Anesthetic effects on regional CBF, BOLD, and the coupling between task-induced changes in CBF and BOLD: An fMRI study in normal human subjects

Subjects and Anesthetic Administration

Sixteen healthy consenting subjects (American Society of Anesthesiologists, physical status class I) 19- to 30-year-old were administered inhalational 0.25 MAC end-tidal sevoflurane. Subjects on psychoactive drugs or any centrally acting medication were excluded. Subjects with a history of renal disease and those with potentially difficult airway were also excluded. All patients fasted for 8 h before the study. After passing the physical and medical history examination, subjects proceeded to the magnetic resonance research center (MRRC) where they were screened for ferromagnetic materials and then connected to standard ASA (American Society of Anesthesiologists) monitors. An intravenous infusion line was started with a 22G cannula for maintenance infusion (lactated ringer at 100 mL/h). Subjects were moved into the magnet with the monitors and intravenous infusion line in place. Anesthesia was induced and maintained with oxygen (5 liters) and sevoflurane 0.5% (0.25 MAC) administered through a semi-closed circuit and a facemask. The facemask was held in place with head straps. During anesthesia of 0.25 MAC sevoflurane, subjects were able to talk and respond to questions. ECG, noninvasive blood pressure, oxygen saturation, end-tidal carbon dioxide, and end-tidal sevoflurane concentration were monitored continuously throughout the study. The monitors used in this study are serviced and calibrated every 6 months, as recommended by the manufacturer, OHMEDA (Madison, WI).

Stimulation

Visual stimulation consisted of an 8 Hz reverse-flashing, full-field black and white checkerboard (shown as an inset in Fig. 3a). The horizontal and vertical extents were approximately 16 and 12 degrees, respectively. During rest intervals subjects fixated on a white plus sign (+) presented against a black background. The auditory stimulus consisted of randomly presented pure tones and a white-noise sound burst, presented through MR compatible headphones. The tones were played out at different frequencies (including 400, 600, 800, 1000, 1500, 2000, and 3000 Hz) in randomized order in a block design experiment. The duration of each tone was 250 ms with an ISI of 250 ms and the sound pressure level of the auditory stimulus was 95 dB, with attenuation by the ear inserts of 15 to 20 dB; this was determined in a pilot study and kept constant for all subjects. The auditory and visual stimuli were presented simultaneously in 4 cycles of 30-s blocks, alternating with 30-s periods of rest; these clusters started and ended with resting periods. Clinical practice and our pilot studies indicated that the end-tidal sevoflurane concentration reached a steady-state approximately 9 minutes after the induction of sevoflurane. Based on ECG readings, it took approximately 9 minutes for the subjects to recover after the withdrawal of the agent. The subject was positioned once at the beginning of the experiment; head straps and pads were used to help control head motion during the experiment. Two anesthesia sessions were interleaved with 3 anesthesia-free sessions. During each session, the stimulation procedure was repeated three times. Perfusion-weighted and BOLD-weighted MRI data were collected during all sessions.

Regional CBF and BOLD Measurements

Imaging was performed on a 3 Tesla (T) whole-body scanner Trio (Siemens Medical Systems, Erlangen, Germany) with a circularly polarized head coil. Pulsed arterial spin labeling imaging was performed to measure regional CBF and BOLD in the presence of auditory/visual stimuli in the anesthesia-free and anesthesia conditions with a modified EPISTAR QUIPSS ASL sequence (23, 24). This involved applying a train of thin-slice saturation pulses at TI₁ =700 ms after the radiofrequency (RF) inversion pulse so as to control the bolus delivery and suppress the intravascular signal from large vessels (25). A slab-selective hyper-secant RF inversion pulse was used for labeling the arterial blood water. The RF pulse was applied to a slab 20 mm inferior to the imaging slab. As a control, the same RF pulse was applied to a slab 20 mm superior to the imaging slab. Interleaved labeling and control images were acquired using a gradient echo-planar imaging (EPI) sequence, followed by a recovery time allowing arterial blood to be refreshed.

