The Effect of Inflow of Fresh Blood on Vascular-space-occupancy (VASO) Contrast

Abstract

In vascular-space-occupancy (VASO)-MRI, cerebral blood volume (CBV) weighted contrast is generated by applying a non-selective inversion pulse followed by imaging when blood water magnetization is zero. An uncertainty in VASO relates to the completeness of blood water nulling. Specifically, radiofrequency coils produce a finite inversion volume, rendering the possibility of fresh, non-nulled blood. Here, VASO-fMRI was performed for varying inversion volume and TR using body coil radiofrequency transmission. For thin inversion volume thickness (δ_tot<10 mm), VASO signal changes were positive (ΔS/S=2.1–2.6%). Signal changes were negative and varied in magnitude for intermediate inversion volumes (δ_tot =100–300 mm), yet did not differ significantly (P>0.05) for δ_tot>300 mm. These data suggest that blood water is in steady-state for δ_tot>300 mm. In this appropriate range, long-TR VASO data converged to a less negative value (ΔS/S=−1.4±0.2%) than short-TR data (ΔS/S=−2.2±0.2%), implying that cerebral blood flow or transit-state effects may influence VASO contrast at short TR.

Introduction

Microvascular cerebral blood volume (CBV) is known to adjust during periods of increased neuronal activity (1–5). Such CBV changes have been reported to occur primarily in arterioles and capillaries (2), with a smaller change possibly occurring in venules as well (6). Since such CBV changes occur in the microvasculature and not large draining veins, identification of CBV adjustment concurrent to neuronal activity should be useful for improving the spatial localization of brain function over conventional blood-oxygenation-level-dependent (BOLD)-fMRI (7,8).

Vascular-space-occupancy (VASO) fMRI (9) is gaining popularity as a method for noninvasively assessing CBV changes accompanying neuronal activity. In VASO, a spatially non-selective inversion pulse is applied, after which an image is acquired at an inversion time (TI) when the longitudinal magnetization (M_z) of blood is expected to be zero, yet M_z of tissue is positive. Typically, a negative VASO-fMRI signal change (ΔS/S) is observed during increased neuronal activity, which has been attributed to a decrease in the detectable tissue signal consequential to the CBV increase (9). Experimental data in humans (9,10) have confirmed VASO localization to gray matter parenchyma, while recent high-resolution data in cats (11) indicate localization of VASO contrast to the middle cortical layer, in agreement with CBV approaches using paramagnetic agents.

However, similar to other fMRI methods, VASO has been criticized for uncertainty regarding the nature of its contrast mechanism. The most important question relates to the contribution of intravascular blood water that is not nulled. Potential causes of this could be exchange between capillary blood water and tissue water (9,12), as well as inflow of fresh (i.e. non-nulled) blood into the image slices. Although a body-coil and adiabatic inversion have been used in VASO applications until now (9,12–14), the width of the inversion volume may differ per instrument and may allow for some fresh blood, which is not in steady-state and therefore not nulled, to enter the imaging volume before acquisition. It has previously been simulated that such fresh blood will cause VASO ΔS/S to become more negative (10). Here, VASO ΔS/S is measured as a function of inversion volume thickness to assess how fresh blood may influence VASO contrast.

Methods

Theory

In VASO, a spatially non-selective inversion pulse is applied to invert all proton magnetization, after which an image is acquired at a TI when longitudinal blood magnetization is expected to be zero. Since tissue water T₁ (1100–1300 ms at 3.0T (13)) is shorter than blood water T₁ (1600–1800 ms at 3.0T (14)), the extravascular tissue signal will be slightly positive at TI. Multiple VASO images can be acquired in a time-efficient manner if a repetition time (TR) shorter than the amount of time required for tissue magnetization to recover to equilibrium is used. In this case, tissue and blood are presumed to be in steady-state and the TI can be adjusted to keep the steady-state blood water magnetization nulled.

