SEARCH

SEARCH BY CITATION

Keywords:

  • arterial spin labeling;
  • acceleration selective;
  • perfusion imaging;
  • spatially nonselective labeling

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

In this study, a new arterial spin labeling (ASL) method with spatially nonselective labeling is introduced, based on the acceleration of flowing spins, which is able to image brain perfusion with minimal contamination from venous signal. This method is termed acceleration-selective ASL (AccASL) and resembles velocity-selective ASL (VSASL), with the difference that AccASL is able to discriminate between arterial and venous components in a single preparation module due to the higher acceleration on the arterial side of the microvasculature, whereas VSASL cannot make this distinction unless a second labeling module is used. A difference between AccASL and VSASL is that AccASL is mainly cerebral blood volume weighted, whereas VSASL is cerebral blood flow weighted. AccASL exploits the principles of acceleration-encoded magnetic resonance angiography by using motion-sensitizing gradients in a T2-preparation module. This method is demonstrated in healthy volunteers for a range of cutoff accelerations. Additionally, AccASL is compared with VSASL and pseudo-continuous ASL, and its feasibility in functional MRI is demonstrated. Compared with VSASL with a single labeling module, a strong and significant reduction in venous label is observed. The resulting signal-to-noise ratio is comparable to pseudo-continuous ASL and robust activation of the visual cortex is observed. Magn Reson Med 71:191–199, 2014. © 2013 Wiley Periodicals, Inc.

Arterial spin labeling (ASL) techniques can be used to quantify local tissue perfusion noninvasively [1, 2]. Conventional ASL methods tag arterial blood spins by saturation or inversion in a plane proximal to the imaging region. After a delay, the inflow time (TI), a label image is acquired. This TI is chosen approximately equal to the T1 of blood for cerebral perfusion imaging, representing a compromise between loss of the label due to longitudinal (T1) relaxation and a transport time long enough for the labeled blood to reach the microvascular bed [3]. Subsequently, the sequence is repeated without labeling, thereby obtaining the so-called control image. Subtraction of the label image from the control image results in a perfusion-weighted image, which reflects solely the magnetization of the inflowing tagged spins without contamination from static tissue. Multiple interleaved averages of the label and control sequence are required to gain sufficient signal-to-noise ratio (SNR).

>A major confound of conventional ASL techniques is that a significant amount of label is lost because of T1-relaxation in the time it takes for the blood to flow from the labeling to the imaging plane, the so-called transit time. Furthermore, in several pathological brain conditions, because of slow or collateral flow the label relaxes almost completely before reaching the microvascular bed, leading to severe underestimation of cerebral blood flow (CBF). Even in normal volunteers the variation in transit time can be up to hundreds of milliseconds across a single slice [4].

Recently, a technique called velocity-selective ASL (VSASL) was introduced, in which spins are tagged based on flow velocity rather than spatial localization [5-7]. This enables labeling within the imaging region where perfusion is measured. By generating the label much closer to the capillaries, smaller and more uniform transit delays are obtained, and therefore VSASL can provide (semi) quantitative CBF maps even under slow and collateral flow conditions [8].

The first velocity-selective labeling module labels all spins that flow faster than the cutoff velocity (VC), irrespective of whether these are located in arterial or venous blood. Using only this labeling approach (which we will refer to as “single VSASL”) the signal is fundamentally cerebral blood volume weighted. The use of a second velocity-selective labeling module just before imaging (dubbed “dual VSASL”) will exclude the venous components, because accelerating spins will be suppressed [6]. Furthermore, the VSASL signal is converted into a CBF-weighted signal. This approach suffers, however, from a reduced SNR by limiting the labeling to spins that decelerate to a velocity below VC before the second labeling module, as well as due to increased T2-relaxation and diffusion weighting (DW) because of the second labeling module.

More recently, it has been shown in magnetic resonance angiography that acceleration-encoded imaging can selectively image the arteries, without contamination from venous signal [9]. Acceleration-dependent preparation differs from a velocity-selective approach in that it does not affect the magnetization from spins flowing at a constant velocity, but dephases spins that are accelerating or decelerating [10]. In the vascular tree, the average velocity of the spins decreases from the arteries toward the capillaries, after which their average velocity increases only slightly while flowing to the veins. Pulsatility is known to be higher in the arteries than in the veins, providing a second origin for the differentiation between arterial and venous blood by using acceleration encoding. Moreover, the tortuosity of the vessels, and especially of the capillaries, is a third origin of labeling. When the vessel orientation changes with respect to the acceleration-encoding direction, there will be incomplete refocusing of the magnetization.