The ASL acquisition parameters were: field of view = 256 × 256 mm²; matrix = 64 × 64; bandwidth = 2056 Hz/pixel; slice thickness = 8 mm; inter-slice spacing = 2 mm. Ten AC-PC aligned slices were acquired from inferior to superior in an ascending order to cover most of the cortex. Acquisition of each slice took approximately 60 ms. The repetition time was TR = 3000 ms; the echo time was TE = 26 ms, and the delay time TD within each TR was adjusted to maximum, which was 900 ms. During each EPI acquisition, fat was suppressed and phase-correction was performed. A bipolar gradient of encoding velocity V_enc = 20 mm/s was applied to the imaging slices for intra-vascular signal suppression.

To get the absolute values for regional CBF, 40 proton density weighted images were acquired with the same perfusion sequence, except for the following changes: TR = 10,000 ms; TD = 0 ms; and within each TR, TI was adjusted to 7900 ms. Mapping for the apparent longitudinal relaxation time T_1app was performed with an ultra-fast Look-Locker echo-planar imaging T1 mapping sequence (26).

Two additional image acquisitions, one three-dimensional (3D) and one 2D, were acquired to aid in multi-subject registration. First, a high-resolution whole brain T1-weighted 3D image was acquired for each subject using MPRAGE (Magnetization Prepared Rapid Acquisition with Gradient-Echo imaging), with the following settings: 160 sagittal slices with field of view (FOV) = 256 × 256 mm²; voxel size = 1 × 1 × 1 mm³; TR = 1500 ms; TI = 800 ms; TE = 2.83 ms; flip angle 15 degrees; and one average. Next, a 2D T1-weighted image was acquired during each MR session using the same slice positions as the perfusion-weighted images and the following additional settings: FOV = 256 × 256 mm²; in-plane resolution 1 × 1 mm²; TR = 300 ms; TE = 3.69 ms; flip angle 60 degrees; and two averages.

Data Processing

Inter-subject Integration

The mean perfusion-weighted image for each condition was used to estimate the map of absolute CBF. The following parameters were used in calculating the absolute CBF map: a longitudinal relaxation time for arterial blood of T_1a =1490 ms; a tissue blood partition coefficient for water of λ=0.9 g⁻¹/mL⁻¹; an RF labeling efficiency of α_π=0.95; and a postlabeling delay time for the first slice of TI = 1400 ms (30). For each subject, maps of task-induced differences in CBF or BOLD were obtained by contrasting all measurements for the two conditions (e.g., visual vs rest). A standard whole brain template (MNI-1mm) was used for subject spatial normalization of the individual data. Subject registration was carried out using the BioImageSuite software package (31) for maps of task-induced changes in the CBF or BOLD values. Two transformations were calculated and used in multiple subject integration: (i) an affine transformation was estimated by co-registering the 2D anatomical image to the high-resolution 3D anatomical image of each individual, and this was then used to transform the individual maps of task-induced changes in CBF or BOLD to the high-resolution 3D anatomical space of that subject; (ii) a nonlinear transformation was used to co-register the high-resolution 3D anatomical image of each individual to the brain template, which enabled warping of all the transformed maps of an individual subject from step (i) to a common brain space. Tri-linear interpolation was used for image re-gridding. The mean, standard deviation, and statistics were estimated in the common template space on the pooled-subject data. In summary, first, the individual maps of task-induced changes in CBF or BOLD were estimated and then transformed to the common reference brain space; and second, voxel-wise contrasts between conditions were estimated in the common space on the pooled-subject data using a t-statistic to test the null hypothesis. Brain regions of interest were defined in the common reference space based on the BOLD t-statistic map in the anesthesia-free condition (see Fig. 1a, top), corrected for multiple-voxel comparisons. Bonferroni correction for multiple voxel comparisons was used in the definition of the auditory and visual regions of interest (ROIs): the per-voxel t-statistics were first calculated on the pooled-subject data and then thresholded with P_vox to obtain a corrected threshold of p = p_vox · V_roi < 0.05 (where V_roi is the size of the ROI considered) (32-36).

We first examined task-induced changes in both BOLD and CBF for the anesthesia-free and anesthesia conditions. Within the visual and auditory ROIs defined, the effect of sevoflurane on baseline CBF and the changes in CBF were evaluated; the coupling between task-induced changes in CBF and BOLD was assessed using analysis of covariance (ANCOVA) and multiple-comparison procedures for the different conditions (37).

