A long post-labeling delay is recommended in ASL to reduce vascular artifacts¹. However, recent studies have suggested that vascular artifacts may still be present at such delay time and, in particular, it manifests as pulsatile signal fluctuations in regions with large blood vessels. In single-shot settings, its impact is that voxels near large vessels show considerably greater (e.g. >2 times) signal variations and this could be mitigated by cardiac-triggered acquisition². In the present study, we asked three further questions. 1) Given that 3D segmented acquisition is the preferred readout scheme in ASL, what is the impact of vascular pulsation effects in these images? 2) While ASL is a small signal to start with, it is even more challenging to detect CBF changes due to manipulations such as hypercapnia. Then, to what extent could cardiac-triggering improve data quality in these settings? 3) In imaging centers where cardiac-triggered ASL sequence is not available, is there any way to reduce the cardiac pulsation effects?

Fourteen healthy subjects (26±5yo, 6M) were scanned on a 3T (Philips). We use the ASL labeling scheme recommended in the white paper¹, which includes pseudocontinuous labeling (labeling duration/post-labeling delay 1.8/1.8s) and background-suppression. Both regular and cardiac-triggered ASL were performed. For the cardiac-triggered ASL, a pulse oximeter was attached to the finger of the subject, yielding R-peaks that serve as a trigger for each TR of the pCASL sequence.

Study 1: Cardiac-triggering in 3D segmented pCASL: In this study (N=5), we employ a 3D segmented spiral TSE acquisition³ in pCASL, with FOV 200×200×120mm³, voxel size 3×3×4mm³, 22 segments, one pair of control/label. To quantify the improvement provided by the cardiac-triggered scheme, both regular and triggered pCASL were repeated 4 times and coefficient-of-variation (CoV) map was calculated by the ratio between standard deviation of each voxel and whole-brain signal mean.

Study 2: Cardiac-triggered ASL in detecting perfusion changes due to hypercapnia: In this study (N=1), cardiac-triggered pCASL was used to evaluate the perfusion change between normocapnia (NC) and hypercapnia (HC). The subject inhaled room air for 3min, followed by HC challenge (5%CO2/21%O2) for 5min (the first 2min allowed the subject to reach new physiological steady state). Only the first and last 3-min worth of data were used to derive the ASL images in NC and HC, respectively. Regular pCASL was also conducted with the same paradigm for comparison. The acquisition parameters are: single-shot 3D GraSE readout, FOV 180×180×104mm³, 13 slices, voxel size 4.5×4.5mm², TR 4.2s.

Study 3: Benefit of recording cardiac phase even without using it for triggering: In N=8 subjects, regular (non-triggered) pCASL was performed using readout schemes described in Study 2. A pulse oximeter was also used but to record cardiac phase during the scan. A previous study² showed that both control/label signal intensities are modulated by cardiac phase with similar curve shape. We therefore propose that, rather than simply averaging/subtracting, one can plot the control/label signals as a function of cardiac phase, and fit the two data group to two identical 2^nd-order Fourier series (see Figure 3a for illustration). The gap between the two curves then corresponds to ASL signal for this particular voxel. 5 repetitions (10 control/label pairs in each repetition) were conducted to allow CoV calculation.

Figure 1 shows a comparison between regular and triggered 3D segmented pCASL images. With regular pCASL, the ASL images are contaminated by ring-shaped artifacts (e.g. yellow arrows). In the triggered data, this was not present. The CoV of the ASL image decreased significantly (p<0.01) by 22±3% when considering all subjects.

For the hypercapnia study, Figure 2 summarizes the ASL images under NC and HC conditions, their difference, and the CoV image, for both regular and triggered schemes. While the CBF difference image using the triggered scan appears similar to that of the regular scan (which is desired as one does not want any bias in the triggered sequence), the data is more stable with less bright/dark vascular artifacts, as revealed by 21% smaller CoV.

Figure 3b shows the results using the cardiac-phase-based analysis method, related to the simple subtraction method. Note that both analyses used regular, non-triggered ASL data. The only difference is that the fitting method also utilized the cardiac-phase. It can be seen that, without modifying the pulse sequence, it was possible to mitigate the signal fluctuations in the traditional subtraction approach, most apparently in large vessel regions. Over the 8 subjects, the CoV was reduced by 18±5% (p<0.01).

We have demonstrated that considerations of cardiac phase can improve ASL data quality in multiple settings.