Functional magnetic resonance imaging of blood oxygenation level-dependent signal changes offers a very promising approach to investigate activated neural networks during anesthesia.
Sixteen healthy male volunteers, assigned into two groups of eight subjects (isoflurane group, control group), were investigated by functional magnetic resonance imaging during different experimental conditions. The isoflurane group successively breathed air (baseline condition), isoflurane in air (0.42 vol% inspiratory; isoflurane condition) and air again (recovery condition) while performing a visual search task, whereas the control group breathed air during all experimental conditions. Functional magnetic resonance images were acquired during the entire experimental session. In addition, reaction times and error rates were recorded.
A significant isoflurane-related decrease (z > 3.1 corresponding to P < 0.001) in task-induced brain activation was found in three distinct cortical regions: the right anterio-superior insula (Talairach coordinates: x = 32, y = 22, z = 8) and the banks of the left and right intraparietal sulcus (Talairach coordinates: x = -34, y = -36, z = 32; x = 22, y = -60, z = 41, respectively). Subcortical structures (lateral geniculate nucleus) and the primary cortices (motor cortex, visual cortex) were not affected. All measured parameters indicated a nearly complete recovery of the affected networks within 5 min.
Our findings indicate that subanesthetic isoflurane affected task-induced activation in specific neural networks rather than causing a global decrease in functional activation.
FUNCTIONAL imaging modalities such as positron emission tomography or single photon emission computed tomography have revealed many details concerning the effect of common anesthetic agents (e.g. , isoflurane) on cerebral blood flow and metabolism. 1–8The effect of anesthesia on task-induced brain activation, however, is less well established. Functional magnetic resonance imaging (fMRI) offers a very promising approach to investigate this phenomenon.
In comparison to positron emission tomography or single photon emission computed tomography, fMRI provides higher intrinsic spatial and temporal resolution, and it does not require the injection of exogenous tracers. 9,10Therefore, a subject can perform a variety of different tasks during various experimental conditions (e.g. , breathing subanesthetic concentrations of a volatile anesthetic agent vs. breathing air) within the same imaging session. Most fMRI studies are based on measuring changes in the blood oxygenation level–dependent (BOLD) signal that arise from a complex interplay between cerebral hemodynamics and oxidative metabolism. 10,11
Investigation of sedative effects produced by anesthetics in conscious volunteers offers the possibility to study the interaction of these drugs with cognitive processes. It should therefore provide further evidence for a better understanding of the general mechanism of anesthetic action. In the present study, we investigated the effect of subanesthetic isoflurane on functional brain activation induced by a visual search task.
Materials and Methods
Sixteen male volunteers between the ages of 20 and 35 yr with no history of trauma or neurologic disease were investigated. All subjects were right-handed. Females were excluded from the study because of the possibility of sex differences in task-induced activation of the brain. 12The study was approved by the local ethics committee of the Medical Faculty, University of Leipzig. Written informed consent was obtained from each subject before the study. Subjects were instructed not to eat or drink 6 h before the experiment and to refrain from consuming alcohol or using any medication on the day of the experiment. All volunteers took part in a familiarization session before the day of the study in which all tasks and proceedings were practiced.
Experimental Design and Procedure
The volunteers were positioned supine on the gantry of the scanner. A polyethylene mouthpiece with 4 m of flexible tubing was used to connect the subjects to an MR-compatible ventilator. We used a semi-open circuit with a fresh gas flow of 10 l/min. Subjects breathed spontaneously throughout the experiment. To facilitate inspiration, a pressure support of 200 Pa was applied. Nasal breathing was precluded by a nose clamp. Foam padding and a “stiff neck” were used to limit head movements. Heart rate, arterial oxygen saturation, end-tidal partial carbon dioxide pressure, and isoflurane concentrations (inspiratory and expiratory) were continuously monitored during the entire session using an MR-compatible monitoring system.