In the cerebral cortex, the VASO signal is proportional to the gray matter parenchymal signal (S_par), which comprises the extravascular tissue (t) and microvascular blood (b) and is assumed to change between a baseline (base) and vasodilatory (act) state:

$$S_{\mathit{\text{par}}}^i(TI)=S_t^i+S_b^i$$ (1)

where i = activated (act) or base.

When using a gradient echo sequence, the tissue signal is given by

$$S_t^i(TI)\approx \left(C_{\mathit{\text{par}}}-CB{V}^i·C_b\right)M_t^i(TI)·e^{-TE/T_{2,t}^{*i}}$$ (2)

, where C_par ≈ 0.89 mL water/mL parenchyma (15) is the water density of gray matter parenchyma, M_t is the longitudinal magnetization of extravascular tissue, TE is the echo time and T_2,t* ≈ 45.3 ms (base) or 46.6 (act) are the effective transverse relaxation times for extravascular tissue as reported in the literature (13). For blood, relaxation times will vary with microvascular compartment. Assuming that an MR voxel comprises 21% arterioles, 33% capillary and 46% venules (8), the blood compartment can be approximated as 38% arteriolar (a) and 62% venular (v),

$$S_b^i=C_b·\lfloor 0.38·CB{V}_a^i·M_{b,a}^i(TI)·e^{-TE/T_{2,a}^{*}}+0.62·CB{V}_v^i·M_{b,v}^i(TI)·e^{-TE/T_{2,v}^{*}}\rfloor$$ (3)

where i=act or base.

Eq. 3 assumes that the CBV change may occur in either the arterioles or venules. C_b≈87 mL water/mL blood is the water density of blood (16). Assuming a microvascular Hct≈37 (10), the transverse relaxation times for blood can be calculated according to T₂*=[A*+B*(1-Y)+C*(1-Y)²]⁻¹ (17,18) where A*=16.6 s⁻¹, B*=37.3 s⁻¹ and C*=99.6 s⁻¹ (which have been extrapolated from recent work in bovine blood (18)). Assuming arteriolar and venular oxygen saturation fractions (Y) of 0.98 and 0.61, respectively, the transverse relaxation times for blood are found to be 57.6 ms (arterioles) and 21.6 ms (venules).

The total VASO signal change (ΔS/S) in such a functional experiment is:

$$\frac{\Delta S_{\mathit{\text{par}}}}{S_{\mathit{\text{par}}}^{\mathit{\text{base}}}}=\frac{S_{\mathit{\text{par}}}^{\mathit{\text{act}}}-S_{\mathit{\text{par}}}^{\mathit{\text{base}}}}{S_{\mathit{\text{par}}}^{\mathit{\text{base}}}}$$ (4)

.

It has commonly been assumed that all blood water in the VASO experiment is nulled and Eq. 3 vanishes (9). However, this is only true for steady-state (ss) blood water and even here this is an approximation since blood T₁ varies between microvascular compartments. In practice, the inversion volume of the body coil is finite, which may lead to inflow of some fresh (i.e. non-steady-state) blood and up to two additional magnetic situations for the blood water. First, if TR is longer than the time to refresh blood, blood water magnetization will appear as being fully relaxed and, instead of a multi-inversion steady state, experience only a single inversion pulse in each scan. This so-called fully-relaxed (f-r) blood will flow into the imaging slice and will have non-zero magnetization at the TI calculated for steady-state blood-nulling. Second, for thin inversion volumes, a fraction of blood in the imaging slice may be completely refreshed; this blood will have magnetization that is the same as at equilibrium (eq). Therefore, longitudinal blood magnetization can be expanded:

$$M_{b,j}(TI)=x_{b,j,eq}·M_{b,j,eq}+x_{b,j,f-r}·M_{b,j,f-r}(TI)+x_{b,j,ss}·M_{b,j,ss}(TI)$$ (5)

where j=arteriole or venule.

