Neuroimaging of cortical development and brain connectivity in human newborns and animal models

Abstract

Significant human brain growth occurs during the third trimester, with a doubling of whole brain volume and a fourfold increase of cortical gray matter volume. This is also the time period during which cortical folding and gyrification take place. Conditions such as intrauterine growth restriction, prematurity and cerebral white matter injury have been shown to affect brain growth including specific structures such as the hippocampus, with subsequent potentially permanent functional consequences. The use of 3D magnetic resonance imaging (MRI) and dedicated postprocessing tools to measure brain tissue volumes (cerebral cortical gray matter, white matter), surface and sulcation index can elucidate phenotypes associated with early behavior development. The use of diffusion tensor imaging can further help in assessing microstructural changes within the cerebral white matter and the establishment of brain connectivity. Finally, the use of functional MRI and resting-state functional MRI connectivity allows exploration of the impact of adverse conditions on functional brain connectivity in vivo. Results from studies using these methods have for the first time illustrated the structural impact of antenatal conditions and neonatal intensive care on the functional brain deficits observed after premature birth. In order to study the pathophysiology of these adverse conditions, MRI has also been used in conjunction with histology in animal models of injury in the immature brain. Understanding the histological substrate of brain injury seen on MRI provides new insights into the immature brain, mechanisms of injury and their imaging phenotype.

Introduction

Understanding human brain development is clinically relevant as many neurobiological disorders and disabilities have their origin in early structural, functional development and plasticity. With the advent of magnetic resonance imaging (MRI), it has become possible to address the question of where, when and how adverse conditions in fetal and early postnatal life and prematurity relate to the maturation of cortical networks (Ment et al. 2009). MRI has entered the research arena due to its non-invasiveness and potential for high-resolution structural brain imaging. It has allowed for the first time the study of in-vivo brain development in premature and healthy newborn infants providing full 3D imaging data sets.The integration of healthy infants in longitudinal studies has become possible thanks to the absence of irradiation or other side-effects, thus allowing a wide and thorough analysis of brain development. This review presents advanced imaging modalities used to study the immature brain and discusses their potential for the in-vivo assessment of cerebral development in both human as well as animal models established to study the pathophysiology of early adverse conditions for brain development.

Conventional magnetic resonance imaging

Conventional MRI techniques, including mainly T1- and T2-weighted images, allow the assessment of brain development in vivo with the highly sensitive assessment of gray and white matter contrast, as well as the differentiation of unmyelinated and myelinated white matter (Battin et al. 1998; Counsell et al. 2002; Ferrie et al. 1999). The immature white matter demonstrates a relatively homogenous low signal on T1-weighted images and a high signal on T2-weighted images compared with gray matter (Fig. 1A). This is mainly due to the higher water content of the immature white matter. With increasing maturation, the signal of the white matter increases on T1-weighted images, which is related to multiple concurrent processes occurring in the developing white matter with decreasing water content, a change in the water : macromolecule ratio caused by the arrival of the lipid precursors of myelination, and finally the process of myelination itself. Myelination is visualized at different rates and times on T1-weighted and T2-weighted images with evidence of myelination determined by a low signal on T2-weighted images and a high signal on T1-weighted images that is already present at 28 weeks of gestation in regions such as the inferior and superior cerebellar peduncles or the ventrolateral nuclei of the thalamus. Myelination of the posterior limb of the internal capsule is seen by 36 weeks with a completion of myelination by 2 years of age (Battin et al. 1998; Counsell et al. 2002).

Fig. 1.

(A) Preterm infant born at 29 weeks of gestation scanned at term equivalent, coronal T2-weighted section with cerebral grey matter and basal ganglia hypointense and non myelinated white matter with a higher water content appearing hyperintense. (B) Rat pup at 9 days of life, coronal T2-weighted section with the same contrast as the preterm infant with hypointense cerebral grey matter and hyperintense cerebral white matter. Image acquired at 9.4 Tesla in collaboration with Gregory Lodygensky and Rolf Gruetter (CIBM, Ecole polytechnique fédérale de Lausanne, Switzerland).

Brain MRI of developing animals in correlation to histological findings gives us the opportunity to better define the substrates of MRI changes during development and following injury. The possibility of imaging mouse pups, although technically challenging, opens a new area of research where specific changes in imaging can be addressed using genetically modified animals (West et al. 2009). The majority of correlation studies between MRI and histology in the developing brain have been carried out on rat pups at 7–14 days of life, an age closer to infants born at term. Imaging of immature animals gives the same contrast in conventional imaging (Fig. 1B) with an inversion of contrast similar to the human as the animals mature (Lodygensky et al. 2008a). Ultra-high field MRI scanners are increasingly used in rodent studies and are now available at up to 14 T, equivalent to 300 000 times the earth's magnetic field. They provide a sufficient signal-to-noise ratio to achieve the required high resolution in immature rodents but bring new challenges such as the assembly of stable coils at the given magnetic fields, worsening susceptibility artifacts and magnetic field inhomogeneity.