A visual search task was used for stimulation. The subjects were first given a cue comprising a red “L” shape of a given orientation (0°, 90°, 180°, 270°). After a delay of 1.2 s, a 10 × 10 array of randomly rotating white “L”s was presented on the black background for a period of 2.2 s. Then four of these shapes became stationary and changed their color from white to red, while the remainder continued to rotate. The possible orientations of the stationary “L” shapes were 0°, 90°, 180°, 270°. The stationary shapes always occurred at the same position. The distance from the position of the stationary shapes to the center of the screen was the same, as was the distances between the stationary shapes. The subjects had to determine whether the orientation of any one of the stationary shapes matched that of the cue and had to give a yes–no response using a two-button key pad. Subjects used the index finger (yes response) and middle finger (no response) of the right hand. To introduce a further level of complexity, the coding of the key depended on the orientation of the cue “L” shape. If it was rotated by 90° or 270°, the first key corresponded to “yes” and the second to “no.” If it was rotated by 0° or 180°, the coding of the keys was reversed. Reaction times and error rates were recorded. The reaction limit was 2 s. Answers exceeding this limit were discarded. The stimulation protocol consisted of five blocks of control interleaved with five blocks of visual stimulation. During the control task, the subject had to fixate on a gray dot positioned at the center of a black screen. The duration of each block was 30 s, and the entire stimulation experiment lasted 5 min. The stimulation program was written in ERTS (BeriSoft Cooperation, Frankfurt am Main, Germany). Using a liquid crystal display projector, the display was projected onto a translucent screen positioned inside the magnet bore behind the subject’s head. The screen was viewed using mirror glasses that were equipped with correction lenses as necessary.
The fMRI imaging protocol is illustrated in figure 1. The 16 volunteers were divided into two groups, each consisting of eight subjects. The first group took part in the actual isoflurane experiment (isoflurane group), and the second group took part in a control experiment (control group). The isoflurane experiment lasted 30 min and included three 5-min stimulation blocks. During the first stimulation block (baseline condition [BC]) performed at 0 to 5 min, the subjects breathed air. The vaporizer was then adjusted at 0.42 vol% isoflurane so that the participants were breathing isoflurane in air at 5 to 20 min. The second block (isoflurane condition [IC]) was performed at 15 to 20 min. A tactile stimulation was performed on the big toe 30 s before the IC started to ensure that every subject was aware of the task and was not asleep during this part of the session. After the IC, the vaporizer was closed and the subjects breathed air again. The third block (recovery condition [RC]) was performed at 25 to 30 min to investigate the recovery from the preceding IC. Functional magnetic resonance images were acquired in a continuous fashion during the entire experimental session with a repetition time of 2 s. The control experiment was identical to the isoflurane experiment, except that the subjects breathed air during the entire experiment.
Functional Magnetic Resonance Imaging
All experiments were performed on a 3 Tesla whole-body scanner equipped with a quadrature birdcage head resonator and an actively shielded whole-body gradient set capable of switching 30 mT/m in 500 μs. The MPIL software package developed at the Max-Planck-Institute of Cognitive Neuroscience in Leipzig was used to drive the console.
Functional magnetic resonance imaging was performed on the basis of a conventional blipped echo-planar imaging sequence. 13Imaging parameters were as follows: matrix size, 64 × 64; field of view, 19.2 × 19.2 cm; in-plane spatial resolution, 3 × 3 mm; slice thickness, 5 mm; repetition time, 2 s; echo time, 30 ms. Sixteen axial slices were acquired covering the entire brain.
Before the fMRI experiment, high-resolution anatomic images were obtained from the same slices using a modified driven equilibrium Fourier transform (MDEFT) imaging sequence 14with the following imaging parameters: matrix size 256 × 256; field of view, 19.2 × 19.2 cm; in-plane spatial resolution, 0.75 × 0.75 mm; slice thickness, 5 mm; inversion time, 0.65 s; recovery time, 0.65 s; echo time, 6.1 ms; total acquisition time, 5.5 min.
In a different imaging session, a three-dimensional (3D) data set was acquired using 3D MDEFT imaging. 15Imaging parameters were as follows: matrix size, 256 × 256 × 128; field of view, 25.0 × 25 × 19.2 mm; spatial resolution, 0.98 × 0.98 × 1.5 mm; inversion time, 0.65 s; recovery time, 0.65 s; echo time, 6.1 ms; total acquisition time, 19 min.