These magnetizations relate to the equilibrium magnetization (M₀) via

$$M_{b,j,eq}=M_0$$ (6)

,

$$M_{b,j,f-r}(TI)=M_0\left(1-2e^{-TI/T_{1,j}}\right)$$ (7)

, and

$$M_{b,j,ss}(TI)=M_0(1-2e^{-TI/T_{1,j}}+e^{-TR/T_{1,j}})$$ (8)

.

The inversion pulse will always invert the imaging slice, and therefore the tissue signal will be in steady-state,

$$M_t(TI)=M_0(1-2e^{-TI/T_{1,t}}+e^{-TR/T_{1,t}})$$ (9)

.

T₁ refers to the longitudinal relaxation time, which is taken to be 1209 ms for tissue (19), 1747 ms for arterioles and 1703 ms for venules at a field strength of 3.0T. Arteriolar and venular T₁ are slightly higher than those used in previous work at the same field strength (1624–1627 ms (10,13)); this is because here we improve the specificity by using microvascular T₁, corresponding to a Hct≈37 and varying oxygenation status (98% for arterioles and 61% for venules) whereas previous work used a larger Hct≈44, corresponding to larger vessels. To understand the influence of residual microvascular blood water, simulations were performed using the optimal TI calculated for macrovascular blood water nulling (T_1,b=1627 ms); this allows for better interpretation of the currently used approaches in the literature.

Experiment

The body coil length of the scanner used in these experiments is 650 mm, which was confirmed by the local MR engineer. Therefore an adiabatic pulse is expected to invert no less than this length. A magnetic field gradient associated with the adiabatic inversion was introduced to allow for a finite volume thickness (δ_tot) of water to be inverted. To test maximal inversion thicknesses with respect to inflowing blood, the imaging slice was not positioned in the center of the coil, but toward the top with the isocenter of the coil centered approximately 1 cm inferior to the volunteer’s mouth. VASO-fMRI experiments were performed for a non-selective inversion pulse (δ_tot≥650 mm), as well as for δ_tot=10, 100, 200, 300, 400, 500, and 600 mm, separately at a short TR (2s) and a long TR (5s). These 16 experiments were pseudo-randomized and repeated on five healthy volunteers (age: 31±6 yrs), all of whom provided informed, written consent to an IRB-approved protocol in accordance with HIPAA guidelines. Scan parameters: Philips 3.0T MR scanner (Philips, Best, The Netherlands), 3.75×3.75×5 mm³ nominal spatial resolution, FOV = 240×240 mm², single-shot gradient echo EPI, TE = 12.7 ms, SENSE-factor = 2.5, TR/TI=2000/711 ms or TR/TI=5000/1054 ms, as based on T_1,b = 1627ms. All slices were co-registered and motion corrected using the FSL Linear Image Registration Toolkit (FLIRT) (20). Subsequently, images were corrected for baseline drift using a cubic spline interpolation algorithm. Slices were reconstructed and zero-filled to a matrix of 256 × 256.

The fMRI paradigm consisted of 30s/30s on/off cross-hair fixation/flashing checkerboard (8 Hz) repeated three times. An additional 30s of data was acquired before the paradigm block began to allow for blood magnetization to reach a steady-state; data acquired during this period were not analyzed. Positive and negative VASO ΔS/S were calculated from voxels in the occipital cortex meeting activation criteria in a z-hypothesis test; negative activation criteria were: z≤−3 (P<0.05), SNR≥20, cluster size≥4. Positive activation criteria were identical, except with z≥3. Time courses were calculated for each δ_tot scan for voxels meeting negative activation criteria in the non-selective VASO experiment. This assured that common voxels were analyzed in all scans. It should be noted that this voxel selection method controls for partial volume effects between scans. Significance of ΔS/S variation between δ_tot values or TR values was based on a paired t-test with significance defined as P<0.05. Finally, number of voxels meeting positive (z≥3) or negative (z≤−3) activation criteria in each scan were separately recorded in order to understand how the proportion of voxels showing positive or negative signal changes was influenced by inversion volume.

