Does cardiac triggering improve pCASL signal stability? Isolation of the effect of the last labeled spins by end-of-labeling triggering and extremely long labeling durations
Jasper Verbree1 and Matthias J.P. van Osch1

1Radiology Department; Leiden Institute for Brain and Cognition; C.J. Gorter Center for High-field MRI, Leiden University Medical Center, Leiden, Netherlands

Synopsis

In pCASL, the blood tagged at the end-of-labeling period is expected to contribute most to the perfusion signal due to T1 recovery of the labeled spins. The influence on PCASL of cardiac triggering at the end-of-labeling was assessed with simulations and subsequently applied in volunteers. Simulations predict a 9% variation in ASL-signal over the cardiac cycle. In-vivo measurements were unable to show the predicted effect nor a difference in tSNR. Combining with earlier findings concerning cardiac triggering, neither triggering start- or end of labeling triggering improves signal stability, suggesting that cardiac triggering is not beneficial for pCASL.

Background

Introducing cardiac triggering at the start of every measurement has been shown to reduce signal variability in pulsed ASL[1,2], but not pCASL[2]. Due to T1 recovery of the labeled spins in pCASL, the blood tagged close to the end of the labeling period is expected to contribute most to the perfusion signal and will therefore be the most significant source of signal variations. The goal of the current study is to investigate the influence of the cardiac cycle on these spins labeled last. To this end, we first assessed the contribution of the end-of-labeling period using simulations. Subsequently, prospective end-of-labeling triggering was implemented and tested in human volunteers. By increasing the labeling duration drastically (to an impractical value of 7500 ms), the effect of cardiac cycle on the earliest labeled spins was minimized.

Simulation methods

The inflow of label into tissue was simulated in MATLAB as a single vessel. For each (1ms) iteration fresh blood entered the artery, thereby shifting the entire blood pool toward the tissue-compartment. Subsequently, T1-relaxation was accounted for in =the artery (T1=1664ms) and tissue (T1=1200ms). The ASL signal at a fixed post labeling delay was assessed by integrating over the tissue-compartment. A simulated pulsatile flow (Figure 1) was generated with a mean heart rate of 50 bpm and constant stroke volume.

Simulations results

The simulated ASL signal over time followed the standard Buxton inflow-decay curve, with flow pulsations superimposed (Figure 1a). Ending the labeling directly after the trigger (ELdelay of 0ms) corresponded with a peak in the Buxton curve (Figure 2a). Increasing the ELdelay to 500ms decreased the ASL signal by 9% (Figure 2b). Further increases of the ELdelay resulted in recovery of the ASL signal to the starting level. In summary, simulations show a potential 9% effect of the cardiac cycle on the ASL-signal.

Implementation details

The start-of-labeling (SL) was always triggered, whereas the time of the end-of-labeling (EL) was varied relative to the finger pulse oximeter signal (PPU-trigger) by ending labelling after a certain delay (ELdelay). To isolate the effect of EL, care was taken that the average labeling duration (7500ms) was similar for all ELs as well as for the untriggered ASL-scan. For such long labeling durations, small variations in labeling duration have only a marginal influence on the ASL-signal (0.5% increase for 500ms longer labeling, assuming a T1 of 1664ms).

In-vivo MRI methods

Data from six volunteers (3 female, 25-56 years) were acquired on a Philips 3T MRI. The blood velocity through the labeling plane was assessed using a retrospectively triggered phase-contrast scan with 15 cardiac phases (venc=200cm/s). The vendor supplied pCASL sequence was modified to incorporate triggering of start- and end-of-labeling (ASLtrig). Three distinct cardiac phases (a), (b) and (c) were determined for each subject according to the flow data (Figure 1). The control scans were triggered at start-labeling and had the same labeling duration as the preceding label acquisition. The triggered scan was acquired continuously, cycling through the cardiac phases (a-b-c-a-b-c….) resulting in 20 label and control images for each cardiac phase.

A non-triggered reference ASL scan (NTASL) was acquired with the same imaging parameters. For quantification purposes an M0 scan was also acquired. ASL-images were quantified [3], aligned to the first dynamic, and scaled using the M0 scan. Average gray matter ASL-signal was calculated for each dynamic. All analyses and statistics were performed in MATLAB with the SPM8 toolbox.

In-vivo results

Heart rate (p=0.26) and mean labeling duration (p=0.47) did not differ significantly between scans. Mean CBF was higher in NTASL (p=0.035) compared to ASLtrig of cardiac phase (b). The temporal standard deviation was comparable between the triggered and non-triggered scans (p=0.44) (Figure 4). The moment of imaging with respect to the cardiac cycle varied equally for triggered (23%) and non-triggered (27%) scans.

Conclusion

Simulations predict a 9% variation in ASL-signal over the cardiac cycle when triggering the end-of-labeling, with maximum signal at end-systole. In-vivo measurements were unable to show either the predicted effect or a difference in tSNR. It should be noted that using end-of-label-triggering with normal labeling durations (≈1800ms) would result in a variation in labeling duration that would need to be accounted for in quantification. Combined with earlier findings concerning cardiac triggering, neither start-[2] nor end of labeling triggering improves signal stability, suggesting that cardiac triggering is not beneficial for pCASL.

Scan Parameters

ASLtrig: minimum labeling duration=7000ms; post labeling delay=1800; TE=14 ms; flip angle=90°; FOV=240x240x119mm3; matrix=80x79; slice thickness=7mm; number of dynamics=60 (20 per condition); EPI factor=35; NTASL: same as triggered, labeling duration=7500; dynamics=20; M0: TR/TE/TE2=2000/11/14ms; FOV=220x220x102mm3; matrix=72x69; slice thickness=6 mm; number of averages=10; EPI factor=25;

Acknowledgements

The authors would like to thank ir. J. Wezel for help with optimizing the simulations.

References

[1] Fushimi, et al. NMR biomed. 2013; (jan) [2] Wu et al. IEEE TMI. 2007;26(1): 84-92 [3] Heijtel et al. MRM. 2014;92(C):182-192

Figures

Figure 1: Total flow though brain feeding arteries with the three cardiac phases indicated at which points the labeling was stopped during in vivo measurements.

Figure 2a: Simulated ASL signals following the Buxton curve over time. Pulsations during the inflow period come from the simulated flow.

Figure 2b: Zoom of figure 2a showing two signals, simulated with end-of -labeling delay of 0 ms (blue) and 500ms (red). Note that the inflow of label was stopped at either the peak or the trough, resulting in different level of ASL-signal at imaging (arrows).

Figure 3: whole brain gray matter perfusion of all six subjects during the three cardiac phases and non-triggered ASL (NT).

Figure 4: mean voxel-wise temporal standard deviation over of all six subjects.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1498