Main

It is well known that the brain's energy expenditure is considerably larger than what is to be expected from its weight alone. Similar observations have been made in the infant brain (1). Moreover, it has also become apparent that the additional brain energy required to respond to changes in the environment is surprisingly small (2). Taken together, these observations have lead researchers to conclude that a large amount of the brain's activity is spent on tasks that as of yet is unaccounted for (3). Several theories related to the functional role of the brain's intrinsic activity have been proposed. Speculatively, intrinsic brain activity might be related to the neuronal activity necessary to retain a sustained level of maintenance and updating of information flow in the brain. To this end, recent development in functional MRI (fMRI) research has provided the opportunity to gain insights into the brain's intrinsic functional architecture by recording spontaneous, low-frequency blood oxygenation level dependent (BOLD) fMRI signal changes that are present during rest, i.e. in absence of any overt behavior (see Ref. 4 for a recent review). Several studies have shown that spontaneous changes in resting-state fMRI activity is related to the changes in neuronal activity (5,6) and that signal synchronicity across widely separated brain areas exist within multiple so-called resting-state networks in the adult brain (7). Previous reports have shown that functional connectivity in the format of resting-state networks in the adult brain spans brain regions that are involved in sensory perception (8–10), language (11), and in the so-called default-mode network (12–14). The default-mode network is believed to be involved in different aspects of self-referential mental efforts (see Ref. 15; for a recent review on the mental capabilities of the newborn infant see Ref. 16). Recently, a number of studies have addressed the possibility to use resting-state functional connectivity as a biomarker for neurologic deficits in spatial neglect after stroke in the parietal cortex (17) as well as its potential role in developmental disorders such as attention deficit hyperactivity disorder and autism (see Ref. 18 for a recent review).

Because the resting-state fMRI does not require any overt feedback by the subjects, it is well suited to study functional connectivity in clinical populations as well as in children and infants. In a recently published study, we could show that spontaneous low-frequency fMRI BOLD signals are present in the sedated infant brain (19). In detail, using a model-free explorative image analysis method, we could show for the first time that the infant brain through spatio-temporal synchronization of spontaneous signal fluctuations exhibited resting-state networks that spanned long-range cortical areas that among others are known to be involved in sensory perception. However, our previous study was conducted at term-corrected age in infants born preterm (gestational age <27 wk) and lightly sedated (choral hydrate, 30 mg/kg). Hence, it is conceivable that preterm birth or prenatal factors might have influenced our results. Similarly, the effect of sedation on resting-state activity is currently under investigation and a modulation of functional connectivity from sedative agents cannot be ruled out (20–23). Therefore, the aim in this work was to investigate resting-state networks in the infant brain during natural sleep in babies born at term.

PATIENTS AND METHODS

Participants.

All 21 infants included in this study were born at term. Only healthy mothers with uncomplicated pregnancies, giving birth to healthy babies with normal Apgar scores, were invited to enter the study. Subjects were recruited from two hospitals in the Stockholm area. Parents with planned caesarean section (CS) were informed about the study. If parents were interested in participating in this study, a doctor further informed the family on the details regarding the MR examination. The reasons for CS were uncomplicated breech presentation, prior CS (acute or planned), history of abdominal surgery or anxiety/fear of delivery (humanitarian CS). No infants were delivered with acute CS. All infants except one were delivered with CS and informed consent was received from all parents. All MR scanning was conducted when the infants were in a stage of natural sleep. To facilitate sleep and successful scanning, all subjects were fed just before the MRI examination. Anatomical MR images were analyzed by an experienced pediatric neuroradiologist and all MR scans were found to be normal without any visible brain abnormalities. The study was approved by the regional ethical committee in Stockholm. The infants comfort during scanning was secured by a pediatrician who resided inside the scanning room during the complete imaging session. Arterial oxygen saturation and heart rate were continuously measured during scanning. All included infants were appropriate for their gestational age with regard to birth weight, height, and head circumference. Further details regarding the infants are given in Table 1.

Table 1 Parameters for the infants included in the study
Full size table

MR image acquisition.