Results

Simulations

The magnitude of different blood magnetizations (Fig. 1a) is simulated as a function of TR. Blood water in equilibrium has a constant magnetization, as expected based on Eq. 7, yet fully-relaxed blood water magnetization is more negative at short TR (Eq. 8). Steady-state blood water magnetization is nearly nulled at all TR values, as is assumed in the VASO literature. Steady-state blood magnetization, however, is not completely nulled (residual magnetization≈−2%) since the nulling time (TI) was based on the macrovascular blood T_1,b=1627 ms, as is the case in current VASO studies, which differs appreciably for the arteriolar and venular blood magnetization at Hct =0.37 (1747 and 1704 ms, respectively). Tissue magnetization increases with TR, as expected based on steady-state considerations.

Figure 1.

(a) Simulations depicting the TR-dependence of two possible fresh blood contributions, namely equilibrium (eq) and fully-relaxed (f-r) blood water magnetizations. The steady-state (ss) blood and tissue magnetizations are also shown. (b) Simulation showing the effect of three separate cases: pure equilibrium (x_eq=1), fully-relaxed (x_f-r=1) and steady-state (x_ss=1) blood water magnetization on VASO signal change (ΔS/S) using CBV^base = 0.05 mL/mL, CBV^act = 0.07 mL/mL. All change is assumed to originate from the arteriolar compartment. (c) Simulation showing the same three magnetization cases, except with vasodilatation assumed to originate in the venular compartment.

In Fig. 1b, the influence of different blood magnetizations on VASO ΔS/S is shown assuming arteriolar CBV changes. VASO ΔS/S is simulated assuming three conditions: 100% equilibrium blood, 100% fully-relaxed blood and 100% steady-state blood. The voxel will likely contain a mixture of each of these blood types; Fig. 1b provides only an overview of how the different blood types will influence the contrast. For equilibrium blood, VASO ΔS/S is positive, becoming smaller at long TR due the exponential decay of the steady-state tissue magnetization. Here, the intravascular blood water BOLD signal change (Eq. 3) dominates over the extravascular blood volume change in the tissue signal term (Eq. 2). For fully-relaxed blood water, VASO ΔS/S is more negative at short TR than at long TR. This is because the fully-relaxed blood null time (TI≈126 ms) is closer to the steady-state blood water null time at long TR (TI≈054 ms) than that at short TR (TI≈11 ms). Finally, for steady-state blood water, VASO ΔS/S is found to be approximately −1.6% and relatively independent of TR. Fig. 1c shows simulation results assuming the same CBV change occurs in the venules (as opposed to arterioles). Notice that the general trends are the same, however the absolute numbers are slightly smaller with respect to absolute magnitude of the signal change.

It is important to realize that in some situations, blood may be in a transit-state, when it is neither in equilibrium, fully-relaxed nor at steady-state. Such a transit-state may exist when blood water begins having one magnetization condition (e.g. equilibrium) and ends with another (e.g. steady-state) and may persist for several TR periods. Fig. 2 shows longitudinal blood magnetization as a function of time for TR=2s (a) and TR=5s (b) VASO experiments. Note that for TR=2s, it takes approximately three inversions before blood is in steady-state. For TR=5s VASO experiments, the transit-state is nearly undetectable due to the much longer times between inversion pulses. The simulations here were performed assuming all blood magnetization experiences an inversion pulse and no excitation pulse, which is true for the majority of blood in single-slice experiments when non-selective inversion pulses are used.

Figure 2.