The aim of this study is to introduce a new ASL method, which is able to image brain perfusion without contamination from venous or cerebral spinal fluid (CSF) signal, based on the principles of acceleration-encoded magnetic resonance angiography. The use of such an acceleration-dependent preparation module for ASL is demonstrated in healthy volunteers. This new method, referred to as acceleration-selective ASL (AccASL), is examined with different cutoff acceleration encodings, obtained by varying the motion-sensitizing gradients (MSGs). In addition, AccASL is compared with VSASL and a conventional spatially selective ASL method, and the feasibility of AccASL in functional MRI (fMRI) is demonstrated.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Acceleration-Selective Labeling Module

The labeling module of AccASL resembles VSASL in the sense that it combines spatially nonselective hard 90° pulses with a pair of identical adiabatic 180° refocusing pulses, hereby correcting for phase shifts due to inhomogeneities of the magnetic field and chemical shifts [11]. MSGs placed in-between these radiofrequency pulses can be used to dephase the magnetization of flowing spins, whereas the polarity of these gradients determines whether flowing spins with constant velocity (VSASL) or accelerating (AccASL) spins are affected. A four-gradient-pulse scheme has been used to reduce eddy-current effects for VSASL [12]. The difference between the velocity-sensitive and acceleration-sensitive sequence is in the sign of the second and fourth gradient as shown in the schematic representation of the labeling modules in Figure 1. Spins with a flow velocity or acceleration above a cutoff velocity (VC) or cutoff acceleration (AC), respectively, are dephased: both parameters are defined as to correspond to a phase change of π.

image

Figure 1. Schematic overview of the labeling module with velocity-sensitizing (a) and acceleration-sensitizing gradients (b), where G is the gradient amplitude, δ is the duration of the gradient, and Δ is the time between the 90° radiofrequency pulses. During the control condition, no motion-sensitizing gradients are used.

Download figure to PowerPoint

The effective first-gradient moment (m1) of the velocity-sensitizing sequence can be calculated to be m1=GΔδ, with G the amplitude of the gradients, δ the gradient duration, and Δ the time between 90° radiofrequency pulses (Fig. 1). By varying these parameters, the cutoff velocity can be set, according to VC=π/(γm1), where γ is the gyromagnetic ratio. Penetrating arterioles exhibit a flow velocity of 1–2 cm/s in human brain [13] and have a laminar flow as their dominant flow pattern, which makes VSASL feasible [14].

The acceleration-sensitizing sequence has an effective first-gradient moment of zero, giving no velocity sensitization, but a second-gradient moment (m2) of m2=4GΔδ. The relation to the cutoff acceleration is AC=2π/(γm2) [9]. Deceleration could be considered as negative acceleration and is targeted with this sequence equally well.

It should be noted that the gradients not only encode motion but also impart a diffusion sensitivity, which becomes substantial for higher moments. This diffusion sensitivity, expressed as the b-value, is dependent on the MSG parameters and scales with the square of G and δ and linearly with Δ.

MR Experiments

A total of 12 healthy volunteers (four males and eight females, age 20–38 years) participated in this study and informed written consent was obtained from each participant. This study was part of a project for protocol development as approved by local Institutional research board.

First, the influences of AC on the signal intensity and the spatial distribution of the labeled blood were evaluated. Second, AccASL was compared with single and dual VSASL and pseudo-continuous ASL (pCASL) with respect to SNR and contamination of venous and CSF signal, the latter being introduced by DW. Finally, the feasibility of AccASL for brain activation studies was tested.

All scans were performed on a 3 T scanner (Philips Healthcare, Best, The Netherlands) using a 32-channel head coil with 17 slices acquired at a 2.75 × 2.75 × 7 mm3 resolution (multislice single-shot two-dimensional echo-planar imaging). The field-of-view was 220×220 mm2 and a sensitivity encoding (SENSE) factor of 2.5 was used. Spectral presaturation inversion recovery was performed to suppress the lipid signal.

