Reliability of Three Dimentional Pseudo-continuous Arterial Spin Labeling: A Volumetric Cerebral Perfusion Imaging with Different Post-labeling Time and Functional State in Health Adults

Mengqi Liu, Zhiye Chen, Lin Ma＊

1Department of Radiology, Hainan Branch of Chinese People’s Liberation Army General Hospital, Sanya 572013, China

2Department of Radiology, Chinese People’s Liberation Army General Hospital, Beijing 100853, China

ARTERIAL spin labeling (ASL) is a non-invasive cerebral perfusion imaging technique to be used to measure cerebral blood flow(CBF) in vivo without exogenous tracers.1-3In clinical practice, ASL sequence is used to diagnose and monitor stroke,4brain tumor,5and cognitive disorders.6Initially, flow-sensitive alternating inversion recovery (FAIR) technique is applied in the functional imaging to measure CBF.7This approach could label the water molecules of flowing blood as a tracer to obtain the CBF information. However, this technique suffered from low signal intensity, limited spatial coverage of the brain and rapid T1 decay of the labeled spins.8

Three dimensional (3D) spiral fast spin echo (FSE)pseudo-continuous ASL (3D pc-ASL) is a novel non-enhancement perfusion sequence on MR750 3.0T. The advantage of this technique includes 3D acquisition,spiral k-space filling, FSE pulse sequence, which would further expand the clinical application of ASL. Therefore, the assessment of the reliability of 3D pc-ASL seems important before the large scale application.

Previous studies9-10demonstrated that continuous and pulsed ASL had a good test-retest reliability of CBF on 1.5T scanners. And compared with15O-water positron emission tomography (PET) in Alzheimer’s disease, 3D pc-ASL provided a reliable whole brain CBF measurement in young and elderly adult on 3.0T scanner.6A reliability study of pseudo-continuous ASL was also performed at 1.5T and 3.0T, suggesting the fluctuations in perfusion signal seen over the longer term at both 1.5T and 3.0T are likely to reflect genuine fluctuations in resting-state perfusion, and the physiological contributions to the variability of the regional ASL perfusion signal should be furtherly clarified.11

Although some reliability studies focused on the different short term inter-scan interval12-14and different scanners,15the reliability of 3D ASL with different post-labeling was not investigated up to now, which would be helpful for the best choice for the different post-labeling time. Besides, it was not assessed that whether the different cerebral functional state may influence the reliability of 3D pc-ASL. In the current study, we hypothesize that a good reliability of 3D pc-ASL could be confirmed with different post-labeling time and functional state. To address this hypothesis,we performed the study as follows: (1) investigate the reliability of 3D pc-ASL with different post-labeling delay time (PLD) at the resting state and right finger taping state over one week interval; (2) clarify the signal fluctuation by different PLD contributions to the test-retest reliability changes of CBF over the whole brain.

SUBjECTS AND METHODS

Subjects

Eight health adults (6 men and 2 women) were recruited from our medical school in April 2016, with a mean age of 23.8 years (ranging from 21 to 33 years). All the subjects were right-handed and highly educated.The exclusion criteria included: cranium trauma, central nervous system inflammatory disease, and use of psychoactive drugs or hormone. All the subjects were scanned twice at the same time each day for one week interval, and none was permitted to do heavy exercise and has caffeinated beverages within one hour of scanning session. Written informed consent was obtained from all subjects and the study was approved by the ethics committee of the local institution.