A Philips Intera 1.5 Tesla scanner equipped with a 6-channel receive-only coil was used to acquire both anatomical and fMRI in all studied infants. MR image acquisition parameters were closely matched to the settings used in our previous study in preterm infants (19). Briefly, fMRI of the infant brain were acquired by means of an echo planar image (EPI) sequence sensitized to T2*-weighted BOLD signal changes (TR/TE/flip, 2000 ms/50 ms/80 degrees; matrix size, 64 × 64; field-of-view, 180 × 180 mm). Each EPI volume comprised 20-axial slices (slice thickness = 4.5 mm, interleaved slice acquisition order) yielding a spatial resolution of 2.8 × 2.8 × 4.5 mm3. Resting-state functional connectivity was assessed by recording BOLD signal changes during 10 min of sleep (300 EPI volumes). Four dummy scans were included in the beginning of each fMRI scanning session to achieve steady-state magnetization. Anatomical MRI included a T1-weighted turbo-spin echo scan, an inversion recovery scan as well as a high-resolution three-dimensional gradient-echo image sequence (TR/TE/flip, 40 ms/4.6 ms/30 degrees; acquired voxel size, 0.9 × 0.9 × 1 mm3; reconstructed to 0.7 × 0.7 × 1 mm3). Additionally, T2-weighted turbo-spin-echo images were acquired in sagittal, coronal, and axial slice orientations. The total scanning time amounted to approximately 50 min. Noise protection was provided using a 4-fold protection system including standard pediatric muffs, a dental putty molded into the hearing canal (Affinis dental putty soft, Coltene AG, Switzerland), minimuffs (Natus Medical Inc, San Carlos, CA), and a custom-made acoustic hood (thickness = 104 mm) made from polyurethane that attached tightly to the upper semicircle of the magnet bore.

Image preprocessing.

Postacquisition image processing closely followed the steps outlined in our previous study (19). In brief, nonbrain voxels were excluded (24) followed by correction for subject movement, spatial normalization to an infant T2-weighted template (25), and spatial smoothing (full width at half maximum = 6 mm) within the SPM2 software package (Wellcome Trust Center for Neuroimaging, University College London, UK, Ref. 26). In 10 of 21 subjects, subject movement during scanning was small throughout the whole scanning session with a maximal translational movement of less than 1 mm and a maximal rotational movement of less than 1.5 degrees across all 10 infants. In nine subjects, the realignment procedure showed one or two episodes of brief, sudden jerk-like movements during which the infants head was tilted away from the original position and within 10–30 s titled back again in a position that was very close in space to the original position. For these nine subjects, the image data were treated in same manner as in our previous study (19). Accordingly, the affected image volumes were removed from the fMRI dataset and the remaining image volumes were treated as one continuous dataset (27). The remaining two subjects showed excessive movement throughout the scanning session and the corresponding datasets were discarded. Thus, our final data analysis was performed on 200 fMRI EPI volumes obtained in 19 subjects.

Image analysis.

Intrinsic brain activity at rest in the infant brain was estimated using the temporal concatenation version of the tensorial probabilistic approach to independent component analysis (PICA) as implemented in the MELODIC (version 4.0) module in the FSL software package (Oxford Centre for functional MRI of the Brain, FMRIB, UK, Ref. 28). Independent component analysis is a multivariate method that does not require any seed regions specified by the user to measure functional connectivity (29). Specifically, PICA divides the four-dimensional EPI data into spatially non-Gaussian processes in the presence of Gaussian noise and uses a Gaussian mixture model to test for significance of the extracted spatial maps (30). The alternative hypothesis was tested at p > 0.5 for “activation” versus null to create thresholded results for each spatial map. Besides the ability to produce spatio-temporal components that are likely to be candidates for resting-state networks, ICA is able to divide resting-state fMRI data into maps that show strong resemblance, in both the spatial and temporal domain, to maps typically assigned to other signal sources such as subject motion and nonneuronal physiologic noise. The independent components here shown to represent resting-state networks were selected on the basis of anatomical localization and frequency content. In detail, information from the anatomical localization was used as far as those components that exhibited activity pattern that mainly resided in known major blood vessels, e.g. in the vicinity of the circle of Willis, were discarded. Similarly, components that showed “rim”-like activation patterns typical of subject motion were not considered further. Components for which approximately more than 50% of the activity was judged to reside in gray matter were considered to be of interest. Furthermore, only components for which the majority of the signal variance resided below 0.1 Hz were considered to be relevant.

