane vmax, plane and the velocity of the blood flow vblood determines how many times the blood will move through the labeling plane and therefore how often the blood’s magnetization experience a certain change in orientation. A computer program written in Matlab (The Mathworks) was used to optimize the labeling parameters in terms of spatial selectivity and labeling efficiency. The program calculates the magnetization of flowing spins after inversion by adiabatic fast passage. Orientation and strength of the labeling gradient as well as the amplitude of the RF pulse were chosen according to a previous study with a fixed labeling plane [3]. T1 is set to 1000ms, T2 to 200ms to represent human blood values at 1.5T. The simulation is based on a stepwise integration of the Bloch equations for a single magnetization vector using a 4 order Runge-Kutta algorithm. The magnetization starts with a value of M=[0,0,1] and travels with a fixed velocity along a linear trajectory. Labeling efficiency was computed as