Since the discovery of the transient subplate zone (Kostovic & Molliver, 1974) and its significance for the normal development of the human telencephalon (Kostovic & Rakic, 1990a), there has been an extensive amount of effort invested in identification of the subplate by MRI methods (Kostovic et al. 2002; Rados et al. 2006; Huang et al. 2006; Prayer et al. 2006; Kasprian et al. 2008; Kostovic & Vasung, 2009). The subplate zone can be clearly distinguished in ex-vivo T1 MRI after 13 weeks of gestation as an area of hypointensity (Kostovic et al. 2002; Rados et al. 2006; Kostovic & Vasung, 2009) and on in-vivo T2 MRI as an area of hyperintensity (Prayer et al. 2006) between the bordering cortical plate and the intermediate zone (Fig. 2). The presence of this transient subplate zone is one of the main signs of cortical immaturity (Kostovic & Vasung, 2009). It represents the major site of the endogenous circuitry reorganization (Kostovic & Jovanov-Milosevic, 2006; Volpe, 2009). The subplate is crucial in developmental processes such as the ingrowth of thalamo-cortical axons (Kostovic & Jovanov-Milosevic, 2006), reorganization of the fetal white matter and establishment of cortical layers. The subplate is rich in extracellular matrix and serves as a ‘waiting’ compartment for the thalamo-cortical and cortico-cortical connections (Kostovic & Rakic, 1990b). Its neurons are known to contribute, as early as 24 weeks of gestation, to the appearance of the first evoked potentials (Hrbek et al. 1973), which play an important role in the functional maturation of the human brain. Furthermore, after 32 weeks of gestation, in parallel to the gradual decrease in size and extracellular matrix content, the subplate zone becomes less visible on in-vivo MRI (Kostovic & Vasung, 2009).

Fig. 2.

Ex vivo T1 and T2-weighted coronal MRI images acquired 3T of a human brain at 24 weeks of gestation. (A) Coronal T1-weighted image with a resolution of 1 × 1 × 2 mm, (B) Coronal T2-weighted image with a resolution of 0.39 × 0.39 × 0.5 mm. Note the subplate zone clearly distinguished from the bordering fetal zones in T2-weighted image as a hyperintense area between the cortical plate and the intermediate zone and on the T1 weighted image as a hypointense area between the cortical plate and intermediate zone. cp, cortical plate; iz, intermediate zone; sp, subplate zone; vz, ventricular zone.

Qualitative image characteristics such as ventricular dilatation, abnormal white matter signal, increased subarachnoid space, white matter cysts and thinning of the corpus callosum have been shown to predict the neurodevelopmental outcome. These imaging characteristics can therefore help the clinician in the assessment of the impact of prematurity and its co-morbidities on brain development and outcome with better accuracy when compared with head ultrasound (Woodward et al. 2006; Huppi et al. 1996; Ramenghi et al. 2007; Huppi, 2002).

Magnetic resonance image analysis: manual segmentation

Manual segmentation of anatomical structures was the first method used in quantitative image analysis of the newborn brain, mostly performed on high-resolution T2-weighted images during the first months of life and on T1-weighted images later in life (Toft et al. 1995). This approach was chosen to evaluate subtle differences in quantitative brain development caused by premature birth. Several structures quantified by manual segmentation are affected by preterm birth especially in children with associated white matter injury. These structures include the cerebellum, corpus callosum, thalamic and lentiform volumes (Anderson et al. 2006; Srinivasan et al. 2007, 2006; Limperopoulos et al. 2005) all showing a marked reduction in size due to premature birth. Intrauterine growth restriction, a condition in which the placental blood supply to the fetus is reduced and results in poor growth of the fetus, was shown to have an impact on hippocampal volume (Lodygensky et al. 2008b). These changes were shown to persist in former preterm infants during childhood or adolescence with significant alterations in hippocampal volume, the cerebellum and corpus callosum (Allin et al. 2001; Lodygensky et al. 2005; Abernethy et al. 2002; Narberhaus et al. 2007).