Ambiguity in fMRI Data Interpretation in the Presence of an Anesthetic

In Figure 3a, the same observed change in CBF, f₁, brings about different changes in the tissue oxygenation, b₁ and b₂, for the anesthesia-free and anesthetized conditions, respectively. Similarly, a single change in tissue oxygenation, b₂, corresponds to different changes in CBF, f₁ and f₂, for the anesthesia-free and anesthetized conditions, respectively. Because the sevoflurane altered the coupling, neither the change in CBF nor the change in BOLD alone could be used to unambiguously infer the change in oxidative metabolism or neuronal activity in the presence of sevoflurane.

Regional CBF, CBV, BOLD, and Quantitative fMRI

Inferring neuronal activity and metabolism based on the MR-measured vascular changes (e.g., CBF and CBV) and BOLD is the goal of quantitative fMRI. Establishing the link between BOLD signal changes and a more solid index of neuronal activity such as neuronal metabolism has been an area of active research for some time (38-42). Given that the BOLD signal is an epi-phenomenon—a result of the complex interplay between neurophysiologic processes involving oxygen supply, oxygen consumption, and the vascular system – interpreting quantitative fMRI has turned out to be a formidable task, especially in cases where a drug may be involved (10). That the task-induced CBF-BOLD coupling is significantly affected by an anesthetic indicates that quantitative fMRI models must be calibrated under the same anesthetic conditions as the testing that will be undertaken. The regional dependence of the change in coupling between the awake and anesthetic conditions suggests that the calibration in one region cannot necessarily be applied to any other brain region. Such calibration may also be dose-specific as differential flow effects have been observed at different dose levels (43, 44).

To obtain more insight into the possible sources that caused the observed changes in this study, data were simulated using an fMRI quantitative model (39) with the following parameters: α = 0.38, β = 1.5, and M = 0.08. Consistent with our experimental observations, the simulation data, shown in Figure 5, indicate that a decrease in α, or an increase in β and/or M, will decrease the slope of the BOLD-CBF coupling. If the parameters of the fMRI model were not affected by sevoflurane, that is, the model or all related couplings remained constant, the mean relative CMRO2 in the visual ROI would be 0.13 in the awake condition, and drop to 0.06 during anesthesia; in the auditory ROI it changed from 0.028 to 0.013. During anesthesia, however, a decrease in α is very probable, because any increase in the vascular occupancy or CBV, as a result of the vasodilative effect of sevoflurane, likely alters the vaso-reactivity. Let us suppose sevoflurane altered the CBV-CBF coupling such that α was changed from 0.38 to 0.1 for example, then during anesthesia the mean relative CMRO2 would be 0.095 for the visual task and 0.02 for auditory. But the iso-CMRO2 lines shown in Figure 5 indicate that increases in α alone cause only a small decrease in the slope of the regression line, which is not as large a change as observed in this study. Changes in β and/or M are also possible. It should be noted that α, β, and M are inter-dependent (39). Any increase in baseline CBV or β will increase M, while any increase in β or M further decrease the slope of the regression lines. It is very likely that the slope change in the slopes of the regression lines observed in this study resulted from a combination of changes in all these three parameters. One-parameter calibration procedure is simply not enough for the quantitative fMRI model to appreciate oxygen consumption in the presence of drugs.

In most quantitative fMRI practices, two model parameters, α and β, are often borrowed from the previous studies (45, 46) and assumed to be global constants; M is calibrated; and ΔCBF/CBF and ΔBOLD/BOLD are measured. The constant α, for example, indicates unaltered CBF-CBV coupling, which is unlikely in cases where a drug is involved. For cases where the CBF-CBV coupling is altered by a drug, two models should be used: one for the case when the drug is present and the other in the drug-free situation. A handful of recent studies raised additional issues with respect to the use of CO2 for model calibration. It has been demonstrated that a significant source of regional variability arises when using hypercapnia for calibration; incorrect calibrated parameters obviously then propagate error into the rest of the experiment (47, 48). The coupling relationship has also previously been shown to be time-varying and stimulus-dependent (49, 50).

We cannot directly compare our results with previous observations (5, 7, 9) in terms of task-induced changes in CBF or BOLD with and without anesthesia. In general, our findings are not consistent with those animal studies however the studies differ on many dimensions. These include differences in the neurophysiology of the human and rate brain, the anesthetic agents applied, the doses, the lack of an awake condition in the animal studies, the type of stimuli applied and its intensity, in addition to the methodology used to calibrate the models.

