With wedge-shaped slice-selective inversion pulses ¹, pulsed arterial spin labeling (PASL) has the potential to place all delivered blood in well-controlled label or control states (full duty cycle (FDC) labeling), using short labeling pulses. In this study, we explore the potential for FDC-ASL to measure blood flow (BF) and transit delay (TD) with improved sensitivity.

An FDC-ASL signal model was generated based on a kinetic ASL model ² assuming that: 1) the labeling duty cycle is 1, that is, the bolus duration (BD) created by each labeling pulse equals the TR; 2) the delivered blood magnetization stays in the voxel and evolves only due to T₁ relaxation and excitation pulses; 3) BF and TD are constant. The numbers of consecutive TRs in the labeled state (nL) can exceed one, and are followed by the same number of TRs in the control state, corresponding to a fundamental driving frequency (fDF) of 1/(2·TR·nL). This approach is similar to Dynamic ASL (DASL) ³, except that here the labeling is pulsed and the TR equals BD, so the acquisition occurs on the same phase of the signal. The response at the fDFs can be obtained from the FT of the signal time course, where the amplitude and the phase provide information on the magnitude of the ASL signal and the TD, respectively. The maximal TD the method can detect without aliasing equals a full cycle of the labeling scheme, i.e., 2·nL·TR.

Simulation without noise: The FDC-ASL signal was simulated with: BD=TR=0.5s, flip angle=45°, tissue signal=5000 (a.u.), BF=100 (a.u.), TI=0s, TD stepping from 0 to 2*nL*TR with a step size of 1ms for nL=1, 2 and 3. 44 TRs were simulated. The first 20 points were discarded and an FT was performed after removing the mean. The amplitude and the phase at the fDFs (1Hz, 0.5Hz and 0.33Hz for nL=1, 2 and 3, respectively) were calculated.

Simulation with Gaussian noise: The above simulations were performed for 260 TRs for FDC-ASL with nL=2 and 3, and BF=1 (a.u.). The amplitude and phase at the fDFs were calculated from the last 240 TRs to form a dictionary. Gaussian noise of four different levels (corresponding to SNRs of 5, 2, 1 and 0.5, respectively, with respect to the ASL signal from a pair of conventional PASL measurements with CBF=100, BD=0.7s and TI=1.8s) was added to the simulated signal time courses with TD=0.8s and CBF=100. TD and BF were recovered from the dictionary. A thousand Monte Carlo simulations were performed at each SNR level for FDC-ASL with nL=2 and 3, conventional PASL (BD=0.7s, TI=1.8s, TR=3s, 40 TRs) and pseudo-continuous ASL (PCASL) ⁴ (labeling duration=1.7s, post-labeling delay=1.8s, TR=4s, 30 TRs). The mean and the standard deviation (STD) of the recovered BF and TD values were calculated. Labeling efficiencies of 0.98 and 0.85 for PASL and PCASL, and T₁ of 1.65s for blood ⁵ were assumed in the simulations.

The simulated signal and the frequency responses with TD=0 are shown in Fig. 1. The sensitivity of the amplitude and the phase at the fDFs to TD are shown in Fig. 2. With nL=1, no phase information can be retrieved due to the critical sampling rate; the measurement became insensitive to BF at some TDs. With nL=2, the sampling rate allowed for estimation of the phase and the TD; and the sensitivity to BF was high across the TD range. With nL=3, the sensitivity to blood flow and the TD detection range were further improved. The mean and STD of the recovered BFs and TDs from the simulation with noise are shown in Fig. 3. Across the four SNR levels, the maximal normalized biases and the STDs (a.u.) of BF estimates were -2.2%, -1.1%, 1.1% and 0.7%, and 43.2, 32.4, 38.9 and 28.4 using conventional PASL, PCASL, UF-ASL with nL=2 and UF-ASL with nL=3, respectively; the maximal normalized biases and the STDs of TD estimates were -0.7% and -1.3%, and 0.15s and 0.13s with FDC-ASL with nL=2 and 3, respectively.The temporal resolution of the FDC-ASL can be high, with an estimate of BF and TD every 2·nL·TR, e.g., every 2s with nL=2 and BD=TR=0.5s. Because the labeling pulses are short, high imaging duty cycle can be supported, unlike for CASL/PCASL. For dynamic changes in BF and/or TD, a signal model that corrects for dynamic changes in BD will be required, as in ⁶.FDC-ASL showed potential of measuring BF and TD with improved sensitivity and efficiency.