Cutoff Acceleration

To evaluate the influence of the value of AC on the signal intensity and the appearance of the perfusion map, AccASL images with a range of different AC were acquired. In four volunteers (two males and two females, age 20–38 years), 10 scans were performed in a random order with different values of G, δ, and Δ, as shown in Table 1 with their corresponding acceleration-encoding parameters. In a scan time of ∼4.5 min, a total of 30 averages of both label and control images were obtained. The postlabeling delay was set to 1600 ms.

Table 1. MRI Acquisition Parameters Used for the AccASL Experiment with Varying Motion-Sensitizing Gradients and Their Cutoff Acceleration and b-Values
Cutoff acceleration (m/s2)Gradient strength (mT/m)Duration labeling module (ms)Gradient duration (ms)b-Value (m/s2)1−exp(−bD) (%, brain)1−exp(−bD) (%, CSF)
  1. The diffusion attenuation of the labeling module was calculated with the typical value of apparent diffusion coefficient (D): 0.8 × 10−3 mm2/s for gray matter and 2.5 × 10−3 mm2/s for CSF.

0.53040322.71.805.51
1.330600.50.970.0770.24
1.3156010.960.0770.24
2.3303011.920.150.48
2.3153021.900.150.47
3.5203010.850.0680.21
3.5103020.840.0670.21
4.7153010.480.0380.12
7103010.210.0170.053
753020.210.0170.053

For analysis purposes, a high-resolution three-dimensional T1-weighted image was obtained with a voxel size of 1.1 mm3. A total of 200 slices with a field-of-view of 240 × 188 × 220 mm3 was acquired in a scan time of about 2.5 min with repetition time (TR) of 7.6 ms and a echo time (TE) of 3.5 ms.

By using the Oxford Centre for Functional MRI of the Brain (FMRIB)'s Software Library (FSL), the images were realigned over all dynamics [15, 16] and corrected for motion with Motion Correction FMRIB's Linear Image Registration Tool (MCFLIRT) [17]. Subsequently, the time averages were calculated and the ASL maps were obtained by subtracting the mean label from the mean control images. An average CBF-weighted scan was generated from all 10 coregistrated images and with FMRIB's Non-linear Image Registration Tool (FNIRT) this average image was registrated to the atlas image of the gray matter (GM) tissue prior. The mean signal ASL maps and the standard deviation maps over the 30 repeated measurements from the 10 different methods were registrated to the image reflecting the GM tissue prior by combining the FLIRT and FNIRT matrices. The same transformation was applied on the mean and standard deviation maps of the control images. The standard tissue prior of GM was thresholded to obtain a binary GM mask to calculate the ASL signal intensity and temporal SNR within the GM region of the ASL maps. Mean region of interest (ROI) signal intensities were compared for the 10 different sequences with a two-way analysis of variance statistical test using Matlab.

Comparison to VSASL and pCASL

Next, the properties of the AccASL labeling module were compared with single VSASL, dual VSASL, and pCASL labeling. Eight volunteers (three males and five females, age 20–38 years) were scanned for this substudy. The labeling module parameters of AccASL were G=30 mT/m, δ=1 ms, and Δ=30 ms corresponding to an AC of 2.3 m/s2.

Single VSASL consisted of one velocity-selective labeling module, which saturated the spins above a VC of 2 cm/s. For dual VSASL, a second velocity-selective labeling module was added 1500 ms after the first VSASL sequence, just before the imaging module. Labeling module parameters, which determine the velocity sensitivity, were G=22 mT/m, δ=1 ms, and Δ=30 ms.

For AccASL, single VSASL, and dual VSASL, the postlabeling delay was set to 1600 ms. Imaging parameters were chosen as previously stated. Background suppression was applied using two adiabatic nonselective inversion pulses at 50 and 1150 ms after labeling to increase the contrast-to-noise ratio [18-20]. For all three techniques, the postacquisition delay, the time between the postacquisition nonselective saturation and the subsequent labeling module, was set to 2000 ms. Label and control acquisitions were interleaved in time and 30 averages of both image types were obtained. TR/TE was 4272/15 ms and the total scan duration was ∼4.5 min. Velocity and acceleration encodings were only applied along the slice-encoding direction. In a final comparison, balanced pCASL was used as the “conventional” ASL method and consisted of a labeling module of 1650 ms labeling with a 1525 ms delay before multislice acquisition with a TR of 3863 ms [21]. The labeling pulse duration is 0.5 ms, pause between the labeling pulses is 0.5 ms, the positive gradient lobe was three times larger than the negative lobe, labeling pulse flip angle is 18°, and averaged gradient in the direction of the blood flow is 0.6 mT/m. The obtained inversion degree was estimated to be ∼85%. Background suppression was applied using two inversion pulses at 1680 and 2760 ms. Total scan duration was ∼4 min with 30 averages of both label and control images.