MR imaging

All the MR data were acquired on a DISCOVERY MR750 3.0T MR system (GE Healthcare, Milwaukee, WI, USA),with a conventional eight-channel phased array head coil. First, fast fluid-attenuated inversion recovery(FLAIR) images with repetition time (TR)/echo time(TE)/inversion time (TI) = 8802 ms/124.3 ms/2200 ms, slice thickness = 4 mm, gap = 1 mm, matrix =256×256, field of view (FOV) = 24 cm×24 cm, and number of acquisition (NEX) = 1 were obtained for general assessment. The structural imaged data were acquired with a high resolution 3D T1-weighted fast spoiled gradient recalled echo (3D T1-FSPGR) sequence, which was used to generate the 244 contiguous axial slices with parameters as follows: TR/TE =8.6 ms/3.5 ms, flip angle = 12°, FOV = 22 cm×22 cm,matrix = 256×256, slice thickness = 1.2 mm, and NEX= 1. Volumetric perfusion imaging was obtained using a pc-ASL tagging scheme with a 3D interleaved spiral FSE readout (3D spiral FSE ASL) with parameters of TR/TE = 5128 ms/15.9 ms, flip angle = 111°, FOV =20 cm×20 cm, x, y matrix = 1024×8 (spiral acquisition), and slice thickness = 3.0 mm. The labeling duration was 1.5 seconds, and PLD was 1.5 seconds and 2 seconds respectively. The first ASL data acquisition was performed with PLD 1.5 seconds, and the second ASL data acquisition with PLD 2 seconds in the uncontrolled resting state for all the subjects, and the third ASL data acquisition was performed with PLD 1.5 seconds with the right finger taping for the subjects. All the subjects were advised to take a finger-thumb taping exercises before MRI scanning, which include repeated, self-paced, rapid (2 Hz or greater), rhythmic taping of the right thumb with the other fingers respectively.16The scan protocols were identical at baseline and follow-up for all subjects.

CBF image processing

All MR structural and ASL data were processed by using Statistical Parametric Mapping 12 (SPM 12) running under MATLAB 7.6 (The Mathworks, Natick, MA, USA)and Advantage Windows workstation (Functool, General Electric, Milwaukee, USA).

ASL data including perfusion weighted images and proton density-weighted images were processed and 50 axial CBF images were acquired based on the following equation according to the reported literatures:17-18

f, flow; λ=0.9 (brain–blood partition coefficient);α=0.85 (labeling efficiency); T1b=1.6 seconds (the T1 value of blood); T1g=1.2 seconds ［the T1 value of gray matter (GM)］; τ=1.5 seconds (labeling duration); Scon,Slbeand Sref, the signal of control, label and reference images, respectively; tsat=2 seconds (the saturation time for proton density images); w, post-labeling delay.

CBF quantification

CBF GM and white matter (WM) were automatically quantified as follows: (1) All the T1-weighted images(raw T1) were checked visually for artifacts and realigned manually according to the anterior commissural-posterior commissural (AC-PC) line to generate the co-registered T1 images (co T1); (2) The individual co T1 images were segmented by Dartel method19to generate the normalized GM and WM images; (3) The individual normalized matrix (subject_id_seg8.mat)generated by segment was applied to the raw CBF map and then generated normalized CBF map; (4) The individual normalized GM and WM were overlapped on the normalized CBF map to obtain the CBF value by the rest tool.20(Fig. 1)

Statistical analysis

The reliability were evaluated using intraclass correlation coefficient (ICC) and Bland-Altman plot for the CBF variance of the GM and WM at the different PLD over one week interval and at different PLD during the same scanning session. ICC was commonly used to assess the reliability of quantitative measurement with different observer or methods measuring the same quantity, which operated on grouped observation while not paired observations. The analysis of variance(ANOVA) method was used to calculate the ICC value.The common quoted guidelines for interpretation for ICC agreement measures were listed as following: ＜0.40, poor; 0.40-0.59, fair; 0.60-0.74, good; 0.75-1.00, excellent.21Bland-Altman plot was used to an-alyze the consistency between two measurements of the same parameter in the form of plot. The combination of ICC and Bland-Altman would provide a good assessment for the reliability. The statistical analysis was performed by using SPSS 19.0 and MedCalc (V11.4.2.0,https://www.medcalc.org/index.php).

Figure 1. Schematic of measurement of CBF value of GM and WM. CBF: cerebral blood flow; GM: gray matter; WM: white matter; raw T1: the raw T1 weighted images (structural images); co T1: co-registered T1 weighted images.