RESULTS

On the basis of the results from the 19 subjects included in the independent component analysis, we could identify six anatomically coherent resting-state networks with the majority of the signal variance residing in the frequency interval of 0.01–0.05 Hz. An example of the resting-state networks typically observed in individual subjects is shown in Figure 1, whereas the group results are shown in Figure 2. Coherent synchronization of low-frequency intrinsic brain activity was found in the middle part of the occipital cortex (Figs. 1A and 2A), bilateral sensorimotor cortex (Figs. 1B and 2B), bilateral temporal cortex (Figs. 1C and 2C), middle and lateral aspects of the parietal cortex (Figs. 1D and 2D), anterior prefrontal cortex (Figs. 1E and 2E), and bilateral basal ganglia (Figs. 1F and 2F). The relative amount of variance explained by each network in relation to the total amount of variance was 1.63% (visual cortex, Fig. 2A), 3.18% (sensorimotor cortex, Fig. 2B), 0.70% (temporal cortex, Fig. 2C), 4.78% (parietal cortex, Fig. 2D), 1.60% (anterior prefrontal cortex, Fig. 2E), and 0.1% (basal ganglia, Fig. 2F).

Figure 1
figure 1

Intrinsic brain networks found in a typical infant born at term based on a probabilistic independent component analysis of fMRI acquired during sleep. Each row shows in a coronal, sagittal and axial view resting-state networks thresholded at p > 0.5 (alternative-hypothesis threshold for activation vs null) superimposed on a T2-weighted MR image infant brain template. At the individual level, resting-state networks were found in primary visual areas (A), bilateral sensorimotor regions (B), bilateral temporal/inferior parietal cortex including the primary auditory cortex (C), posterior lateral and midline aspects of the parietal cortex (D), medial and lateral parts of the prefrontal cortex (E), and in bilateral subcortical regions (F). The left side of the image corresponds to the left side of the brain. The color-bar shows the corresponding t-value.

Full size image
Figure 2
figure 2

Group consistent intrinsic brain networks in the infant born at term based on an independent component analysis of fMRI acquired during sleep. Each row shows in a coronal, sagittal and axial view resting-state networks thresholded at p > 0.5 (alternative-hypothesis threshold for activation vs null) superimposed on a T2-weighted MR image infant brain template. Resting-state networks at a group level were found in primary visual areas (A), bilateral sensorimotor regions (B), bilateral temporal/inferior parietal cortex including the primary auditory cortex (C), posterior lateral and midline aspects of the parietal cortex (D), medial and lateral parts of the prefrontal cortex (E), and in the bilateral basal ganglia (F). The left side of the image corresponds to the left side of the brain. The color-bar shows the corresponding t-value.

Full size image

DISCUSSION

Overall, our current results are in good agreement with our previous investigation of resting-state activity in sedated preterm infants (19). The spatial distributions of the resting-state networks that encompasses primary sensory cortices (Figs. 1A–C and 2A–C) are very similar to the networks obtained in preterm infants (see for reference Fig. 3A–C in Ref. 19). Further, this observation is in agreement with previous investigations that have shown resting-state networks present in visual, auditory, and sensorimotor areas in the adult brain (7–9,31). Moreover, the network consisting of the medial as well as the lateral parts of the parietal cortex depicted in Figures 1D and 2D was also found in sedated preterm infants, although no involvement of the cerebellum was found in naturally sleeping infants born at full term. Similarly, the anterior network shown in Figures 1E and 2E was also found in sedated preterm infants as well as in adults (7,32,33). Interestingly, the resting-state network shown in Figures 1F and 2F that resides predominately in the basal ganglia was not detected in sedated infants born preterm. This is a novel finding in infants born at term and it is in accordance with a recent study that found a strong bilateral resting-state pattern in the caudatus and putamen in adults (34). At this stage, we consider it too early to relate the absence of resting-state activity in the basal ganglia in our previous study to any factor related to preterm birth or sedation. This is due to the fact that primarily the study performed in preterm babies used a rather small sample size, yielding a too limited sensitivity to adequately address this discrepancy. Further studies using larger cohorts of subjects need to be carried out to address this issue. Thus, the overall impression from our current findings is that they corroborate our previous results in that low-frequency fMRI signal fluctuations are present and mediate long-range functional connectivity already in the infant brain.