For the comparison of the AccASL technique to single VSASL, dual VSASL, and pCASL, the images were coregistrated to a standard brain with FSL. The images were realigned over all dynamics [15, 16] and motion corrected with MCFLIRT [17]. Subsequently, the time averages were calculated and the ASL maps were obtained by subtracting the mean label from the mean control images. An average CBF-weighted scan was generated from all four coregistrated images and with FNIRT this average image was registrated to the standard tissue prior of GM. The ASL maps from the four different methods were registrated to the standard tissue prior of GM by combining the FLIRT and FNIRT matrices. The standard tissue prior of GM was thresholded to obtain a binary GM mask to calculate the ASL signal intensity and temporal SNR within the GM. The temporal SNR of the pCASL scan was corrected for the shorter total scan time by multiplying by the square root of the relative scan times. To calculate the mean and standard deviation of the ASL signal in the CSF, a mask, which only covered the ventricles, was created from the standard tissue prior of CSF. In a control image, the central voxel of the sagittal sinus was selected in eight slices and these were averaged to calculate the venous signal. Mean ROI signal intensities were compared for the four different sequences with a two-way analysis of variance statistical test using Matlab.

Quantification of absolute blood flow was performed similar to the procedure described by Chalela et al. [22], which is based on the general kinetic model with the equilibrium magnetization of arterial blood (M0a) estimated by using CSF as reference, although the model was slightly adopted to correct for the finite labeling duration of 1.65 s. The T1 of arterial blood was assumed to be 1.664 s [23], the T2(*) of arterial blood was assumed to be 50 ms, ρ was set to 1.05 g/mL, and the labeling efficiency and loss of label due to the background inversion pulses were, respectively, set to 0.85 and 0.83 [22]. Quantification of perfusion in dual VSASL is very similar to quantification of pulsed, spatially ASL, although a small adjustment is necessary for the fact that the VSASL applies saturation rather than inversion of the labeled spins [3, 24].

Multiple Postlabeling Delay Times

To get an impression of the changes in the spatial distribution of the label over time, a single slice was imaged with look-locker sampling scheme after labeling with an Acc or single VS labeling module in a single volunteer (female, age 22 years). The 16 postlabeling delay times varied from 300 to 3300 ms in steps of 200 ms, TR/TE was 200/14 ms, and the look-locker flip angle was 20°. The other imaging and labeling module parameters were the same as described above. A total of 30 averages were acquired in about 10 min.

To reconstruct the multi-TI ASL maps, label images were subtracted from the control images and averaged per postlabeling delay time in Matlab. The ASL maps were corrected for the flip angle used in the look-locker imaging strategy by multiplying each voxel by (cos(α)−1)n−1, where α is the flip angle and n is the number of the look-locker acquisition. In this way, the signal loss due to the excitation pulses is corrected for and the signal intensity will be similar to that as if each scan was acquired sequentially with a 90° flip angle.

Visual Stimulation

For mapping of neuronal activation, 70 dynamics of AccASL (total scan time of 10 min) were used to acquire five cycles of on- and off-periods with equal duration. Visual stimulation used an 8-Hz yellow and blue radial flashing checkerboard during the on-period and a black background with a white fixation point at the center during the off-period. Two volunteers (female, age 21–24 years) viewed the projections of the stimulation via a mirror attached to the head coil. The AccASL technique was performed with AC=2.3 m/s2, executed with G=30 mT/m, δ=1 ms, and Δ=30 ms, and 17 slices were acquired.

To aid the analysis, a blood oxygenation level-dependent (BOLD) scan was acquired for each volunteer with 150 dynamics (TR/TE of 2000/30 ms and total scan time of 5 min) with the same imaging region and voxel size as the AccASL scan, with the visual stimulation as described above, except that the duration of all blocks was only 30 s.

The fMRI data were analyzed in SPM8 and Matlab. Motion correction was performed on the label and control images from the AccASL data and images of the BOLD data separately. Then, the AccASL images were registrated to the BOLD images and all images were smoothed with an 8-mm3 full-width-half-maximum gaussian kernel. A threshold on the BOLD activation map was chosen to include only the visual cortex, where the highest activation was found, and these activated voxels from the BOLD data were used to mask the ASL signal from the subtracted control and label images of the AccASL data. The signal was averaged over all voxels within the mask per time point and averaged over the stimulation and rest blocks, respectively.