RESULTS

Reliability of 3D pc-ASL in CBF measurement at the different PLD, control state and scanning session

Table 1 showed that in the resting state, ICC of GM at PLD 1.5 seconds (0.84) was lower than that of WM(0.92), and ICC of GM (0.88) at PLD 2.0 seconds was also lower than that of WM (0.94), in addition it could be noticed that ICC of both GM and WM at PLD 1.5 seconds was lower than that at PLD 2.0 seconds in the resting state. After being exerted the right finger taping, ICC of GM (0.88) showed a slightly increase compared with that (0.84) in the resting state at PLD 1.5 seconds, and ICC of WM showed no changes.

The Fig. 2A showed a positive correlation between the CBF value difference and average CBF value of GM at PLD 1.5 seconds, while there was no correlation between the CBF value difference and average CBF value of WM at PLD 1.5 seconds (Fig. 2B). The CBF value difference had rarely correlation with averageCBF value of GM and WM at PLD 2.0 seconds (Fig. 2C and 2D). Fig. 2E indicated that the CBF difference of GM was negatively correlated with average CBF value,and there was no evident correlation between the CBF difference and average CBF value of WM at PLD 1.5 seconds with right finger taping (Fig. 2F).

Table 1. Test-retest reliability of CBF values at different PLD in the resting state and right finger taping state for the brain over one week interval in 8 healthy volunteers§

Reliability of 3D pc-ASL in CBF measurement at different PLD in the same scanning session

Table 2 demonstrated that in the resting state ICC of GM and WM were 0.71 and 0.78 for PLD 1.5 seconds and PLD 2.0 seconds in the first scan, and ICC of GM and WM were 0.83 and 0.79 in the second scan. Fig. 3 indicated that a positive correlation between the CBF difference and average CBF of GM or WM.

DISCUSSION

Our results demonstrated that 3D pc-ASL had a reliable CBF measurement for different brain tissues with different PLD over one week interval at 3.0T scanner, which suggested that this sequence could reflect the intrinsic CBF state. ICC of GM was lower than that of WM at the same PLD, which indicated that CBF of WM was relatively reliable and stable at the same PLD,and could be considered as a reference value in the brain perfusion study.

The CBF variance could be influenced by random noise and physiological noise. Although higher field strength could increase the perfusion signal intensity and reduce the variability arising from random noise, it would increase the physiological noise. Table 1 demonstrated that ICC of GM and WM was higher at PLD 2.0 seconds than that at PLD 1.5 seconds, which may be associated with the labeling efficiency at different PLD. It was reasonable to speculate that short PLD might increase the physiological noise, which would decrease the reliability of the CBF value for GM and WM at high field strength.Therefore, the detailed mechanisms should be investigated further.

Figure 2. Bland and Altman plot of CBF difference of GM (A, C, E) and WM (B, D, F) for the normal subjects at different PLD and control state for the brain over one week interval. A and B. PLD 1.5 seconds in the resting state; C and D. PLD 2.0 seconds in the resting state; E and F. PLD 1.5 seconds in right finger taping state. circle: the subject; purple dotted line:regression line of difference; light yellow dotted line: 95% confidence interval.

Table 2. Reliability of CBF values of GM and WM acquired at the different PLD during the same scanning session in 8 healthy volunteers§

Previous study demonstrated that regional CBF was more reliable when measured on separate days with pseudo-continuous ASL,22and the high reliability was also confirmed in Alzheimer’s disease.6The current study demonstrated that the high reliability of CBF measurement with 3D pc-ASL in young healthy adults based on the measurements of whole brain. Besides,we also observed that controlled state, such as motion,could influence the reliability of CBF measurements with different PLD for different brain tissues. It has been demonstrated that 3D PC-ASL had a higher reliability for CBF measurements of WM with the different PLD under rest state, while it had a higher reliability to detect CFB of GM with the different PLD at control state (such as motion), which may be associated with the contribution of physiological noise to reliability of CBF measurements. Therefore, it should be cautious as explaining CBF value with different control state.