(a) Simulated blood magnetization in a TR=2s VASO experiment. Note that the blood magnetization does not reach steady state for approximately three inversion cycles. In the first TR period, the blood is fully-relaxed; however, in the next two TR periods it is in a transit-state, where it is neither fully-relaxed nor in steady-state. After the third TR period, the blood magnetizatiton is in steady-state. (b) Simulated blood magnetization in a TR=5s VASO experiment. Note that here the transit-state is nearly undetectable. The vertical dashed lines depict the expected blood nulling times assuming steady-state blood magnetization (711 ms after inversion at TR=2s and 1054 ms after inversion for TR=5s).

Experiment

Fig. 3a shows how the VASO inversion thickness (δ_tot, blue + red) was varied experimentally. Note that the inversion was centered at the imaging slice (yellow); therefore, the distance below the slice (δ_in, red), which is relevant to inflowing arterial blood water, is half of δ_tot. Representative z-maps for short and long TR for a single subject are shown for varying δ_in (Fig. 3b); all inversion values are within the range of the coil, especially with the slice positioned higher in the coil. Note that the slices for short and long TR are slightly different since the two scans were performed at different times; we found that dividing the experiment into two sessions allowed for better volunteer compliance over the long fMRI paradigm (e.g. two sessions of eight fMRI scans vs. one session of 16 fMRI scans). For small δ_in=5 mm, the visual cortex shows primarily positive z, indicating that the voxels contain largely equilibrium blood.

Figure 3.

(a) Location of the VASO imaging slice (yellow) and corresponding total inversion slab thickness (δ_tot) which was varied; δ_in (red) shows the inversion volume relevant to inverting inflowing arterial blood water. (b) Representative z-maps for a single subject for TR=2s (above) and TR=5s (below). Note that for small δ_in=5 mm, activation (z) in the occipital cortex is largely positive, whereas it is primarily negative for larger δ_in. (c) The effect of the inversion volume thickness (δ_in) on VASO signal changes (ΔS/S). Values are mean±standard error (n=5). Three regions are demarcated. Region I likely contains significant fractions of fully-relaxed and equilibrium blood, region II contains a mixture of fully-relaxed and steady-state blood and region III contains blood that is primarily in steady-state. (d) A surface plot showing the effect of δ_in on VASO ΔS/S for the three separate task periods for a representative subject. Notice that the signal changes plateau to approximately the same value consistently in each of the three paradigm blocks.

Fig. 3c shows ΔS/S as a function of δ_in for the two TR values studied. Signal changes represent the same subset of voxels, which met the negative activation criteria in the non-selective VASO experiment (see Methods). ΔS/S traverses three regions (I–III). In region I (5 mm ≤ δ_in ≤ 50 mm), mean ΔS/S (n = 5) is positive at small inversion widths and becomes negative when increasing the thickness of the slab. In region II (50 mm < δ_in ≤ 150 mm), ΔS/S becomes more negative at short TR, while it plateaus at long TR. In region III (δ_in > 150 mm), ΔS/S does not change significantly (P>0.05). In region III, ΔS/S was statistically more negative at TR=2s than at TR=5s (P<0.05). Fig. 3d shows a surface map for a representative subject which depicts the relationship between the time, δ_in, and ΔS/S for TR=2s. In all three task periods, ΔS/S is most negative for small δ_in (dark blue) and becomes approximately equal in all task periods for δ_in>150 mm.

Table 1 shows the number of voxels meeting activation criteria in each scan and also compares VASO ΔS/S for the two repetition times used. Note that for both short and long TR, the thin inversion widths contain many positively activated voxels due to equilibrium blood contributions, whereas the thick inversion volumes contain consistently more negatively activated voxels. The proportion of negatively activated to positively activated voxels was not statistically different (P>0.05) between TR periods at any δ_in value. However, more voxels meeting both negative and positive activation criteria were found at short TR. The P value corresponding to the significance between the short and long TR ΔS/S is shown (*). ΔS/S is more negative at short TR than at long TR, an observation that is most significant (P<0.05) at δ_in=100 mm due to fully-relaxed blood contributions, but is also significant (P<0.05) for δ_in≥200 mm. It should be noted that the transition from δ_in=100 mm to δ_in=150 mm is the last statistically significant transition, as VASO signal changes for larger inversion volumes are non-distinguishable at a P value of 0.05. When assessing whether the measured ΔS/S at a certain δ_in value is significantly different from the ΔS/S measured at the thinner δ_in value (#), no significant change is observed for δ_in≥100 mm, with the highest P values generally being in region III.