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Cutoff Acceleration

For all encoded cutoff accelerations (as presented in Table 1), a clear white matter–GM contrast typical for perfusion images was observed, as shown in Figure 2, which shows a typical example of a single volunteer. In the image corresponding to AC=0.5 m/s2 the signal intensity was decreased by a factor-of-four. For the signal from the CSF, the duration between the MSG plays an especially important role. With increasing Δ, and therefore decreasing AC, the CSF contamination increases in the ventricles and in white matter. In general, the mean ASL signal and SNR of the four subjects varied depending on the MSG parameters, as can be seen in Figure 3. Lower AC resulted in more intersubject variations of the ASL signal in GM, while the standard deviation of the SNR was similar for the whole range of AC.

image

Figure 2. Representative set of single-slice ASL maps with a range of cutoff accelerations achieved by variation of the acceleration parameters: gradient amplitude, the time between the 90° radiofrequency pulses, and the duration of the gradient, described as G/Δ/δ, respectively. All have the same scaling (ranged from 0 to 100 a.u.), except that AC=0.5 m/s2, which was scaled in a range of 0–400 a.u. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

image

Figure 3. The mean SNR (left bar), ASL signal (middle bar), and the signal that theoretically could be attributed to diffusion weighting (DW-signal, right bar) in gray matter for a range of cutoff accelerations achieved by different contributions of acceleration-encoding parameters. A gray matter mask was used to extract the voxels of interest. The error bars indicate the standard deviation of four subjects. The SNR was relatively constant for all ACs. The percentage of ASL signal attributable to the level of diffusion attenuation from the labeling pulses is shown underneath the acceleration-encoding parameters.

Download figure to PowerPoint

Comparison of AccASL to VSASL and pCASL

The properties of AccASL were tested and compared with single VSASL, dual VSASL, and pCASL. The ASL maps in a single volunteer from the four different ASL methods are shown in Figure 4. All maps have the same signal intensity scaling. AccASL and pCASL showed the highest ASL signal in GM, whereas the lowest signal was observed in the images acquired with dual VSASL.

image

Figure 4. Representative ASL maps averaged across 30 label/control pair images in a single volunteer. From top left clockwise: AccASL, pCASL, dual VSASL, and single VSASL. Identical color scales were used for the four sets of images. The acceleration encoding was AC=2.3 m/s2 for AccASL and the velocity encoding was VC=2 cm/s for both VSASL methods.

Download figure to PowerPoint

These visual observations are confirmed by the results of statistical testing as shown in Figure 5, where the mean SNR in GM and the mean ASL signal in CSF and the sagittal sinus are depicted along with a two-way analysis of variance statistical comparison. This showed that the SNR for single VSASL in GM was significantly lower (P<0.05) compared with AccASL, and the SNR for dual VSASL in GM was significantly lower (P<0.05) compared with all other methods. The ASL signal in the CSF and the sagittal sinus was significantly higher (P<0.05) for single VSASL compared with the other methods; no significant differences were observed between AccASL, pCASL, and dual VSASL for either region. Quantification of pCASL and VSASL resulted in CBF values of 55.2±12.3 and 92.6±17.2 mL/min/100 mL GM, respectively.

image

Figure 5. Mean SNR in gray matter (a), ASL signal in CSF (b), and ASL signal in sagittal sinus (c) for AccASL, single VSASL, dual VSASL, and pCASL, where # represents P<0.05 between single VSASL and AccASL, and * represents P<0.05 between this method and all other methods. The acceleration encoding was AC=2.3 m/s2 for AccASL and the velocity encoding was VC=2 cm/s for both VSASL methods. The error bars indicate the standard deviation of four subjects.