Figure 3. Bland and Altman plot of CBF difference of GM (A) and WM (B) for the normal subjects in the resting state at different PLD during the same scanning session. circle: the subject; purple dotted line: regression line of difference; light yellow dotted line: 95% confidence interval.

Bland and Altman plot analysis showed a positive correlation of CBF value difference with average CBF value of GM, while there was no correlation of CBF value difference with average CBF value of WM at PLD 1.5 seconds, which suggested that the reliability of CBF measurement for GM was more easily affected by the perfusions state. It should be careful to explain the results when the lesions in the GM had a higher CBF value, since it might be related to the hypervascularity of GM.23Fig. 2 also presented that there was no significant correlation between CBF value difference and average CBF value of GM and WM at PLD 2.0 seconds in the resting state, which also suggested that the CBF measurements for GM and WM were reliable when PLD was set at 2.0 seconds. However, the variance of CBF value of GM presented negative correlation with the average CBF of GM in right finger taping state, which indicated that the CBF value of GM was influenced by the control state (motion state). This study demonstrated that the CBF value of WM measured with 3D pc-ASL was reliable in finger taping state. The detailed mechanisms of difference between the CBF value of GM and WM in control state should be investigated further.

There were some limitations in our study.First, the sample size was relatively small. Second,more controlled state (cognitive state, visual stimulation, etc) should be applied. Third, the subjects were all young adults, and elderly adults should be included in the future study. Last, the reliability should also be evaluated in different disease entities because the CBF could be affected by different disease state.

In conclusion, 3D pc-ASL offered a good reliability for CBF measurement over the whole brain at different PLD in the resting state or controlled state.

Conflicts of interest statement

The authors have no conflicts of interest to disclose.

1. Gai ND, Chou YY, Pham D, et al. Reduced distortion artifact whole brain CBF mapping using blip-reversed non-segmented 3D echo planar imaging with pseudo-continuous arterial spin labeling. Magn Reson Imaging 2017; 44:119-24. doi: 10.1016/j.mri.2017.08.011.

2. Inoue Y, Tanaka Y, Hata H, et al. Arterial spin-labeling evaluation of cerebrovascular reactivity to acetazolamide in healthy subjects. AJNR Am J Neuroradiol 2014; 35(6):1111-6. doi: 10.3174/ajnr.A3815.

3. Amann M, Achtnichts L, Hirsch JG, et al. 3D GRASE arterial spin labelling reveals an inverse correlation of cortical perfusion with the white matter lesion volume in MS. Mult Scler 2012;18(11):1570-6. doi: 10.1177/1352458512441984.

4. Shinohara Y, Kato A, Kuya K, et al. Perfusion MR imaging using a 3D pulsed continuous arterial spin-labeling method for acute cerebral infarction classified as branch atheromatous disease involving the lenticulostriate artery territory. AJNR Am J Neuroradiol 2017;38(8):1550-4. doi: 10.3174/ajnr.A5247.

5. Ozsunar Y, Mullins ME, Kwong K, et al. Glioma recurrence versus radiation necrosis? A pilot comparison of arterial spin-labeled, dynamic susceptibility contrast enhanced MRI, and FDG-PET imaging. Acad Radiol 2010; 17(3):282-90. doi: 10.1016/j.acra.2009.10.024.

6. Xu G, Rowley HA, Wu G, et al. Reliability and precision of pseudo-continuous arterial spin labeling perfusion MRI on 3.0 T and comparison with15O-water PET in elderly subjects at risk for Alzheimer’s disease. NMR Biomed 2010; 23(3):286-93. doi: 10.1002/nbm.1462.

7. Gutjahr FT, Gunster SM, Kampf T, et al. MRI-based quantification of renal perfusion in mice: improving sensitivity and stability in FAIR ASL. Z Med Phys 2017;27(4):334-9. doi: 10.1016/j.zemedi.2017.02.001.