Table 1.

VASO activated voxels, signal changes (ΔS/S) and P values as a function of inflowing inversion volume thickness (δ_in).

	Activated Voxels				ΔS/S (%)		PTR2 vs TR5*	Pδin vs δin-before#
	Positive		Negative
δin (mm)	TR=2s	TR=5s	TR=2s	TR=5s	TR=2s	TR=5s		TR=2s	TR=5s
5	507±87	619±52	99±41	27±3	2.6±1.2	2.1±1.2	0.40	-	-
50	256±89	321±41	265±90	77±20	−1.7±0.7	−0.1±0.5	0.07	<0.01	0.02
100	155±45	32±6	352±97	133±31	−2.8±0.6	−1.3±0.3	<0.01	0.03	0.03
150	89±23	78±34	317±87	248±23	−2.4±0.5	−1.8±0.3	0.19	0.13	0.14
200	83±22	65±43	368±96	185±35	−2.3±0.5	−1.3±0.3	0.02	0.38	0.18
250	95±28	39±2	258±85	195±43	−2.1±0.3	−1.3±0.3	0.02	0.13	0.29
300	87±30	51±16	332±46	227±101	−2.0±0.3	−1.4±0.3	0.04	0.33	0.28

Activated voxels and signal changes (ΔS/S) represent the mean over five subjects ±standard error.

^*Significance of difference between ΔS/S measured at TR=2s and TR=5s for the given δ_in.

^#Significance of a given ΔS/S having the same sample mean as ΔS/S from the previous δ_in measurement. Changes that are significant at P<0.05 are italicized.

Discussion

The proportions of fresh and steady-state blood in VASO experiments

The data presented here demonstrate how VASO contrast is influenced by non-nulled intravascular blood water contributions due to the inflow of fresh (equilibrium and/or fully-relaxed) blood. Simulations predict two counteracting effects. First, contributions from fresh equilibrium blood will cause VASO ΔS/S to become positive, with this effect being largest at short TR. Second, fresh fully-relaxed blood will cause VASO ΔS/S to become more negative, but only at short TR where the null times for steady-state and fully-relaxed blood vary. The experimental short-TR ΔS/S trends as a function of inversion volume thickness therefore provide important clues regarding the relative extent of these two different blood magnetizations in the total VASO signal. Specifically, in region I, for δ_in=5 mm, mean VASO ΔS/S values are 2.6±2% and 2.1±2% at TR=2s and TR=5s, respectively. These consistent positive values indicate that equilibrium blood is dominating for this thinnest inversion volume. As can be seen in Fig. 3b and Table 1, positive and negative ΔS/S varies with voxel location, indicating that the amount of fully-relaxed and equilibrium blood varies spatially. In region II, VASO ΔS/S becomes more negative indicating that the CBV contribution to the VASO effect dominates. However, a large difference between short and long TR can be seen, indicating that fully-relaxed blood contributions are dominating for short TR. When small coils are used, it is possible that VASO ΔS/S will be more negative at short TR due to this limitation. In region III (δ_in>150 mm), ΔS/S approximately plateaus (P>0.05) owing to the fraction of fully-relaxed blood decreasing and the fraction of effectively nulled, steady-state blood increasing. Therefore, these data suggest that a volume of 150–300 mm below the imaging slice should be sufficient for steady-state conditions of blood nulling to hold.

In order to control for partial volume effects between scans, common voxels were analyzed for each inversion volume VASO scan (see Methods) and mean signal change within these voxels was recorded. In theory, such an approach could give an inaccurate view of the signal changes if equal number of positive and negative signal changes are occurring within the voxels, yielding an overall small net signal change. However, we observed largely positive signal changes at thin inversion volumes and largely negative signal changes at thick inversion volumes, as can be seen in Fig. 3b and Table 1. The only inversion volume where approximately equal amounts of positively activated and negatively activated voxels were found was for δ_in=50 mm, where significant amounts of both equilibrium and fully-relaxed blood were presumed to be present. This counteracting effect was most pronounced at short TR (Table 1) since fully-relaxed blood water contributions are more significant at short TR.

As was shown in Fig. 2, it is possible for blood magnetization to occupy a transit-state, in which magnetization is not fully-relaxed, in equilibrium, or in steady-state. Furthermore, it was shown that this transit-state is more significant at short TR than at long TR. In the present experiments, data acquired during the first 30s were not analyzed, thereby allowing for blood magnetization to reach a potential steady-state. Such a waiting time is very important in VASO experiments and VASO data acquired using only one or two measurements, especially at short TR, should be interpreted with caution. It is of interest to note that, even when deleting data acquired during the first period, transit effects could occur when the steady-state changes during brain activation. These effects are presumably small because CBV changes are expected to occur locally. However, changes in blood velocity may slightly affect the presumed steady-state. Importantly, when blood water cannot reach a steady-state due to thin inversion volumes, the transit-state may persist for a much longer period, and will depend complexly on slice position, inversion volume, blood velocity and anatomical variation. As is shown in Fig. 2, one way to minimize such transit-state effects is by imaging at long TR. Similarly, fully-relaxed blood contributions can be minimized by imaging at long TR, as was demonstrated in Fig. 3c. Therefore, to desensitize VASO contrast to transit-state and fresh blood uncertainties, long TR (e.g. TR=5s) acquisitions are recommended.

Given that it takes approximately three inversions for blood to reach steady-state in a TR=2s VASO experiment (Fig. 2), it is important to consider the mean velocity of blood within the typical VASO inversion volume (δ_in=300 mm). Note that the above steady-state condition requires two inversions after the initial inversion, meaning that only 4s is required after the initial inversion for blood to reach steady-state. Therefore, the mean velocity of inflowing blood must be less than 75 mm/s. Flow velocity will vary significantly between inflowing vascular compartments (e.g. arteries, arterioles and capillaries), depending on diameter. Blood velocity in large arteries is approximately 600–800 mm/s (5–10 mm diameter), falling to 100–300 mm/s in smaller arteries (~1 mm diameter), 1–10 mm/s in arterioles (10–100 µm diameter), and 0.7 mm/s in capillaries (see Piecknik et al. (21) and references therein). Using a multi-compartment vascular tree model, it has been proposed that it takes approximately 4–5s for blood in large inflowing vessels (10 mm diameter) to reach the capillary exchange site. The precise velocity and inflow time is complex since the vascular tree contains an array of vessel sizes and orientations; also, plasma and red blood cells have different flow velocities, and both will contain water protons. Therefore, the range of velocities for intravascular blood water will clearly vary. Recently, using arterial spin labeling with multiple inversion times, it was shown that occipital arterial arrival time is longer than mean arterial arrival time in cerebral cortex (22); these results are in agreement with PET literature which has shown that mean transit time in occipital cortex is longer than in most other cortical regions (23). Therefore, it is more likely that blood will be in steady-state in visual experiments, such as were conducted here, than in other fMRI experiments using motor or auditory stimuli. It may be noticed that there is an apparent slight upward trend in Fig. 3c, suggesting that if inversion volumes were increased beyond what is possible with standard body coils, the TR=2s signal change would converge to the TR=5s signal change. However, a statistical test (Table 1) shows no significant difference between consecutive signal change measurements over the range δ_in=150–300; therefore, at present it can only be claimed that although there appears to be an upward trend, it is not significant.