Download figure to PowerPoint

Multiple Postlabeling Delay Times

In Figure 6, the ASL maps acquired with AccASL and single VSASL are shown for the 16 postlabeling delay times. In the ASL maps at the first time points the label mainly resides in arteries and GM for AccASL, but also in large veins. During the first postlabeling delay times for the ASL maps acquired with single VSASL the signal is especially visible in arteries and veins, and lower signal intensity can be observed in GM. The label reaches the GM at an earlier time point for AccASL and is visible with a higher intensity compared with single VSASL. Although the signal intensity in the sagittal sinus in the AccASL maps decreases over time, and vanishes around 2 s and increases again around 2.5 s, the signal in the sagittal sinus stays present at a high intensity level for single VSASL at all postlabeling delay times.

image

Figure 6. Single-slice AccASL (left) and single VSASL (right) ASL maps at multiple postlabeling delay times with a 200-ms interval, acquired with a look-locker sampling scheme. The acceleration encoding was AC=2.3 m/s2 for AccASL and the velocity encoding was VC=2 cm/s for the VSASL methods. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Visual Stimulation

The average signal–time curve of the activated voxels is shown in Figure 7. These curves show a block pattern with an increase of 29% in signal during the visual stimuli compared to the baseline ASL signal during rest.

image

Figure 7. The signal–time curve detected by AccASL with AC=2.3 m/s2 for the visual stimulation fMRI experiment. Only the voxels that showed activation in the BOLD images are included. Rest and visual stimulation were altered in periods of seven label and control pair images. The curve shows the mean signal of the included voxels averaged over the five different rest and activation periods. The error bars indicate the standard error of means of two subjects. The increase in signal in the activated region during the visual stimuli compared to the baseline signal during rest is ∼29%. The gray line indicates the duration of the visual stimulus. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

In this study, we have proposed acceleration selection in ASL for labeling the arterial blood compartment within the imaging region without contamination from venous blood or CSF, and this sequence was successfully demonstrated in healthy volunteers to obtain resting-state perfusion maps as well as for fMRI.

By using a method with labeling not based on the spatial position of the spins, but on their acceleration, it was possible to label in the same plane as imaging, similar to VSASL. This creation of the label closer to the capillaries provides shorter and more uniform transit delays compared with traditional ASL, which makes this method theoretically suitable to use under slow or collateral flow conditions. This could also enable the use of shorter postlabeling times, resulting in a higher SNR, because there would be less loss of label through relaxation before it arrives in the microvasculature. This should, however, be confirmed by further research. Compared with conventional ASL, where the label is created by inversion of the magnetization, the use of saturation leads to a lower signal difference between the label and control condition, leading to a SNR penalty of a factor-of-two similar to VSASL.

The origin of the label in the blood created by AccASL is the result of at least three different processes. First, the flow velocity in the vessels is not constant, but is fluctuating because of the cardiac cycle. Acceleration-dependent preparation dephases the flowing spins magnetization that are accelerating or decelerating and a pulsatile flow has both of these properties. This characteristic was used previously in magnetic resonance angiography to achieve a good artery–vein separation, because blood flow in arteries is known to be more pulsatile than in veins [9]. Second, the general distribution of blood flow velocities over the cerebral vasculature is known to decrease for smaller vessels [25, 26]. In general, closer to the capillaries, the vasculature branches with vessels becoming smaller in diameter, but with the total surface area becoming larger, resulting in a lower blood velocity. Therefore, as the blood flows from the arterial compartment to the capillaries the blood will on average decelerate, whereas it will accelerate when flowing into the venous compartment. Third, the tortuosity of vessels, for example, at bifurcations of vessels or due to branching, but especially at the level of the capillaries, can also be a source of labeling in AccASL. When the directionality of the vessels changes with respect to the acceleration-encoding gradient direction, there will not be (complete) refocusing. Because of the structure of the capillary bed, more label is created compared with VSASL, which cannot be targeted to this region, as there is no laminar flow present [27]. All these three processes are introducing some cerebral blood volume weighting instead of the signal being only CBF-weighted, and further research should be performed to further study the specific hemodynamic contrast that is imaged by AccASL.

However, AccASL was shown to enable effective labeling of a large part of the arterial vasculature as proven by a similar SNR when compared with pCASL. The ASL signal intensity in GM was found to be comparable to pCASL and significantly higher compared with single and dual VSASL. Even though AccASL and VSASL both create label in the imaging plane, based on blood flow characteristics, it appears that more label is created with AccASL. Moreover, the signal maximum of the multi-TI measurements with AccASL and single VSASL is observed at the first time point, which indicates that the signal is specifically cerebral blood volume weighted, with filters of |A|>AC and |V|>VC, respectively. Also was demonstrated that at shorter postlabeling delay times more label was present in the GM of the ASL maps acquired with AccASL compared with single VSASL. This could make it feasible to reduce the postlabeling delay and thereby improve the time resolution, although future studies should prove this. Furthermore, the GM signal intensity also remained high at the later time points, proving that the label was indeed created in the arterioles and the microvasculature.