8. Campbell AM, Beaulieu C. Comparison of multislice and single-slice acquisitions for pulsed arterial spin labeling measurements of cerebral perfusion. Magn Reson Imaging 2006; 24(7):869-76. doi: 10.1016/j.mri.2006.03.011.

9. Floyd TF, Ratcliffe SJ, Wang J, et al. Precision of the CASL-perfusion MRI technique for the measurement of cerebral blood flow in whole brain and vascular territories. J Magn Reson Imaging 2003; 18(6):649-55.doi: 10.1002/jmri.10416.

10. Jahng GH, Song E, Zhu XP, et al. Human brain: reliability and reproducibility of pulsed arterial spin-labeling perfusion MR imaging. Radiology 2005;234(3):909-16. doi: 10.1148/radiol.2343031499.

11. O’Gorman RL, Coward HJ, Zelaya FO, et al. Reproducibility of pseudo-continuous ASL at 1.5T and 3T. Proc Intl Soc Mag Reson Med 2007; 15:1419.

12. Parkes LM, Rashid W, Chard DT, et al. Normal cerebral perfusion measurements using arterial spin labeling:reproducibility, stability, and age and gender effects.Magn Reson Med 2004; 51(4):736-43. doi: 10.1002/mrm.20023.

13. Yen YF, Field AS, Martin EM, et al. Test-retest reproducibility of quantitative CBF measurements using FAIR perfusion MRI and acetazolamide challenge.Magn Reson Med 2002; 47(4):921-8. doi: 10.1002/mrm.10140.

14. Jiang L, Kim M, Chodkowski B, et al. Reliability and reproducibility of perfusion MRI in cognitively normal subjects. Magn Reson Imaging 2010;28(9):1283-9.doi: 10.1016/j.mri.2010.05.002.

15. Huang D, Wu B, Shi K, et al. Reliability of three-dimensional pseudo-continuous arterial spin labeling MR imaging for measuring visual cortex perfusion on two 3T scanners. PLoS One 2013; 8(11):e79471. doi:10.1371/journal.pone.0079471.

16. Buijink AW, Broersma M, van der Stouwe AM, et al.Rhythmic finger tapping reveals cerebellar dysfunction in essential tremor. Parkinsonism Relat Disord 2015;21(4):383-8. doi: 10.1016/j.parkreldis.2015.02.003.

17. Jarnum H, Steffensen EG, Knutsson L, et al. Perfusion MRI of brain tumours: a comparative study of pseudo-continuous arterial spin labelling and dynamic susceptibility contrast imaging. Neuroradiology 2010;52(4):307-17. doi: 10.1007/s00234-009-0616-6.

18. Dai W, Garcia D, de Bazelaire C, et al. Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields. Magn Reson Med 2008; 60(6):1488-97. doi: 10.1002/mrm.21790.

19. Chen Z, Li L, Sun J, et al. Mapping the brain in typeⅡ diabetes: voxel-based morphometry using DARTEL. Eur J Radiol 2012; 81(8):1870-6. doi: 10.1016/j.ejrad.2011.04.025.

20. Song XW, Dong ZY, Long XY, et al. REST: a toolkit for resting-state functional magnetic resonance imaging data processing. PLoS One 2011; 6(9):e25031. doi:10.1371/journal.pone.0025031.

21. Hallgren KA. Computing Inter-rater reliability for observational data: an overview and tutorial. Tutor Quant Methods Psychol 2012; 8(1):23-34.

22. Pfefferbaum A, Chanraud S, Pitel AL, et al. Volumetric cerebral perfusion imaging in healthy adults: regional distribution, laterality, and repeatability of pulsed continuous arterial spin labeling (PCASL). Psychiatry Res 2010; 182(3):266-73. doi: 10.1016/j.pscychresns.2010.02.010.

23. Kavanagh EC. The reversal sign. Radiology 2007;245(3):914-5. doi: 10.1148/radiol.2453050112.