Image Processing and Statistical Analysis
Functional data sets were acquired with a repetition time of 2 s per volume. Since each volume consisted of 16 slices acquired in a serial manner, the temporal delay between slices was corrected by sinc-interpolation. In addition, data were corrected for bulk motion using a rigid-body realignment algorithm. Image time series were then displayed as a movie to ensure that no substantial through plane motion was present. A spatial Gaussian filter with a full width at half maximum of 5 mm was applied to each image. Finally, a baseline correction of the image series was performed by applying a high-pass filter with a cutoff frequency of 1/120 Hz.
Statistical evaluation was based on a least-squares estimation using the general linear model for serially autocorrelated observations. 16–19First, for each individual subject, statistical parametric maps were generated. The design matrix was generated with a boxcar (square wave) function and a response delay of 6 s. The model equation, including the observation data, the design matrix, and the error term, were convoluted with a Gaussian kernel with a dispersion of 4 s (full width at half maximum). The model includes an estimate of temporal autocorrelation that is used to estimate the effective degrees of freedom. For each experimental condition (BC, IC, RC), contrasts between visual search and baseline task were calculated using t statistics. Subsequently, t values were converted to z values. The resulting z maps were coregistered onto the corresponding 3D anatomic data sets using the two-dimensional high-resolution MDEFT images as a reference, and finally transformed into a stereotactic Talairach coordinate system. 20
The primary statistical analysis described above yielded 48 z maps (Zc,s,g) corresponding to the three conditions (c = BC, IC, RC), the 8 subjects per group (s = 1, 2...8), and the 2 groups (g = isoflurane group, control group). It is reasonable to assume that differences between experimental conditions in the isoflurane group are not solely caused by anesthesia but may also reflect carryover effects (e.g. , habituation, fatigue, learning). However, using an appropriate statistical design, separation of isoflurane from carryover effects can be achieved by comparison with a suitable control group. For this purpose, we used the concept of the “split-plot” factorial design, 21using planned contrasts to increase statistical sensitivity. We expected a specific interaction between group and condition: isoflurane effects should be dominant during the IC and may be still present during the RC of the isoflurane group, whereas carryover effects in both groups should be the same. The latter assumption is reasonable because both groups underwent the same experimental procedure, except that the control group was breathing air during the entire experiment. To separate isoflurane from carryover effects, we calculated the following planned contrasts between the experimental conditions (c = BC, IC, RC) from the individual z maps for each of the 16 subjects:
These contrasts are quantitative measures that describe the difference between the IC in comparison to the mean of the BC and RC. In the isoflurane group, these differences may be caused by both isoflurane and carryover effects, whereas in the control group, differences may only arise as a result of carryover effects. The effect of isoflurane can therefore be separated from carryover effects by comparing the contrasts obtained in both groups. For this purpose, we applied a t test for independent samples. The resulting t map was finally converted into a z map. Hereafter we refer to this z map as “interaction map” because it displays the isoflurane-induced interaction between the two groups.
However, we were not only interested in the isoflurane-induced changes in the task-induced activation pattern, but also in the activation pattern itself. We therefore calculated six activation maps (Ac,g) corresponding to the three experimental conditions (c = BC, IC, RC) and the two groups (g = isoflurane group, control group) by multisubject analysis of the individual z maps 22:
Multisubject analysis was used for two reasons: to increase statistical sensitivity and to reduce the amount data.
Finally, the interaction map showing isoflurane-related changes in task-induced activation was superimposed onto an activation map (Ac=BC,g=IG) that indicates the entire task-induced activation pattern during the BC. Significant task-induced activation (z > 5.5, uncorrected for multiple comparisons) was color-coded in uniform blue, whereas a red–yellow color scale was used to display the level of significance in isoflurane-related changes. Correction for multiple comparison was omitted because of this reasonable high threshold. Color-coded regions were then overlaid onto a 3D anatomic data set that was obtained by multisubject averaging over all individual 3D anatomic data sets (n = 16) in the Talairach coordinate system.
Image processing and statistical analysis were performed on the basis of the LIPSIA software package developed at the Max-Planck-Institute of Cognitive Neuroscience in Leipzig. 23The routine for generating the interaction map was written in IDL (Research Systems, Boulder, CO).
Physiological data (heart rate, breathing rate, arterial oxygen saturation, end-tidal carbon dioxide values) of the isoflurane group were compared between experimental conditions (BC, IC, RC). A one-way repeated-measures analysis of variance was used to compare normally distributed data (heart rate, breathing rate, arterial oxygen saturation), whereas a repeated-measures analysis of variance on ranks (Friedman test) was applied to analyze data that were not normally distributed (end-tidal carbon dioxide).
Average changes in reaction times were calculated for each group (isoflurane group, control group) comparing IC versus BC and RC versus BC. The t statistic was used to analyze the statistical significance of the results. In addition, the error rates (number of actual errors/number of possible errors) were calculated for each group and condition. For a statistical analysis of the results, we applied a one-way analysis of variance on ranks (Kruskal-Wallis test), because data were not normally distributed.
Functional Magnetic Resonance Imaging
To separate isoflurane-related effects from carryover effects (e.g. , habituation, fatigue, learning), we performed a statistical comparison between the isoflurane group and the control group (see Methods). The resulting interaction map is depicted in figure 2. The red-to-yellow color scale displays the level of significance. The threshold was set to z = 3.1, which corresponds to a probability of P = 0.001. Significant decrease in activation is visible in three distinct cortical regions: the right anterio-superior insula, the left intraparietal sulcus (ascending–horizontal branch), and the right intraparietal sulcus (horizontal branch); a 3D representation of these regions is shown in figure 3. Voxels with a significant increase in task-induced activation were not found at this level of significance. Regions of significant task-induced activation (z > 5.5) that were not affected by isoflurane are shown in uniform blue. These regions include the lateral geniculate nucleus, the striate and extrastriate visual cortex, the motor cortex, the supplementary motor area, and the left putamen, as well as large parts of the anterior insula and the banks of the intraparietal sulcus.
A quantitative analysis of isoflurane-related changes is shown in figure 4. Representative voxels were selected in the three regions affected by isoflurane (anterio-superior insula, left intraparietal sulcus, right intraparietal sulcus) as well as in three regions not affected by isoflurane (lateral geniculate nucleus, visual cortex, motor cortex). For each group (isoflurane, control), the group-specific z values calculated according to equation 2are plotted versus the experimental condition (BC, IC, RC). In the anterio-superior insula, left intraparietal sulcus, and right intraparietal sulcus, the isoflurane group shows a clear decrease in functional activation during the IC in comparison to the control group. During the RC, functional activation almost reached BC. A statistical comparison of this “decrease–recovery” behavior between both groups using a split-plot factorial design (see Methods) yielded significant differences (z > 3.1). However, no significant isoflurane-related effects were detectable in visual cortex, motor cortex, and lateral geniculate nucleus. The plot corresponding to motor cortex nicely demonstrates that a control group is indispensable for a correct interpretation of isoflurane-related effects. The isoflurane group shows a distinct decrease–recovery behavior that could be misinterpreted as an isoflurane-related effect. However, the control group shows a comparable time course. The observed decrease–recovery behavior is therefore a result of carryover effects (e.g. , habituation, fatigue, learning) and not related to isoflurane.
Reaction Times and Error Rates
Figure 5shows the average change in reaction times comparing IC versus BC and RC versus BC. The isoflurane group showed a significant increase of 26.5 ± 10.6% during the IC (P < 0.0002). During the RC, reaction time was still prolonged: 4.9 ± 3.7% (P < 0.01). No significant changes in the reaction times were found for the control group. The corresponding changes in reaction time of the control group are not significant.
The average error rates under BC, IC, and RC were 5.7 ± 8.6, 25.1 ± 23.1, and 14.9 ± 18.0%, respectively, for the isoflurane group and 6.5 ± 3.7, 10.0 ± 7.1, and 8.5 ± 6.6%, respectively, for the control group. Differences between error rates are not significant as proven by a statistical comparison of both groups.
Behavioral State of the Subjects
No subject fell asleep during the IC, as proved by reaction times. Most of the subjects compared their behavioral state during the isoflurane experiment with a slightly intoxicated state. During an interview conducted after the experimental session, they reported that details of the search task became blurred.