An additional uncertainty relates to which value for T₁ of blood water is correct to use when calculating the blood nulling TI. Here, it was shown that for steady-state blood water, microvascular blood magnetization is small (≈−2%; Fig. 1a) when T₁ for macrovascular blood water is used to calculate TI. The present study used only a single TI value, and therefore it is not possible to say how VASO signal changes will vary experimentally with choice of TI. Assuming that a given voxel contains a mixture of arteriolar, capillary and venular blood, it is reasonable to assume that a TI calculated from the average macrovascular T₁ is a reasonable choice. However, as higher resolution VASO fMRI becomes available and vasodynamics of different microvascular compartments are more fully understood, it is likely that VASO experiments may be employed with TI specific to different microvascular compartments. This would be more possible at higher magnetic fields, where the difference between arterial and venous blood T₁ is expected to increase if the same trend for oxygenation dependence holds as between 1.5T and 3T (14).

Additional contributions to VASO contrast

These data demonstrate how inflow of non-nulled blood water influences VASO contrast. However, two additional effects should be considered. First, partial volume contributions from CSF (5–15%) will cause VASO ΔS/S to become slightly more negative at long TR than at short TR (10). However, this contribution is small and will influence both short and long TR VASO ΔS/S, and therefore cannot fully account for the discrepancy observed between short and long TR VASO ΔS/S at δ_in=150–300 mm. As shown in previous work, this discrepancy may be due to a second effect, namely changes in tissue M_z consequential to exchange between the blood water spins and tissue water spins, similar to arterial spin labeling (ASL). This ASL effect, which has been shown to make VASO ΔS/S more negative (12,24), is most pronounced at short TR. This CBF influence to VASO contrast may provide a basis for future experiments to extract this important hemodynamic parameter, concurrently with CBV, from VASO data (24). This extraction of physiological parameters is not trivial as the exchange between capillary blood and tissue may also affect blood nulling. Future work should aim to understand such complex exchange effects in VASO contrast. To successfully accomplish this, it will be necessary to generate multi-compartment vascular models which account for vascular dilatation, in addition to exchange effects within the capillary compartment.

Experimental limitations

Several experimental limitations should be noted. First, a single slice through the visual cortex was analyzed here, as is the case in many VASO-fMRI studies (9,10). For VASO-fMRI performed in other brain regions, the local vascular geometry may cause the arrival time for fully-relaxed and equilibrium blood contributions to differ. Second, a body coil-administered adiabatic inversion is expected to invert a length no less than the length of the body coil. Here, the maximum inversion volume used was δ_tot=600 mm, which is less than the length of the current body coil (δ_tot=650 mm). However, body coil length will vary with scanner and vendor and care should be taken in performing VASO measurements when using smaller body coils or head coils for RF administration. Finally, T₁ varies considerably between blood compartments; therefore, the choice of null time will influence the amount of residual blood signal that remains from different blood compartments, however, as noted above, this residual blood magnetization is expected to be small.

Conclusions

This work shows how VASO signal changes may vary with the thickness of the inversion volume and suggests that for sufficiently large body coils (δ_tot≥300 mm), VASO ΔS/S is independent of inversion volume thickness, implicating that intravascular blood water is in steady-state and largely nulled. Secondly, at such thick inversion volumes, a discrepancy between short and long TR VASO ΔS/S persists, indicating that additional effects, such as exchange between capillary blood and tissue or transit-state effects may alter VASO contrast. Care should be taken when interpreting VASO ΔS/S on smaller body coil systems, or when head coils are used for transmission, due to possible inflow of fresh blood. A simple way to minimize the effect of fully-relaxed blood and transit-state uncertainties in VASO contrast is to image at long TR, where the equilibrium and steady-state blood null times are more similar than at short TR.