Focusing on the venous signal, the AccASL maps of the multi-TI measurements showed a decreasing signal intensity in the sagittal sinus up to 2.5–3 s postlabeling, whereas for single VSASL a high signal in the sagittal sinus was visible at all postlabeling delay times. Therefore, it appears that only a small signal is created in the venous compartment. This showed that an efficient elimination of venous signal can be accomplished with AccASL: no significant difference in ASL signal in the sagittal sinus was measured between AccASL, dual VSASL, and pCASL at a postlabel delay of 1600 ms, while venous contamination was clearly present in the image acquired with single VSASL. Nevertheless, elimination of the venous signal can be achieved with VSASL by the use of a second labeling module, applied just before imaging. For AccASL a waiting period is not required in order to exclude signal from the venous compartment, because labeling is based on the acceleration/deceleration during the labeling encoding module of typically 30–50 ms, while dual VSASL requires 1.0–1.5 s between two velocity-selective labeling modules to allow the spins to decelerate below VC. Additionally, the use of two labeling modules enables quantification of VSASL, because the temporal width of the labeling function is fixed, i.e., the amount of label created is being controlled. However, the use of a second labeling module in dual VSASL will also diminish the ASL signal by reduction of the labeling bolus by the second velocity-selective labeling module. Although elimination of the venous signal with AccASL was comparable to that using dual VSASL, in GM the ASL signal was significantly higher for AccASL. In future, the use of a second labeling module in AccASL should be examined to determine if this could make a valuable contribution to the quantification of the AccASL signal.

In AccASL and both VSASL sequences the same background suppression pulses were used. Nevertheless, the CSF signal was of comparable intensity for AccASL, pCASL, and dual VSASL, whereas it was higher for single VSASL. Therefore, the amount of CSF contamination is determined not only by the background suppression but also by the labeling sequence. A greater amount of CSF was apparently labeled with single VSASL.

The fMRI study showed that AccASL is able to detect hemodynamic changes associated with neuronal activation. An increase of ∼30% in signal during the visual stimuli was observed compared with the baseline ASL signal during rest. ASL methods are in general considered more suitable than BOLD for studying slow variations in brain function over periods greater than a minute, because ASL shows stable noise characteristics over the entire frequency spectrum [28, 29]. Furthermore, ASL-based fMRI has been advocated over BOLD because of improved localization. Vascular space occupancy (VASO) techniques, which are also cerebral blood volume weighted, improve localization of the activated regions during stimulation in fMRI, although at the cost of a lower SNR [30].The great advantage compared to conventional ASL techniques, which create label proximal to the image plane, is that with AccASL the label is generated in the same region as where the neuronal activation is located, so the label is created closer to the region where the local blood flow is altered. AccASL has therefore the potential to increase the temporal resolution of fMRI techniques based on hemodynamic imaging sequences. As the tagging is spatially nonselective, arterial transit time is not a limitation for anatomical coverage.

An artifactual contribution of the AccASL signal is caused by DW, because the MSGs of the labeling module also introduce diffusion sensitivity into the labeling process. The diffusion sensitivity of the sequence is characterized by the b-value and scales with the same parameters that determine the AC, but whereas the AC scales linearly with G and δ, the b-value scales quadratically. By varying the gradient parameters, the same AC can be obtained with different b-values and thus different contaminations by DW. Table 1 indicates how much change in signal intensity due to the use of MSG could theoretically be attributed to DW. To calculate contribution of the DW to the signal, the following formula was used: S/So=exp(−bD), in which S is the average signal in GM during the label condition, So is the average signal in GM during the control condition, b is the b-value corresponding to the gradient parameters, and D=0.0008 mm2/s for in GM [31-34]. The results from this calculation are shown in Figure 3. In this study, a maximum of 9% difference of the ASL signal was observed because of DW, which is within acceptable limits, except for the lowest AC with a b-value of 22.7 m/s2, where a 34% difference in ASL signal can be attributed to DW. Similar DW was found for AccASL when compared with VSASL [7].

In conclusion, the use of an acceleration-dependent preparation module for ASL with spatially nonselective labeling was demonstrated to be able to image brain perfusion without contamination from venous and CSF signal.

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES