Metabolic connectivity mapping reveals effective connectivity in the resting human brain

Edited by Peter A. Bandettini, National Institutes of Mental Health, Bethesda, MD, and accepted by the Editorial Board December 4, 2015 (received for review July 13, 2015)
December 28, 2015
113 (2) 428-433

Significance

Noninvasive brain imaging of humans identifies prominent networks related to sensory and cognitive functions in the resting state; however, the signaling hierarchy and directionality among these networks remain largely unknown. Integrating simultaneously recorded measures for network identification and regional energy metabolism, we propose metabolic connectivity mapping (MCM) as a novel measure to reveal signaling directionality in the human brain. Comparing simple “eyes closed” and “eyes open” conditions in healthy subjects, MCM revealed stable bidirectional signaling among visual cortices and top-down signaling from parietal control regions. Additional top-down signaling from a salience network occurs only during the eyes open condition. We propose MCM as a measure for investigating signaling directions among brain networks in healthy and diseased brains.

Abstract

Directionality of signaling among brain regions provides essential information about human cognition and disease states. Assessing such effective connectivity (EC) across brain states using functional magnetic resonance imaging (fMRI) alone has proven difficult, however. We propose a novel measure of EC, termed metabolic connectivity mapping (MCM), that integrates undirected functional connectivity (FC) with local energy metabolism from fMRI and positron emission tomography (PET) data acquired simultaneously. This method is based on the concept that most energy required for neuronal communication is consumed postsynaptically, i.e., at the target neurons. We investigated MCM and possible changes in EC within the physiological range using “eyes open” versus “eyes closed” conditions in healthy subjects. Independent of condition, MCM reliably detected stable and bidirectional communication between early and higher visual regions. Moreover, we found stable top-down signaling from a frontoparietal network including frontal eye fields. In contrast, we found additional top-down signaling from all major clusters of the salience network to early visual cortex only in the eyes open condition. MCM revealed consistent bidirectional and unidirectional signaling across the entire cortex, along with prominent changes in network interactions across two simple brain states. We propose MCM as a novel approach for inferring EC from neuronal energy metabolism that is ideally suited to study signaling hierarchies in the brain and their defects in brain disorders.
Complex cognition emerges by integrating upstream sensory information with feedback signaling from higher cortical regions (14). Networks related to sensory processing or cognition reliably occur in the human brain even at rest (5, 6). These networks are identified by synchronous signal fluctuations, or functional connectivity (FC), among brain regions when neuronal activity is recorded by functional magnetic resonance imaging (fMRI). In recent years, various FC patterns have emerged as reliable indicators of different brain states, because they have been found to adapt to recent behavior or cognition (712) and to be disrupted in patients suffering from specific psychiatric disorders (13, 14). Further knowledge about important aspects of cognition and diseases could be gained from a better distinction between feedback and feedforward communication. Our understanding of the signaling hierarchy in different brain states remains incomplete, however.
Although FC captures correlations among neuronal signals, only effective connectivity (EC) describes the influence exerted by one neuronal system over another (15). Recent approaches to modeling EC during different brain states appear promising (16, 17), but face problems inherent to fMRI. First, EC is estimated directly from the time-varying fluctuations or cross-spectra of the observed fMRI signal, and thus is prone to confounds from different hemodynamic responses across groups, particularly when studying patient populations (15, 17). Second, analyses are usually restricted to a limited number of brain regions, owing to the need for complex computations. Here we propose a novel approach integrating FC with simultaneously measured energy metabolism from positron emission tomography (PET) to derive a voxel-wise, whole-brain mapping of EC in humans.
Energy consumption is an essential aspect of neuronal communication. Consistently across species, the greatest amount of energy metabolism is dedicated to signaling, with the remaining part dedicated to housekeeping functions (18). Up to 75% of signaling-related energy is consumed postsynaptically, i.e., at the target neurons (1922). Scaled to the systems level, we assume that an increase in local metabolism reflects an increase in afferent EC from source regions. We hypothesize that the spatial profile of this relationship is expressed in terms of spatial correlations between metabolic activity and long-range FC, which we term metabolic connectivity mapping (MCM). We simultaneously acquired fMRI and PET data for the glucose analog 18F-fludeoxyglucose (FDG) during two different brain states, as reported previously (10). In individual subject space, we performed spatial correlation analyses of voxel FC and FDG to test whether the metabolic profile indicates the target area of communication between functionally connected regions (Fig. 1).
Fig. 1.
MCM reveals EC in the human brain. (A) FC reveals undirected pathways of synchronous fMRI signal fluctuations between two regions, X and Y. For each subject, FC is calculated as the temporal correlation, [r] between the cluster time series. In our example, we identified FC pathways during conditions of eyes closed and eyes open. (B) In a next step, we identified EC (i.e., the directionality of signaling) in a given functional pathway. For each region, we calculated the spatial correlation, [r], between voxel FC and FDG values of simultaneously acquired fMRI and PET data. According to cellular recordings (see text), the majority of signaling-related energy is consumed postsynaptically, i.e., at the target region. Thus, MCM identifies afferent EC in the signaling pathway between region X and region Y.
Vision is the only sensation that can be interrupted volitionally in a natural way. Opening the eyes is a fundamental behavior for directing attention to the external world, i.e., changing from an interoceptive state to an exteroceptive state (3, 23, 24). Current knowledge of the signaling hierarchy in the extended visual system has emerged from animal and tracer studies. These data reveal reciprocal (bottom-up and top-down) connections along the ventral and dorsal visual stream (25), including top-down projections from frontal back to early visual cortices (3, 4, 26, 27). To test this signaling hierarchy in humans, we scanned healthy human subjects in two brain states, lying with either eyes closed or eyes open in darkness, and calculated EC using our integrated approach. Consistent with previously reported data, MCM revealed persistent and bidirectional interactions between visual stream areas during both the “eyes closed” and “eyes open” conditions, but frontal top-down modulation of early visual areas only during the eyes open condition.
In the present study, we used FDG to inform undirected FC from fMRI with a directional measure of postsynaptic neuronal activity. Our results indicate that the integrated measure of MCM serves as a proxy for EC in brain states. Our approach might be particularly useful for investigating other signaling hierarchies in higher cognition or in brain disorders involving, e.g., hippocampal-cortical circuits in Alzheimer’s disease (28) or fronto-midbrain interactions in major depression (29).

Results

We investigated the signaling hierarchy between an early visual area and functional regions across the entire cortex during conditions of eyes closed and eyes open in a three-step procedure. We first identified regions of interest (ROIs) in visual and association cortices, then determined the FC pathways between the early visual ROIs and all other ROIs, and finally integrated voxel FC and FDG values to reveal the directionality of signaling along these functional pathways via MCM.

Regions Belonging to Visual Stream and Prefrontal Association Networks.

Using independent component analysis (ICA) of fMRI data, we parcellated the cortex into clusters sharing characteristic temporal BOLD dynamics. Each cluster had to fulfill certain properties with respect to size and stability of occurrence (Materials and Methods). We then mapped these clusters onto recently defined networks from a 1,000-subject dataset (Fig. 2, colored outline). Previous work identified two networks (dark violet, green) as part of an extended visual hierarchy and three networks covering prefrontal association areas (30). According to this nomenclature, we identified five ROIs along ventral and dorsal visual streams (Fig. 2, Left) and five ROIs in frontal association networks (Fig. 2, Right). Table 1 lists all anatomic structures covered by each ROI. We selected the calcarine sulcus (CaS) as the early visual area for all subsequent analyses of FC and MCM, because it fully covers the area of greatest metabolic increase when subjects open their eyes (10) (Fig. 2, Middle, blue).
Fig. 2.
Brain regions belonging to visual stream and prefrontal association networks. Spatial maps (P < 0.05, corrected) are projected onto the inflated cortical surface of a standard brain. Major networks of a recent 1,000-subject analysis are illustrated in the middle panel (top, lateral surface; bottom, medial surface) using different colors and serve as a reference for the assignment of each of our ROIs to its closest match. Reprinted from ref. 30. Ventral/dorsal visual stream ROIs include the CaS, StC, ExC, ITG, and SPL. Prefrontal association network ROIs include the PFCa, INS, PFCdl, MCC, and PFCvm. Table 1 lists the anatomic structures covered by each ROI. The CaS was chosen as the early visual region for further analyses, because it fully covers the cluster of greatest increase in metabolic activity during the eyes open condition from an earlier analysis (10).
Table 1.
Brain regions and associated networks as illustrated in Fig. 2
LabelAnatomyNetwork
CaSCalcarine sulcusVisual (dark violet)
StCStriate cortex, lingual gyrus, lateral occipitalVisual (dark violet)
ExCExtrastriate cortex, inferior parietalVisual (dark violet)
ITGFusiform gyrus, inferior temporalVisual (dark violet)
SPLSuperior parietal/frontal eye fieldsSpatial attention (green)
PFCaAnterior prefrontal cortexSalience (purple)
INSInsular cortex, frontal operculumSalience (purple)
PFCdlDorsolateral prefrontal cortexCentral executive (yellow)
MCCMiddle cingulate cortexSalience (purple)
PFCvmVentromedial prefrontal cortexDefault mode (red)

FC Pathways Between Regions.

In a second step, we identified FC pathways between the CaS and all other ROIs via pairwise temporal correlations between individual cluster time-series (FCcl) from the ICA. For each subject, we first calculated the FCcl between the CaS and all ROIs of the visual stream and submitted these values to ANOVA with region [striate cortex (StC), extrastriate cortex (ExC), inferior temporal gyrus (ITG), and superior parietal lobe (SPL)] and condition (eyes closed/open) as factors (Fig. 3A). We found a main effect for region (F1,21 = 3.41, P = 0.02), but not for condition (F1,21 = 3.27, P = 0.075), and no interaction of region and condition (F3,19 = 2.14, P = 0.1). One-sample t tests, however, showed a positive FCcl between the CaS and each of the visual stream ROIs across both conditions (StC: t21 = 5.72, P = 1.1 × 10−5; ExC: t21 = 10.06, P = 1.7 × 10−9; ITG: t21 = 6.62, P = 1.5 × 10−6; SPL: t21 = 8.12, P = 5.8 × 10−8). Thus, persistent functional pathways between CaS and all regions along the ventral and dorsal visual streams exist during both the eyes closed and eyes open conditions.
Fig. 3.
FC pathways during the eyes closed and eyes open conditions. (A) Stable FC is seen between the CaS (image to the right of the graph) and all other visual regions (StC, ExC, ITG, and SPL) independent of condition (results of one-sample t tests; see text). (B) FC between the CaS and all regions of the salience network (purple) increases only during the eyes open condition (two-sample t tests; PFCa: P < 0.05; INS: P < 0.0005; MCC: P < 0.005). The two other regions (PFCdl and PFCvm) show no FC with the CaS (P > 0.1). Note that only salience regions INS and MCC show increased FC when a stringent Bonferroni correction is applied for possibly nine two-sample t tests in this FC analysis. Boxplots indicate median, 25th–75th percentiles (box), and extreme data points (whiskers).
Similarly, we tested whether functional pathways exist between the CaS and any prefrontal association region. We calculated the FCcl between the CaS and each frontal ROI and submitted these measures to ANOVA with region [anterior prefrontal cortex (PFCa), insular cortex (INS), dorsolateral prefrontal cortex (PFCdl), middle cingulate cortex (MCC), and ventromedial prefrontal cortex (PFCvm)] and condition (eyes closed/open) as factors. We found main effects for region (F1,21 = 11.1, P = 1.7 × 10−7) and condition (F1,21 = 19.14, P = 3.0 × 10−5), along with an interaction of region and condition (F4,18 = 5.54, P = 0.0005). This means that the FCcl between the CaS and prefrontal ROIs strongly differs for selected regions between conditions. Post hoc two-sample t tests evaluating the two-way interaction indicated that the FCcl between the CaS and PFCa (t21 = 2.58, P = 0.02), INS (t21 = 4.17, P = 0.0005) and MCC (t21 = 4.07, P = 0.0006) increased during the eyes open condition (Fig. 3B); interestingly, these three frontal regions all belong to the salience network (shown in purple). Note that only salience regions INS and MCC showed increased FC when using a stringent Bonferroni correction for possibly nine two-sample t tests in this FC analysis. Thus, FCcl between early visual and salience regions increases when subjects open their eyes. On the other hand, we found no difference in or persistent FCcl with the PFCdl (P > 0.1) and PFCvm (P > 0.1), which belong to the central executive and default mode networks, respectively. Taken together, our FCcl analyses reveal persisting visual pathways across conditions and additional salience pathways once subjects open their eyes.

MCM Reveals EC Between Regions.

In a final step, we investigated EC among these visual and salience pathways. Here MCM identifies whether an area receives signaling input, or afferent EC, by integrating voxels’ FC and energy metabolism. For a given pathway, e.g., between the CaS and StC, we first used a cluster time course of the CaS to calculate FCvox for each voxel in the StC. The spatial correlation analysis between FCvox and FDGvox then revealed whether the StC receives afferent, or bottom-up, signaling (Fig. 4, red bars). A similar analysis in the CaS tested for the inverse direction, i.e., top-down signaling (Fig. 4, blue bars). Importantly, we performed this integrated analysis on maximally unsmoothed data by omitting spatial normalization or smoothing of both the fMRI and FDG data. This allowed us to calculate the signaling hierarchy in each subject individually. Illustrations of alignment between coregistered EPI and PET images are provided in Fig. S1.
Fig. 4.
EC between regions using MCM. (A) Stable bidirectional (bottom-up/top-down) signaling between the CaS and all other visual regions (StC, ExC, ITG, and SPL) and top-down modulation from the SPL (P < 0.0005) independent of condition. (B) All frontal regions of the salience network (purple) exert top-down modulation of the CaS only during the eyes open condition (PFCa: P < 0.0005; INS: P < 0.005; MCC: P < 0.05). Boxplots indicate median, 25th–75th percentiles (box), and extreme data points (whiskers).
Fig. S1.
Illustration of alignment between individual EPI and FDG-PET images in four subjects. Shown is the alignment between FDG-PET (gray) and EPI (red overlay) images in individual subject space after rigid-body coregistration. Note the full overlap between both imaging modalities in regions analyzed in this study, particularly occipital (axial slice), insular (coronal slice), and cingulate (saggital) cortices; however, typical signal dropouts owing to field inhomogeneity can be found in the medial inferior frontal cortex of the EPI images. Voxel coordinates: 50, 63, and 32.
First, we examined the signaling hierarchy along the visual stream pathways. We calculated MCM for all visual ROIs on both eyes closed and eyes open datasets and submitted these measures to ANOVA with region (StC, ExC, ITG, and SPL) and direction (bottom-up/top-down) as factors. We found main effects of region (F1,21 = 3.92, P = 0.01) and direction (F1,21 = 4.62, P = 0.033), along with a two-way interaction between region and direction (F3,19 = 5.21, P = 0.002), indicating that signaling direction differs for only certain pathways. Post hoc t tests evaluating this two-way interaction revealed unidirectional signaling for the SPL pathway (t21 = 4.1, P = 0.0002; P = 0.001 Bonferroni-corrected for seven two-sample t tests in MCM analysis), but not for any of the other connections (P > 0.1). As shown in Fig. 4A, the SPL pathway is driven by top-down signaling (blue bar), whereas communication in all other visual stream pathways is characterized by bidirectional EC.
Similarly, we investigated the EC along the salience pathways during the eyes open condition. In individual subject spaces, we calculated MCM and submitted these measures to ANOVA with region (PFCa, INS, and MCC) and direction (bottom-up/top-down) as factors. We found a main effect of direction (F1,21 = 41.53, P = 4.1 × 10−9), but no main effect of region (P > 0.1) or any interaction effects (P > 0.2). Thus, there is consistent unidirectional communication along the salience pathways. Post hoc t tests revealed that all salience regions exert a top-down influence on the CaS (PFCa: t21 = 4.78, P = 0.0001; INS: t21 = 3.26, P = 0.004; MCC: t21 = 3.04, P = 0.007; all P < 0.05 Bonferroni-corrected) (Fig. 4B). Taken together, these MCM values reveal a distinct signaling hierarchy along visual stream and salience pathways. Independent of condition, bidirectional communication occurs between early and higher visual regions, but top-down signaling occurs from a parietal region. Once subjects open their eyes, additional top-down signaling from a salience network occurs.
Finally, in two control analyses, we evaluated the effects of voxel and ROI size on the results of MCM. To diminish the effect of spatial smoothing, which might artificially increase spatial correlations between voxel values, we omitted spatial normalization and smoothing during preprocessing. Nonetheless, spatial sampling and the extent of ROI size in the target area might influence the strength of MCM. Fig. S2 presents MCM data for varying ROI and voxel sizes (Fig. S2A), both of which show very consistent results. We found identical directionality with larger voxels (Fig. S2B), as well as with smaller target ROIs down to one-quarter of the original ROI size (Fig. S2C).
Fig. S2.
Influence of voxel and ROI size on MCM results. (A) Varying ROI size of the target region during MCM calculation using an increasing z-value threshold (z > 3–7; light-blue to dark-blue bars) from initial ICA maps (2 × 2 × 2 mm). For comparison, we also calculated MCM on a larger voxel size (3 × 3 × 3 mm) at z > 3 (green bars). (B) Results of MCM calculation on 3 × 3 × 3 voxel size between the CaS and all other regions, corresponding to Fig. 4. Color coding: light green, bottom-up EC; dark green, top-down EC. Note that the directionality of all connections matches the results presented in Fig. 4. (C) Results of MCM calculation with varying target ROI sizes (see A) between the VISa and all other regions, corresponding to Fig. 4. Color coding: light-green to dark-green bars, z > 3–7. Note that the results correspond to MCM calculations presented in Fig. 4 up to z > 6, corresponding to approximately one-quarter of the original ROI size. Increasing variability of MCM occurs only at z > 7.

Discussion

Here we have established a new framework for identifying brain state EC from simultaneously acquired fMRI and FDG-PET data. We first characterized functional pathways among visual, parietal, and frontal regions by calculating FC from fMRI data. We then identified the direction of communication among these pathways by integrating neuroenergetics data obtained from FDG-PET. We found permanent and bidirectional communication between early and higher visual regions of the occipital cortex and top-down signaling from parietal regions. Only during the eyes open condition did additional top-down signaling from a frontal salience network emerge. We have identified dynamically changing signaling pathways during simple brain states in humans, offering a method to further investigate directed communication in healthy and diseased brains.
In human brain imaging, the visual networks are among the most consistent networks in FC analyses (30, 31). Our FC analysis revealed functional pathways between the early visual CaS ROI and all ROIs along the ventral and dorsal visual streams, including inferior temporal and parietal regions. These functional pathways were stable across conditions, suggesting interaction between these regions beyond functional processing of visual input, possibly reflecting underlying anatomic connections. In the monkey brain, retrograde tracer studies have shown feedforward and feedback pathways among striate, extrastriate, and parietal regions (4, 26, 27). Structural imaging with diffusion tensor imaging in humans also has identified a strong overlap between FC and anatomic pathways (32). Our FC analysis supports the notion of functionally distinct, yet highly connected, areas along the visual processing stream.
The frontal cortex includes a core set of networks—default mode, central executive, and salience networks—that have been assigned to higher cognitive functions and are prominently affected in neuropsychiatric disorders (13, 33). We evaluated functional pathways between the early visual cortex and all three networks, but identified FC only with regions of the salience network and then only during the eyes open condition. This finding is interesting, because it is consistent with the behavioral relevance ascribed to each network. Current knowledge assigns self-related memory processing to the default mode network (34), event-related and error-based processing to the central executive network (35), and set maintenance or salience processing of internal and external states to the salience network (36). For instance, FC in the salience network correlates with prescan anxiety ratings in humans (37). The switch from eyes closed to eyes open might induce a change in brain state reflected by salience FC and prepare the organism for event- or self-related processing in the central executive and default mode networks once visual input is present.
We next integrated independent voxel measures of FC and FDG to reveal EC along the visual stream and salience pathways. For the visual pathways, we identified bidirectional communication among regions in the occipital cortex and top-down signaling from fronto-parietal regions. Whereas early anatomic studies suggested a bottom-up hierarchy along the visual streams, recent work has stressed the importance of feedback signaling onto early visual areas. The StC receives considerably more feedback and lateral input than feedforward thalamic afferents (27, 38). This is achieved by a dense network of cortico-cortical connections (1), and feedback signaling occurs not only from extrastriate cortices, but also from parietal cortices (39, 40). Direct feedback connections onto V1 have even been shown for the frontal eye fields (4) that we identified as the most frontal region within the dorsal visual stream (SPL). Taken together, our results largely reflect anatomic data from nonhuman primate studies of strong reciprocal connections along the extended visual processing pathway. Our data only indicate the directionality of communication between major hubs of functional networks, however, and we cannot rule out the possibility of intermediate brain regions below the spatial resolution of fMRI that might mediate the influence on early visual areas.
Finally, we also found a top-down influence from all prefrontal regions of the salience network. Again, this communication might occur along downstream regions that we have missed with current human imaging methods. Nonetheless, previous work revealed top-down modulation of visual cortex activity by frontal regions. In humans, inhibitory transcranial magnetic stimulation of the inferior frontal cortex has been shown to decrease the performance of early visual processing (41). Moreover, recent studies in monkeys using novel tracer techniques have identified previously unknown long-range connections between the prefrontal and early visual cortex (42). Our finding of top-down signaling also might reflect a more general notion about the underestimated influence of cortical feedback, particularly onto the early visual cortex (3, 24). A recent study identified twice as many feedback connections as feedforward connections for cortico-cortical pathways, and also showed that feedback connections tend to be more long-range, whereas feedforward pathways are comparatively short-range (26).
Certain methodological factors must be taken into account when interpreting our MCM data. First, our experimental setup has several limitations. The instruction to keep eyes closed or open is easy for subjects to follow, particularly with respect to scanning; however, resting brain states obviously differ across subjects and entail neuronal processes across many cognitive domains that cannot be controlled for. We chose the resting state on the basis of the extensive literature in this area, and because of the self-evident difficulties in designing tasks that engage sustained and consistent processing over extended time periods (necessary to acquire PET data). A positive aspect, however, is the fact that a basic set of networks emerged consistently in the resting state across subjects (43). Our study design also had to accommodate the fact that temporal resolution of PET is limited to one saturated image of accumulated tracer after 30 min of scanning. Although recent fMRI studies suggest temporal variability of FC (i.e., dynamic FC) within and between networks (44), the MCM approach is based on FC expressed (possibly intermittently) consistently over extended periods. This property ensures that the time scale of coupling is commensurate with the time scale of metabolic (PET) measurements. In other words, MCM pertains to the EC mediating consistent FC that may or may not be fluctuating over time. It might be interesting to revisit this issue by using the average of dynamic FC, as opposed to FC per se. A final limitation of the present study is that interpretation of MCM results might be confounded by varying levels of arousal during the eyes closed and eyes open conditions. In other words, frontal top-down signaling during eyes open might indicate anticipatory signaling from salience regions, or frontal uncoupling during eyes closed may occur simply because subjects have lower levels of arousal.
Second, MCM relies on the assumption that voxel values of FC and FDG spatially correlate with each other in connected regions. Underlying vascular heterogeneity might induce false-positive or -negative spatial correlations, however. In the resting state fMRI literature, the influence of cerebral blood flow (CBF) and vessel size on BOLD-FC is increasingly discussed (45), and recent studies suggest strong BOLD-CBF coupling (46, 47). There is also evidence that these CBF fluctuations indicate reactivity to neuronal activity rather than to heterogeneous vascularity, however. Tak et al. (46) also acquired data on large-vessel volume fraction and found that BOLD-FC was inversely related to macrovascular volume fraction. The stronger relationship with small- to medium-sized vessels (closer to neuronal activity) was particularly prominent for major FC hubs. In summary, spatial variation in neurovascular mechanisms will lead to spatial inhomogeneities in measured BOLD signals that will clearly affect spatial correlation over voxels. To a certain extent, our use of FC (as opposed to BOLD signal per se) mitigates this problem, in the sense that the correlation between the BOLD signals in two regions does not change with their amplitude (e.g., neurovascular sensitivity). This assumes that the signal-to-noise ratio of BOLD time series is spatially invariant, however. If it is not, then MCM may be prone to false-negative results.
As noted by our reviewers, the potentially confounding effects of neurovascular heterogeneity cannot be invoked to explain the condition-specific changes in EC that we observed. This means that the assessment of changes in EC may be more robust to vascular heterogeneity. On the other hand, the combination of two modalities with similar voxel resolution also has distinct advantages with respect to sensitivity. Integrating data acquired simultaneously and independently from the same subject allowed us to avoid spatial distortion of imaging data, which commonly occurs before statistical analysis (48). We omitted spatial normalization to a standard brain and also spatial smoothing as we integrated the two datasets in individual subject space. Moreover, our control analyses showed that MCM was robust against variations in voxel and ROI size. Thus, this approach might prove sufficiently sensitive for even single-subject analyses.
Third, an ongoing discussion in the PET literature surrounds the issue of how to interpret FDG signals. FDG uptake reflects energy consumption and to a large part is related to neuronal signaling (49); however, a critical question is whether neurons themselves or astrocytes consume most of the energy provided by glucose. Numerous studies have provided evidence for both cell types, but a very recent study “identified neurons as the principle locus of glucose uptake as visualized by functional brain imaging” (50). According to that study, FDG-PET is an ideal surrogate for energy consumption in neurons and specifically at synapses. Furthermore, this idea supports the general assumption about the neuronal basis of the BOLD signal. Earlier studies of simultaneous electrophysiology and fMRI have concluded that BOLD signal fluctuations reflect synaptic activity rather than spiking activity (51, 52). Using complementary methods like PET, we can now confirm this assumption directly in humans.
Although we and others found a partial spatial overlap between FDG and fMRI measures in previous studies (10, 53), MCM is a novel approach using energy consumption as a proxy for directional signaling. The question remains as to which aspects of EC (and endogenous fluctuations) best predict metabolic activity as measured by FDG. There are a number of competing hypotheses here. For example, a decrease in regional self-inhibition may correspond to increased excitability and an increase in metabolism (and spatial correlation with FC). Conversely, metabolic increases may be mediated directly by afferent EC from source regions. Finally, the best predictor of metabolic activity may be the amplitude of neuronal fluctuations mediated by changes in EC. In a future study, we will apply stochastic and spectral dynamic causal modeling (DCM) to the data described above (17). The parameters (intrinsic and extrinsic connectivity) and the amplitude of inferred neuronal fluctuations will be used to predict regional metabolism (and its spatial correlation with FC) to identify the key determinants that underlie MCM. The use of spectral DCM here may be interesting, because this approach explicitly parameterizes the spectral profile of endogenous neuronal fluctuations, aa well as their implicit time constants. We hope to test the hypothesis that the determinants of MCM rely not just on the amplitude of neuronal fluctuations (mediated by EC), but also on the characteristic time constants as modeled with power law scaling.

Materials and Methods

Our analysis was performed on 24 healthy subjects (16 males, 8 females; average age, 54.7 ± 9.9 y; all right-handed) participating in a simultaneous FDG-PET/fMRI brain imaging study. All participants gave informed consent to procedures approved by the Ethics Review Board of the Klinikum Rechts der Isar, Technische Universität München. Scanning was performed on an integrated Siemens Biograph mMR scanner capable of simultaneously acquiring PET and MRI (3 T) data using the vendor-supplied 12-channel phase-array head coil.
We simultaneously measured FDG-PET activity and BOLD-fMRI signals during resting conditions of either eyes closed (CLOSED) or eyes open (OPEN) during the initial 10 min immediately after bolus injection of FDG tracer. At 30 min postinjection, we recorded the saturated list mode PET dataset for 10 min. All scanning was performed in a dimmed environment obtained by switching off all lights, including those in the scanner bore. Subjects were instructed to keep their eyes closed/open, to relax, to not think of anything in particular, and to not fall asleep. We assessed each subject’s level of arousal and anxiety using the State Trait Anxiety Inventory before the start of the scanning session (CLOSED/OPEN: 35.5 ± 6.9/30.8 ± 5.6; P = 0.18), and measured blood pressure and venous glucose level (<140 mg/dL in all subjects).
We applied two different preprocessing pipelines for the initial FC analysis of fMRI data, and for the final integrated MCM analysis of fMRI and PET data in single-subject space.

FC Analysis.

We first identified ROIs according to a standard set of networks that Yeo et al. (30) recently established in a 1,000-subject dataset. We followed our previous processing pipeline for ICA of fMRI data (54). We applied slice-time correction, realignment, normalization to a standard template, and spatial smoothing using a Gaussian kernel with an FWHM of 6 mm. We then subjected the concatenated data of all subjects to a group ICA using GIFT (Medical Imaging Analysis Lab, The Mind Research Network) and identified 70 spatially independent components (ICs) with corresponding signal time series (31). Only the most stable ICs with a stability index > 0.9 were identified by iteratively running 30 ICAs using the ICASSO procedure (55). We automatically selected ICs as relevant ROIs by running spatial regression against the spatial maps of the seven-network decomposition from Yeo et al. (30). Finally, we identified functional pathways of FC between an early visual area (CaS) and all other visual stream ROIs (StC, ExC, and ITG, SPL) and prefrontal association ROIs (PFCa, INS, PFCdl, MCC, and PFCvm). For each subject separately, we calculated the temporal Pearson’s correlation between the cluster time series of the CaS and that of all other ROIs.

Multimodal MCM Analysis in Single-Subject Space.

In this analysis, we integrated voxel-wise measures of FC and FDG activity in single-subject spaces to reveal EC among previously established FC pathways. We used the temporally (slice time corrected) and spatially corrected EPI volumes of each subject, but omitted normalization and smoothing procedures to achieve the least distorted volumes. We then coregistered the mean FDG-PET volume to the mean EPI volume, scaled PET data to normalized FDG activity by whole-brain FDG uptake values, and resampled both datasets to common voxel dimensions of 2 × 2 x 2 mm3. We then submitted the fMRI and FDG-PET data to a two-step procedure to calculate MCM (Fig. 1).
Based on the energy model of neuronal signaling, a positive spatial correlation between the voxel profiles of FC (FCvox) and FDG activity (FDGvox) identifies a region as the input area of a pathway. FDGvox is derived from FDG activity maps, and FCvox is calculated as the temporal correlation between each voxel time series within the (target) ROI and the cluster time series of the connected (seed) region, i.e., FCvox in the StC with cluster time series of the CaS. The spatial extent of each ROI was defined by the initial ICA maps (z > 3) back-projected into individual space.
More detailed information is provided in SI Materials and Methods.

SI Materials and Methods

Subjects.

Twenty-four healthy subjects (16 males and 8 females; average age, 54.7 ± 9.9 y, all right-handed) participated in this simultaneous FDG-PET/fMRI brain imaging study, two of which had to be excluded during scanning (see Study Design). Female (>40 y) and male (>35 y) participants were recruited from the greater Munich community, received an expense allowance, and gave informed consent to procedures approved by the Ethics Review Board of the Klinikum Rechts der Isar, Technische Universität München. Exclusion criteria included a history of psychiatric or neurologic conditions, psychoactive medications, pregnancy, or renal failure (critical for PET imaging), as well as common MRI-related exclusion criteria.

Study Design.

We simultaneously measured FDG-PET activity and BOLD-fMRI signals during resting conditions of either eyes closed (CLOSED) or eyes open (OPEN). The amount of radiation exposure per single-shot FDG tracer precluded repeated FDG application in the same subject; thus, we assigned subjects to either the CLOSED group or the OPEN group matched across age (CLOSED/OPEN: 56.9 ± 9.9 y/52.2 ± 10.2 y; P = 0.31), sex (CLOSED/OPEN: 3/4 female), and handedness (all right-handed).
We designed a scanning protocol to achieve the best contrast between CLOSED and OPEN with both imaging techniques. In resting-state fMRI studies, scanning is usually limited to approximately 10 min, because subjects tend to fall asleep during prolonged rest, resulting in altered brain network configuration (56). In FDG-PET imaging, a steady-state condition is usually maintained for at least 30 min, during which the tracer distributes equally across the entire brain (57, 58). After 30 min, a saturated image is acquired that reflects the accumulated tracer distributed across the entire brain. However, the most sensitive period in FDG imaging is the accumulation phase in the initial minutes after tracer injection, when large amounts of FDG tracer accumulate in activated brain regions depending on current glucose consumption (59). Using the modified FDG-PET design of Vlassenko et al. (59), we achieved the greatest contrast between the CLOSED and OPEN conditions during the first 10 min of simultaneous fMRI/FDG-PET imaging. We then performed the remaining 20 min of PET imaging during identical conditions of eyes closed for all subjects.
We began the scanning with an initial period of 4 min during which the subject was instructed to keep his or her eyes closed. This was followed by the functional CLOSED/OPEN run lasting 10 min. Immediately before that run, we instructed the subject to keep his or her eyes closed/open, to relax, to not think of anything particular, and not to fall asleep. Together with bolus injection of FDG tracer, we started simultaneous fMRI data acquisition. Immediately after the CLOSED/OPEN run, we interrogated the subject as to whether he or she had stayed awake. We had to exclude two subjects who expressed doubts about staying awake, reducing the image analysis to 22 subjects. Then, with the subject’s eyes closed again, we performed structural imaging and diagnostic screening MRI sequences. At 30 min postinjection, we recorded the saturated list-mode PET dataset for 10 min, reflecting FDG tracer distribution after homogeneous and complete distribution in the brain.
All scanning was performed in a dimmed environment obtained by switching off all lights, including those in the scanner bore. Light was allowed to pass through the window of the console room to allow for accurate FDG injections during MRI scanning. Subjects had to fast during the 4-h period preceding imaging. Immediately before scanning, we performed a neuropsychiatric examination to exclude current neurologic or psychiatric diseases, and explained the study protocol to the subject. We assessed the subject’s level of arousal and anxiety using the State Trait Anxiety Inventory in each subject before the start of the scanning session (CLOSED/OPEN: 35.5 ± 6.9/30.8 ± 5.6; P = 0.18), and measured blood pressure and venous glucose level (<140 mg/dL in all subjects). Subjects were prepared with a venous catheter to receive the tracer injection during scanning while lying in the scanner bore.

Data Acquisition.

Scanning was performed on an integrated Siemens Biograph mMR scanner capable of simultaneously acquiring PET and MRI (3 T) data using the vendor-supplied 12-channel phase-array head coil. PET volume, T2*-weighted echo-planar-imaging (EPI) MRI, and magnetization-prepared rapid acquisition gradient echo (MPRAGE) T1-weighted anatomic volume data were acquired. Parameters for PET were as follows: list-mode acquisition; 128 slices (gap, 0.5 mm) covering the whole brain; field of view (FoV), 450 mm; matrix size, 192 × 192; and voxel size, 3.7 × 2.3 × 2.7 mm3. Scanning parameters for EPI included repetition time (TR)/echo time (TE)/flip angle, 2,000 ms/30 ms/90°; 35 slices (gap 0.6 mm) aligned to the anterior-posterior commissure covering the whole brain; FoV, 192 mm; matrix size, 64 × 64; and voxel size, 3.0 × 3.0 × 3.0 mm3. Each measurement consisted of 300 acquisitions in interleaved mode, with a total scan time of 10 min 8 s. Scanning parameters for MPRAGE included TR/TE/flip angle, 2,300 ms/2.98 ms/9°; 160 slices (gap, 0.5 mm) covering the whole brain; FoV, 256 mm; matrix size, 256 × 256; and voxel size, 1.0 × 1.0 × 1.0 mm3. The total scan time was 5 min, 3 s. During the initial 4 min, we acquired arterial spin labeling data, which were not analyzed for the present study.

Data Processing.

DICOM images were first converted to 3D-NIfTI volumes and then preprocessed using SPM8 (Wellcome Trust Center for Neuroimaging). We applied two different preprocessing pipelines for the initial FC analysis of fMRI data and the final integrated MCM analysis of fMRI and PET data in single-subject space.

Identification of brain regions and FC analysis between regions.

We first identified ROIs according to a standard set of networks that Yeo et al. (30) recently established in a 1,000-subject dataset. For this analysis, we followed our previous processing pipeline for ICA of fMRI data (54). The first three volumes of each subject’s EPI time series were deleted, and slice time correction was applied to the remaining EPI volumes to account for differences in slice acquisition time. EPI volumes were then realigned to the subject’s mean EPI volume using 6° rigid-body transformation to correct for head motion during fMRI scanning. To allow analysis across individuals, each subject’s brain data were then normalized to a standard template provided by the Montréal Neurological Institute using the SPM8 unified segmentation routine. Finally, EPI volumes were spatially smoothed using a Gaussian kernel with an FWHM value of 6 mm.
We then subjected the concatenated data of all subjects to a group ICA using GIFT (Medical Imaging Analysis Lab, The Mind Research Network) to identify common clusters of BOLD coherence. We performed a high-order data decomposition and identified 70 ICs with corresponding signal time series (31). Only the most stable ICs, with a stability index > 0.9, were identified by iteratively running 30 ICAs using the ICASSO procedure (55). The ICA comprises three consecutive steps (60): data reduction with principal component analysis, ICA using the Infomax algorithm, and back-reconstruction of the group ICs into the single-subject space using GICA3 (61). The same set of ICs exists for each subject, each with an individual spatial extent and an individual BOLD signal time series. We automatically selected ICs as relevant ROIs by running spatial regression against the spatial maps of the seven-network decomposition from Yeo et al. (30). Only the most coherent ICs, with consecutive clusters of >1,000 voxels at a z-threshold > 3, were selected. For each relevant IC, we calculated a whole-brain group statistical parametric map (SPM) of voxel-wise one-sample t tests with a familywise error-corrected P value < 0.05 at the voxel level using Gaussian random-fields theory. For illustration, we projected each ROI’s SPM onto an inflated single-subject high-resolution brain image using CARET (Van Essen Laboratory, Washington University).
Finally, we identified functional pathways of FC between an early visual area (CaS) and all other visual stream ROIs (StC, ExC, ITG, and SPL) and prefrontal association ROIs (PFCa, INS, PFCdl, MCC, and PFCvm). We chose the CaS as the early visual area because it fully covers the region with the greatest metabolic increase when subjects open their eyes (10). For each subject separately, we calculated the temporal Pearson correlation between the cluster time series of CaS and that of all other ROIs and subjected these measures to two distinct ANOVAs (visual/frontal pathways) with network and condition (eyes closed/open) as factors.

Multimodal MCM analysis in single-subject space.

In this analysis, we integrated voxel-wise measures of FC and FDG activity in single-subject space to reveal EC among previously established FC pathways. We used the temporally (slice time corrected) and spatially corrected EPI volumes of each subject, but omitted normalization and smoothing procedures to achieve least distorted volumes. We then coregistered the mean FDG-PET volume to the mean EPI volume, which produced a subject-specific coregistered multimodal dataset. Finally, we scaled PET volume data to normalized FDG activity by whole-brain FDG uptake values, and then resampled both datasets to common voxel dimensions of 2 × 2 x 2 mm3. This oversampling should have reduced the partial volume effects on voxel values during coregistration of both modalities. We then submitted the fMRI and FDG-PET data to a two-step procedure to calculate MCM (Fig. 1).
Based on the energy model of neuronal signaling, a positive spatial correlation between the voxel profiles of FC (FCvox) and FDG activity (FDGvox) identifies a region as the input area of a pathway. FDGvox is derived from FDG activity maps, and FCvox is calculated as the temporal correlation between each voxel time series within the (target) ROI and the cluster time series of the connected (seed) region, i.e., FCvox in the StC with a cluster time series of the CaS. The spatial extent of each ROI was defined by the initial ICA maps (z > 3) back-projected into individual space. Finally, we submitted MCM measures to two distinct ANOVAs (visual/frontal pathways) with network and direction (bottom-up/top-down) as factors.

Acknowledgments

We thank Edward Bullmore for discussions during the process of data analysis. We also thank our colleagues at the Technische Universitaet Muenchen, particularly Claus Zimmer, Markus Schwaiger, Igor Yakushev, Anne Winter, Sylvia Schachoff, Brigitte Dzewas, Michael Herz, and Mathias Lukas. This work was supported by a university grant from the Kommission fuer klinische Forschung (Grant 8762754, to V.R. and A.D.) at the Klinikum Rechts der Isar der Technischen Universitaet Muenchen.

Supporting Information

Supporting Information (PDF)
Supporting Information

References

1
DJ Felleman, DC Van Essen, Distributed hierarchical processing in the primate cerebral cortex. Cereb Cortex 1, 1–47 (1991).
2
PS Goldman-Rakic, Topography of cognition: Parallel distributed networks in primate association cortex. Annu Rev Neurosci 11, 137–156 (1988).
3
CD Gilbert, W Li, Top-down influences on visual processing. Nat Rev Neurosci 14, 350–363 (2013).
4
S Clavagnier, A Falchier, H Kennedy, Long-distance feedback projections to area V1: Implications for multisensory integration, spatial awareness, and visual consciousness. Cogn Affect Behav Neurosci 4, 117–126 (2004).
5
SL Bressler, V Menon, Large-scale brain networks in cognition: Emerging methods and principles. Trends Cogn Sci 14, 277–290 (2010).
6
MD Fox, ME Raichle, Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci 8, 700–711 (2007).
7
CM Lewis, A Baldassarre, G Committeri, GL Romani, M Corbetta, Learning sculpts the spontaneous activity of the resting human brain. Proc Natl Acad Sci USA 106, 17558–17563 (2009).
8
NB Albert, EM Robertson, RC Miall, The resting human brain and motor learning. Curr Biol 19, 1023–1027 (2009).
9
MTR van Kesteren, G Fernández, DG Norris, EJ Hermans, Persistent schema-dependent hippocampal–neocortical connectivity during memory encoding and postencoding rest in humans. Proc Natl Acad Sci USA 107, 7550–7555 (2010).
10
V Riedl, et al., Repeated pain induces adaptations of intrinsic brain activity to reflect past and predict future pain. NeuroImage 57, 206–213 (2011).
11
M Hampson, NR Driesen, P Skudlarski, JC Gore, RT Constable, Brain connectivity related to working memory performance. J Neurosci 26, 13338–13343 (2006).
12
T Harmelech, S Preminger, E Wertman, R Malach, The day-after effect: Long-term, Hebbian-like restructuring of resting-state fMRI patterns induced by a single epoch of cortical activation. J Neurosci 33, 9488–9497 (2013).
13
V Menon, Large-scale brain networks and psychopathology: A unifying triple network model. Trends Cogn Sci 15, 483–506 (2011).
14
M Greicius, Resting-state functional connectivity in neuropsychiatric disorders. Curr Opin Neurol 21, 424–430 (2008).
15
KJ Friston, Functional and effective connectivity: A review. Brain Connect 1, 13–36 (2011).
16
AJ Bastos-Leite, et al., Dysconnectivity within the default mode in first-episode schizophrenia: A stochastic dynamic causal modeling study with functional magnetic resonance imaging. Schizophr Bull 41, 144–153 (2015).
17
KJ Friston, J Kahan, B Biswal, A Razi, A DCM for resting-state fMRI. Neuroimage 94, 396–407 (2014).
18
F Hyder, DL Rothman, MR Bennett, Cortical energy demands of signaling and nonsignaling components in brain are conserved across mammalian species and activity levels. Proc Natl Acad Sci USA 110, 3549–3554 (2013).
19
D Attwell, SB Laughlin, An energy budget for signaling in the grey matter of the brain. J Cereb Blood Flow Metab 21, 1133–1145 (2001).
20
D Attwell, C Iadecola, The neural basis of functional brain imaging signals. Trends Neurosci 25, 621–625 (2002).
21
H Alle, A Roth, JRP Geiger, Energy-efficient action potentials in hippocampal mossy fibers. Science 325, 1405–1408 (2009).
22
JJ Harris, R Jolivet, D Attwell, Synaptic energy use and supply. Neuron 75, 762–777 (2012).
23
T Jao, et al., Volitional eyes opening perturbs brain dynamics and functional connectivity regardless of light input. Neuroimage 69, 21–34 (2013).
24
JC Anderson, H Kennedy, KAC Martin, Pathways of attention: Synaptic relationships of frontal eye field to V4, lateral intraparietal cortex, and area 46 in macaque monkey. J Neurosci 31, 10872–10881 (2011).
25
M Mishkin, LG Ungerleider, KA Macko, Object vision and spatial vision: Two cortical pathways. Trends Neurosci 6, 414–417 (1983).
26
NT Markov, et al., Anatomy of hierarchy: Feedforward and feedback pathways in macaque visual cortex. J Comp Neurol 522, 225–259 (2014).
27
L Muckli, LS Petro, Network interactions: Non-geniculate input to V1. Curr Opin Neurobiol 23, 195–201 (2013).
28
L Pasquini, et al., Link between hippocampus’ raised local and eased global intrinsic connectivity in AD. Alzheimers Dement 11, 475–484 (2015).
29
S Lui, et al., Resting-state functional connectivity in treatment-resistant depression. Am J Psychiatry 168, 642–648 (2011).
30
BTT Yeo, et al., The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 106, 1125–1165 (2011).
31
EA Allen, et al., A baseline for the multivariate comparison of resting-state networks. Front Syst Neurosci 5, 2 (2011).
32
CJ Honey, R Kötter, M Breakspear, O Sporns, Network structure of cerebral cortex shapes functional connectivity on multiple time scales. Proc Natl Acad Sci USA 104, 10240–10245 (2007).
33
AC Chen, et al., Causal interactions between fronto-parietal central executive and default-mode networks in humans. Proc Natl Acad Sci USA 110, 19944–19949 (2013).
34
JR Andrews-Hanna, JS Reidler, J Sepulcre, R Poulin, RL Buckner, Functional-anatomic fractionation of the brain’s default network. Neuron 65, 550–562 (2010).
35
NUF Dosenbach, DA Fair, AL Cohen, BL Schlaggar, SE Petersen, A dual-networks architecture of top-down control. Trends Cogn Sci 12, 99–105 (2008).
36
V Menon, LQ Uddin, Saliency, switching, attention and control: A network model of insula function. Brain Struct Funct 214, 655–667 (2010).
37
WW Seeley, et al., Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 27, 2349–2356 (2007).
38
JM Budd, Extrastriate feedback to primary visual cortex in primates: A quantitative analysis of connectivity. Proc Biol Sci 265, 1037–1044 (1998).
39
P Barone, A Batardiere, K Knoblauch, H Kennedy, Laminar distribution of neurons in extrastriate areas projecting to visual areas V1 and V4 correlates with the hierarchical rank and indicates the operation of a distance rule. J Neurosci 20, 3263–3281 (2000).
40
JW Lewis, DC Van Essen, Corticocortical connections of visual, sensorimotor, and multimodal processing areas in the parietal lobe of the macaque monkey. J Comp Neurol 428, 112–137 (2000).
41
TP Zanto, MT Rubens, A Thangavel, A Gazzaley, Causal role of the prefrontal cortex in top-down modulation of visual processing and working memory. Nat Neurosci 14, 656–661 (2011).
42
NT Markov, et al., The role of long-range connections on the specificity of the macaque interareal cortical network. Proc Natl Acad Sci USA 110, 5187–5192 (2013).
43
BB Biswal, et al., Toward discovery science of human brain function. Proc Natl Acad Sci USA 107, 4734–4739 (2010).
44
EA Allen, et al., Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 24, 663–676 (2014).
45
TT Liu, Neurovascular factors in resting-state functional MRI. Neuroimage 80, 339–348 (2013).
46
S Tak, JR Polimeni, DJJ Wang, L Yan, JJ Chen, Associations of resting-state fMRI functional connectivity with flow-BOLD coupling and regional vasculature. Brain Connect 5, 137–146 (2015).
47
X Liang, Q Zou, Y He, Y Yang, Coupling of functional connectivity and regional cerebral blood flow reveals a physiological basis for network hubs of the human brain. Proc Natl Acad Sci USA 110, 1929–1934 (2013).
48
ME Shaw, et al., Evaluating subject specific preprocessing choices in multisubject fMRI data sets using data-driven performance metrics. Neuroimage 19, 988–1001 (2003).
49
ME Raichle, A brief history of human brain mapping. Trends Neurosci 32, 118–126 (2009).
50
I Lundgaard, et al., Direct neuronal glucose uptake heralds activity-dependent increases in cerebral metabolism. Nat Commun 6, 6807 (2015).
51
A Viswanathan, RD Freeman, Neurometabolic coupling in cerebral cortex reflects synaptic more than spiking activity. Nat Neurosci 10, 1308–1312 (2007).
52
NK Logothetis, J Pauls, M Augath, T Trinath, A Oeltermann, Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150–157 (2001).
53
D Tomasi, G-J Wang, ND Volkow, Energetic cost of brain functional connectivity. Proc Natl Acad Sci USA 110, 13642–13647 (2013).
54
A Dovern, et al., Intrinsic network connectivity reflects consistency of synesthetic experiences. J Neurosci 32, 7614–7621 (2012).
55
J Himberg, A Hyvärinen, F Esposito, Validating the independent components of neuroimaging time series via clustering and visualization. Neuroimage 22, 1214–1222 (2004).
56
E Tagliazucchi, et al., Automatic sleep staging using fMRI functional connectivity data. Neuroimage 63, 63–72 (2012).
57
PT Fox, ME Raichle, MA Mintun, C Dence, Nonoxidative glucose consumption during focal physiologic neural activity. Science 241, 462–464 (1988).
58
A Drzezga, et al., Neuronal dysfunction and disconnection of cortical hubs in non-demented subjects with elevated amyloid burden. Brain 134, 1635–1646 (2011).
59
AG Vlassenko, MM Rundle, MA Mintun, Human brain glucose metabolism may evolve during activation: findings from a modified FDG PET paradigm. Neuroimage 33, 1036–1041 (2006).
60
VD Calhoun, T Adali, GD Pearlson, JJ Pekar, A method for making group inferences from functional MRI data using independent component analysis. Hum Brain Mapp 14, 140–151 (2001).
61
EB Erhardt, et al., Comparison of multi-subject ICA methods for analysis of fMRI data. Hum Brain Mapp 32, 2075–2095 (2011).

Information & Authors

Information

Published in

Go to Proceedings of the National Academy of Sciences
Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 113 | No. 2
January 12, 2016
PubMed: 26712010

Classifications

Submission history

Published online: December 28, 2015
Published in issue: January 12, 2016

Keywords

  1. simultaneous PET/fMRI
  2. energy metabolism
  3. directional signaling
  4. effective connectivity
  5. resting state

Acknowledgments

We thank Edward Bullmore for discussions during the process of data analysis. We also thank our colleagues at the Technische Universitaet Muenchen, particularly Claus Zimmer, Markus Schwaiger, Igor Yakushev, Anne Winter, Sylvia Schachoff, Brigitte Dzewas, Michael Herz, and Mathias Lukas. This work was supported by a university grant from the Kommission fuer klinische Forschung (Grant 8762754, to V.R. and A.D.) at the Klinikum Rechts der Isar der Technischen Universitaet Muenchen.

Notes

This article is a PNAS Direct Submission. P.A.B. is a guest editor invited by the Editorial Board.

Authors

Affiliations

Valentin Riedl1 [email protected]
Department of Neuroradiology, Klinikum rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Department of Nuclear Medicine, Klinikum rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Lukas Utz
Department of Neuroradiology, Klinikum rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Institute of Advanced Studies, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Gabriel Castrillón
Department of Neuroradiology, Klinikum rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Instituto de Alta Tecnologia Medica, 050026 Medellin, Colombia;
Timo Grimmer
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Department of Psychiatry and Psychotherapy, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Josef P. Rauschecker
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Institute of Advanced Studies, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Laboratory of Integrative Neuroscience and Cognition, Georgetown University, Washington, DC 20057;
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Department of Neurology, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
The Wellcome Trust Centre for Neuroimaging, University College London, London WC1N 3BG, United Kingdom;
Alexander Drzezga
Department of Nuclear Medicine, Uniklinik Köln, 50937 Köln, Germany
Christian Sorg
Department of Neuroradiology, Klinikum rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Neuroimaging Center, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;
Department of Psychiatry and Psychotherapy, Klinikum Rechts der Isar, Technischen Universitaet Muenchen, 81675 Muenchen, Germany;

Notes

1
To whom correspondence should be addressed. Email: [email protected].
Author contributions: V.R., A.D., and C.S. designed research; V.R., T.G., and A.D. performed research; V.R., L.U., and G.C. analyzed data; and V.R., J.P.R., M.P., K.J.F., A.D., and C.S. wrote the paper.

Competing Interests

The authors declare no conflict of interest.

Metrics & Citations

Metrics

Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.


Citation statements




Altmetrics

Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

Cited by

    Loading...

    View Options

    View options

    PDF format

    Download this article as a PDF file

    DOWNLOAD PDF

    Get Access

    Login options

    Check if you have access through your login credentials or your institution to get full access on this article.

    Personal login Institutional Login

    Recommend to a librarian

    Recommend PNAS to a Librarian

    Purchase options

    Purchase this article to get full access to it.

    Single Article Purchase

    Metabolic connectivity mapping reveals effective connectivity in the resting human brain
    Proceedings of the National Academy of Sciences
    • Vol. 113
    • No. 2
    • pp. 237-E257

    Media

    Figures

    Tables

    Other

    Share

    Share

    Share article link

    Share on social media