The presence of bilateral functional connectivity patterns in primary sensory brain areas are also supported by a recent study by Lin et al. (35) that reported bilateral resting-state connectivity in sensory areas in 38 neonates. However, our recurrent finding of a bilateral resting-state activity pattern in the sensorimotor region in the infant brain as well as the results presented by Lin et al. are to some extent in contrast to the results reported recently by Liu et al. (36). In the Liu study, fMRI was performed in sleeping 1-y-old children and they reported a bilateral activity pattern similar to the pattern shown here in Figures 1B and 2B in only two of 11 subjects, whereas the remaining subjects showed a predominately intrahemispheric resting-state pattern in the sensorimotor region. The authors suggested that the fact that we investigated infants born preterm as well as different levels of sleep were induced, i.e. sedation versus nonsedation, might have contributed to the observed differences. However, both factors were addressed in this study and our present findings of strongly interhemispheric resting-state patterns corroborate our previous results obtained in sedated infants born preterm. It therefore seems unlikely that the unilateral resting-state pattern reported by Liu versus the bi-laterality found in our studies as well as in the Lin study is due to any of the suggested factors. It should also be noted that although the sensorimotor resting-state network obtained in the majority of the infants investigated by Liu et al. showed a predominately lateralized pattern, a small but still significant cluster was also found in the contralateral hemisphere (see Fig. 2A and B in Ref. 36. for comparison). It is likely that the relatively small sample sizes used together with small differences in image analysis strategies might have contributed to the observed differences.

One of the resting-state networks commonly observed in the adult brain that have received much attention is the so-called default-mode network (12–14,37). The default-mode network comprises the medial aspects of the prefrontal cortex, precuneus/posterior cingulate cortex, bilateral parietal cortex, and the lateral and medial temporal cortex. There is a growing interest to understand its role in the brain's functional architecture and how the default network may be affected in disease (see refs. 15 and 18 for recent reviews). Similar to our previous investigation in infants born preterm, we observed what tentatively might be called a proto-default network in the infant brain with a rather strong resting-state functional connectivity between the precuneus/posterior cingulate cortex and the bilateral parietal cortex. However, at a group level, no significant functional connectivity between the posterior/medial aspects of the parietal cortex and the medial prefrontal cortex was observed and no significant connectivity with the temporal cortex was found (Figs. 1D and 2D). This finding is consistent with the idea that the default network matures gradually throughout infancy, childhood, and adolescence (38). The relative strong functional connectivity between the medial and the lateral parietal cortex in the proto-default network (Figs. 1D and 2D) is interesting in the view of recent research on cortical hubs in the adult brain. It has been hypothesized that cortical hubs, i.e. brain regions that show a high degree of centrality in cortical networks including a large degree of anatomical and functional connectivity, play a pivotal role in coordination, and integration of information flow in the brain (39). With regard to the default network, recent MRI work in the adult brain have shown that the precuneus/posterior cingulate cortex exhibit a high degree of both anatomical (40,41) and functional connectivity (42), thus making it a likely candidate for being a cortical hub in the default network. On the basis of the present findings in infants, we tentatively suggest that the maturation of functionally central cortical hubs starts already in infancy.

It is also worth noting that the observed resting-state functional connectivity in infants born at term was recorded during natural sleep. To this end, it is well known that natural sleep consists of several stages, each characterized with its own electroencephalographic signature. Previous work using EEG as well as fMRI has shown that the depth of sleep have an effect on resting-state functional connectivity (43). A recent study have shown that a transition from wakefulness to light sleep introduce a slight increase in functional connectivity in networks that include higher order cortices (44). Similarly, during light sleep compared with wakefulness, increases in functional connectivity in networks residing in primary sensory cortices have been reported (22). Hence, it is conceivable that differences in arousal are related to the observed differences in resting-state connectivity between adults and infants. Moreover, the depth of sleep was not controlled for in this study, and it cannot be ruled out that some of the differences in functional connectivity between naturally sleeping infants born at full-term and sedated infants born preterm can be attributed to depth of sleep.

In conclusion, we could show that the infant brain born at term hosts resting-state networks driven by spontaneous changes in neuronal activity. Our results showed that neither preterm birth nor light sedation seems to have a detrimental influence on the resting-state patterns in the infant brain. We could corroborate our previous findings from the infants born preterm in that several cortical as well as subcortical long-range functional networks mediated via synchronous spontaneous low-frequency BOLD fMRI signal fluctuations are present in the infant brain born at term. Because it has recently been shown that resting-state functional connectivity reflects structural connectivity in the adult brain (45,46), we suggest that mapping of resting-state networks can potentially be used as a tool to probe white matter integrity in the brain. In future studies, we aim to investigate the relationship between anatomical connectivity and resting-state functional connectivity in the infant brain as well as examining the influence from white matter abnormalities typically observed in preterm infants on resting-state functional connectivity.