Some recently studies on neurovascular coupling showed that vascular changes were altered and they were unnecessarily coupled to neuronal activity in the presence of a large variety of drugs. Dutka et al. investigated changes in task-induced CBF-BOLD coupling in rats before and after multiple exposures of transient hypercapnia (51). During hypercapnia exposures, the baseline CBF increased progressively over time with a trend of increasing significance, whereas the baseline BOLD quickly reached a plateau and the increase was not significant. They also found that the task-induced changes in CBF increased progressively after each exposure and the changes were significant; while the task-induced changes in BOLD were approximately the same after all three transient hypercapnia exposures and none of them are significant compared with the control. Duong et al. used O₂ and CO₂ as chemical stimulants to elicit CBF and BOLD changes in rats which were either awake or anesthetized by isoflurane (52, 53). They found isoflurane attenuated autonomic responses to hypoxia, hypoxia-induced hypocapnia dominated CBF changes, tissues in awake conditions were better oxygenated, and severe hypoxia reduced oxygen metabolism. Their observations suggested marked differences in the BOLD and CBF responses to hypoxia between the awake and anesthetized conditions. In their hypocapnia study, substantially higher BOLD and CBF responses were observed under the awake condition, suggesting that cerebrovascular reactivity was suppressed by the anesthetic agent. In a study done by Liu et al., BOLD responses elicited by a visual stimulus showed distinct characteristics between transit-state and steady-state hypercapnia in normal subjects (54). In rats anesthetized by different anesthetic agents, the coupling between stimulated BOLD and local field potential recordings remained linear; however, the slopes of the regression lines were dependent on the particular anesthetic used—urethane or α-chloralose (55). BOLD increases during hypercapnia were investigated during remifentanil administration and in general remifentanil did not have obvious effects on CO₂ responsiveness in the cerebral vasculature (56) in human. Regional variations of neurovascular coupling were also inspected recently (47, 57) in anesthetized rats, and field potentials and changes in CBF evoked by typical forepaw stimulation were measured. Under 1 MAC isoflurane the field potential and BOLD changes were markedly less than under α-chloralose (45 mg/kg/h) and the relative amplitudes of these changes depended on the stimulation parameters (58), suggesting that different anesthetics affect the field potential and CBF responses in different manners. In general there appears to be a growing consensus suggesting an altered coupling relationship across a large variety of agents.

Some Methodological Considerations

Physiological and physical limits on the number of slices for each radio frequency labeling make whole brain coverage difficult (59-61). In a recent study (62), single slice PASL MRI measurements were performed as a function of in-plane spatial resolution and postlabeling delay, and their results indicated that, when using postlabeling delays shorter than 1500 ms, higher MRI gray matter flow values may be observed owing to signal contamination from the remaining arterial blood water label. For delays above 1500 ms, regional PASL-based CBF values from frontal gray matter and occipital gray matter were comparable with PET-based measurements which can be obtained by using spatial resolutions comparable with PET (5–7.5 mm in-plane). At high resolution (2.5 × 2.5 × 3 mm³), compared with that at low resolution (7.5 × 7.5 × 3 mm³), gray matter CBF values were found to increase by 10–20%, and this difference was attributed to a reduction in partial volume effects. To obtained full coverage of the visual and auditory cortices, the slice thickness was set to 8 mm, with 2 mm gap between slices. A calculation based on the voxel sizes suggests that an 8–16% underestimation for CBF is possible with this resolution. Some of the disadvantages of ASL imaging include low signal-to-noise ratio (63), contamination from the intravascular signal (64), variability of the arterial transit time (65-67), and physiological and physical limits on the number of slices for each radio frequency (RF) label. Recently Woolrich et al. (68) modeled the double-echo ASL process using a general linear model and a Bayesian statistical inference model. Their data indicate that the fractional change in BOLD obtained by averaging tag/control pairs from a single-echo ASL experiment could be underestimated by up to 25%. This underestimation is caused by the residual signal components of static magnetization from the labeled or unlabeled blood. Because in this study, we focused on the anesthetic effects on the task-induced CBF-BOLD coupling, relative to the anesthesia-free condition, the bias in CBF or BOLD measurement will unlikely affect our observation if it is present similarly with or without anesthesia. Further studies are required to clarify these issues.