Physiologic data (heart rate, oxygen saturation, and breathing rate) of the isoflurane group did not show any significant difference between the three experimental conditions (table 1). Breathing of isoflurane resulted in an inspiratory concentration of 0.42 vol% at the beginning and 0.44 vol% at the end of the IC. The corresponding expiratory concentrations were 0.29 and 0.31 vol%. The time courses of the inspiratory and expiratory concentrations of isoflurane during the entire session are shown in figure 6. The end-tidal carbon dioxide values were slightly increased (1.5 mmHg) during the IC (P < 0.05;fig. 7).
In most brain regions, our study reveals no isoflurane-related changes in task-induced activation, including all areas associated with primary information processing (e.g. , lateral geniculate nucleus, primary visual cortex, motor cortex;fig. 2). However, in three focal cortical regions (anterio-superior insula, left intraparietal sulcus, and right intraparietal sulcus), isoflurane significantly decreased task-induced activation (fig. 4). This finding indicates that subanesthetic isoflurane affected task-induced activation in specific neural networks rather than causing a global decrease in functional activation.
The parietal association areas, i.e. , the intraparietal sulci, are known to be involved in visuo-spatial attention. 24,25Recently, the intraparietal sulcal cortex has been suggested to reflect a common neural substrate underlying multiple modes of visual selective processing and attention. Furthermore, there is evidence that the parietal cortex, particularly the banks along the intraparietal sulci, represent an area responsible for suppressing irrelevant distractors to maintain visual attention. 25The anterior insula plays an important role in visual working memory as shown by Courtney et al. 26Dove et al. 27found bilateral activation of the anterior insula related to a task-switching paradigm. Pollmann and von Cramon 28reported activation of the anterio-superior insula (especially in the right hemisphere) associated with the difficulty of a visual search task.
The quantitative analysis of the affected regions (fig. 4) shows an isoflurane-related decrease in task-induced activation that is accompanied by a significant increase in the average reaction time (fig. 5). This increase in reaction time clearly demonstrates an impairment of cognitive performance. Despite this impairment, all subjects were conscious and able to respond to the stimuli. Considering the isoflurane-related decrease in task-induced activation in the parietal association cortex and the insula (see anterio-superior insula, left intraparietal sulcus, and right intraparietal sulcus regions in fig. 4), we interpret the observed deterioration of cognitive performance by a selective impairment of attention and working memory processes in the visual domain. However, the preserved functional activation at the thalamic level in the early visual cortices and motor cortices during isoflurane suggests a regular perception and processing of the stimuli in these regions (see lateral geniculate nucleus, visual cortex, and motor cortex regions in fig. 4).
With regard to the difference in spatial extent and threshold used for displaying the activation and interaction maps (figs. 2–4), the following two factors should be considered. First, the activation maps were derived using a fixed-effect model, whereas for the interaction maps a second-level analysis was required (see Methods). Second-level analysis usually shows lower statistical power, mainly because error terms are estimated on the basis of interindividual rather than intraindividual variance. Second, the activation maps depict the difference between an experimental task and a resting baseline, whereas the interaction maps depict the interaction between identical tasks. For the interaction maps, the conditions compared are more similar than they are in the case of the activation maps.
The measured changes in reaction times and task-induced activation almost reached baseline within 5 min after exposure to 0.42 vol% isoflurane (figs. 4 and 5), which suggests a nearly complete recovery of cognitive functions. Recently published investigations based on psychomotor tests have demonstrated a similar rapid recovery after inhalation of subanesthetic isoflurane in healthy young volunteers. 29,30
For the interpretation of our results, a discussion of the underlying biophysical mechanisms 31–33and the possible interactions with isoflurane seems useful. Increased neural activity is accompanied by a regional increase in oxidative metabolism, cerebral blood flow, and cerebral blood volume. All of these parameters affect the regional deoxyhemoglobin concentration. Because of its paramagnetic properties, deoxyhemoglobin causes distortions of the local magnetic field that can be visualized by BOLD-based fMRI.
Changes in the regional deoxyhemoglobin concentration arise from a complex interplay between cerebral hemodynamics and oxidative metabolism. 10,11An increase in oxygen consumption or venous blood volume results in an increase in deoxyhemoglobin and in a decrease of the BOLD signal, whereas an increase in blood flow produces a decrease in the regional deoxyhemoglobin concentration, resulting in an increase in the BOLD signal. Typically, the BOLD signal is found to be increased during periods of neural stimulation. This implies that changes in the deoxyhemoglobin concentration caused by blood flow exceed changes caused by oxidative metabolism and cerebral blood volume. Therefore, regional cerebral blood flow is the dominant parameter in BOLD-based fMRI.
The measurable BOLD signal change depends on two different factors. First, the change in neural activation induced by the stimulus, and second, the coupling between oxidative metabolism and cerebral hemodynamics. At this point the question arises whether the measured isoflurane-related decrease in functional activation solely reflects a decrease in neuronal activation or was partly caused by a change in the coupling between oxidative metabolism and cerebral hemodynamics. As demonstrated in rats, however, the coupling between cerebral blood flow and metabolism is intact during isoflurane anesthesia. 34This suggests that the measured changes in BOLD contrast during isoflurane reflect regional changes in neural activity rather than alterations in the coupling between cerebral hemodynamics and oxidative metabolism.
For the isoflurane group, end-tidal carbon dioxide was slightly increased (1.5 mmHg) during the IC (fig. 7). We therefore cannot exclude a moderate increase in baseline cerebral blood flow (i.e. , the cerebral blood flow value during the baseline task). An additional moderate increase in baseline cerebral blood flow could result from the vasodilatory properties of the anesthetic agent. 35However, the magnitude of absolute cerebral blood flow changes induced by visual or motor stimulation was demonstrated to be independent of the baseline values. 36–38During hypercapnia induced by breathing of 6% CO2, Maximilian et al. 36found a fairly constant task-induced cerebral blood flow response among highly variable baseline values in the visual cortex. Similar results were obtained by Li et al. 37in the motor cortex. Shimosegawa et al. 38reported a significant effect of hypocapnia on task-induced cerebral blood flow response in the visual cortex but found no differences between normocapnia and hypercapnia. An additional argument is that even if global hemodynamic changes had a significant effect on the measured BOLD signal changes, this would be global in nature and therefore could not explain the observed regional changes in task-induced activation.
The effect of isoflurane on task-induced activation is dependent on the dose of the anesthetic agent. At 0.3% (expiratory), Logothetis et al. 39found no significant isoflurane-related changes in the monkey brain during visual stimulation. At a comparable expiratory concentration (fig. 6), we observed a significant decrease in task-induced activation in three distinct cortical regions of the human brain; however, most brain regions were not affected by isoflurane (fig. 2). No activation was found at 1.3% in the human somatosensory cortex during sensory stimulation 40and at 2.0% in the monkey cortex during visual stimulation. 39Combination of these results supports the hypothesis of a dose-dependent gradual decrease of task-induced activation. The question of how anesthetic drugs affect brain activation during the transition from the conscious to the unconscious state opens a fascinating field for future research.
In conclusion, subanesthetic isoflurane affected information processing during a visual search task by an impairment of distinct parietal cortical structures and of the right insular cortex, resulting in a reduced cognitive performance. The information transfer and processing in subcortical structures at the thalamic level (lateral geniculate nucleus) as well as in the primary cortices (primary visual cortex, motor cortex) was not impaired. These results suggest that neural networks with highly differentiated functions in the association cortices are more sensitive to sedative concentrations of isoflurane than cortical areas with a lesser degree in functional specialization. Functional data as well as reaction times demonstrate a nearly complete recovery of the brain from subanesthetic isoflurane within 5 min.
The authors thank Derk Olthoff, M.D., Ph.D. (Department of Anesthesiology and Intensive Care Therapy, University of Leipzig, Leipzig, Germany), D. Yves von Cramon, M.D., Ph.D. (Max-Planck-Institute of Cognitive Neuroscience, Leipzig, Germany), David G. Norris, Ph.D. (Max-Planck-Institute of Cognitive Neuroscience), Stefan Pollmann, Ph.D. (Max-Planck-Institute of Cognitive Neuroscience), and Volker Bosch, Ph.D. (Max-Planck-Institute of Cognitive Neuroscience) for valuable comments; and Grit Wenge (Max-Planck-Institute of Cognitive Neuroscience), Katrin Wiesner (Max-Planck-Institute of Cognitive Neuroscience), Sabine Busker-Rössig (Max-Planck-Institute of Cognitive Neuroscience), and Pjotrr S. Zbgyev (Max-Planck-Institute of Cognitive Neuroscience) for technical support.