Magnetic resonance image analysis: automated computational techniques

High-resolution T1- and T2-weighted images are the basis for the application of mathematically-based segmentation techniques that allow volume measurement of total cortical gray matter, white matter and basal ganglia. Segmentation techniques for the neonatal brain are still challenging due to: (i) an inversion of the contrast between gray and white matter and changes during development prohibiting the use of adult segmentation toolkits; (ii) a strong signal heterogeneity secondary to myelination affecting both cortical structures and white matter regions; (iii) the small size of the brain requiring high-resolution MRI, which leads to a low signal-to-noise ratio; and (iv) thin structures like the cortex, prone to partial volume effects, which further complicate segmentation. The combination of a signal-based k-means classification with a mathematical morphology approach for shape recognition is currently the preferred method for neonatal brain segmentation (Anbeek et al. 2008; Cachia et al. 2003; Weisenfeld & Warfield, 2009).

The recent development of specific newborn brain segmentation tools has allowed the quantification of in-vivo human brain growth showing a fourfold increase in cortical gray matter, a fivefold increase in myelinated white matter, a linear steady growth rate of the unmyelinated white matter (Huppi et al. 1998b) and a 70% growth of the basal ganglia (Mewes et al. 2006). These major changes underline the importance of this period in brain growth and its vulnerability as shown by the toxicity of dexamethasone on cortical gray matter growth (Murphy et al. 2001) or intrauterine growth restriction with a reduction of 30% in cortical gray matter volume (Borradori Tolsa et al. 2004). This approach has shown for the first time that white matter injury in preterm infants was associated with a significant impact on cortical gray matter development (Inder et al. 1999, 2005).

Mathematical morphology: primary cortical folding

One of several possible approaches to investigating cortical folding in preterm infants is based on a mathematical morphology approach that processes and analyzes shape. The computational approach quantifies both surface area and cortical gyration through curvature measurements from 3D reconstruction of the interface between the developing cortex and white matter (Dubois et al. 2008a,b;). The inner cortical surface is identified between the cortical gray matter and white matter. A smooth triangle-based mesh of this surface is then computed and the global area of this inner cortical surface can be measured. Finally, the local surface curvature is estimated from the mesh local geometry; positive curvatures correspond to the gyrus's top and negative curvatures to the fold's bottom (Cachia et al. 2003). The sulci are then defined as connected components of negative curvature. These sophisticated image analysis tools need high-resolution primary input data with no motion artifacts.

Cortical folding in preterm infants

In the human brain, the morphology of cortical gyri and sulci is complex and variable among individuals with an established asymmetry appearing very early on (Dubois et al. 2010). A significant amount of cortical folding and gyrification takes place during the last trimester with a steep increase of brain surface and degree of sulcation index (Dubois et al. 2008b). Conditions such as twinning and intrauterine growth restriction have been shown to alter sulcation and decrease the sulcation index (Dubois et al. 2008a) (Fig. 3). Alterations in size and cortical morphology might reflect abnormal functioning or vice versa, as shown in correlations between surface, gyrification index at birth and the Assessment of Preterm Infants' Behavior score at term equivalent age (Dubois et al. 2008a), a neonatal neurobehavioral test shown to predict later neurofunctional outcome (Feldman & Eidelman, 2006). So far, the precise mechanisms responsible for such alterations in cortical phenotype during intrauterine or postnatal development are still poorly understood and therefore studies of cortical folding in selected animal models will provide further insight into the mechanisms underlying cortical folding (Barnette et al. 2009) (Fig. 4).

Fig. 3.

3D representation of the inner cortical surface for a singleton and a twin of equivalent age. SI1 represents the average of the sulcation index. Note the altered cortical gyrification shown to be significantly affected in twins when compared to singleton of the same gestational age. Modified from Dubois et al. (2008a).

Fig. 4.

(A) Cortical surface of ferret brains at 4, 10, 17 days of life and in an adult. (B) Cortical surface of human brains at 25, 30, 33, 39 weeks of gestation and in an adult. Courtesy of Barnettte et al. (Barnette et al., 2009).

Diffusion tensor imaging

The technique of diffusion tensor imaging

Diffusion tensor imaging (DTI) assesses water molecule displacement at a microstructural level with displacement of water in the order of 10 μm. This technique has been used in the exploration of the structural basis of white matter development (Neil et al. 1998; Huppi et al. 1998a; Mukherjee et al. 2002) and cortical maturation (McKinstry et al. 2002a).

Based on the tensor model, the main DTI parameters are the three eigenvalues λ1, λ2 and λ3 representing diffusion along the three principal axes of an ellipsoid in each voxel. Eigenvector maps and RGB color-coded maps indicate the orientation of the major eigenvector providing an indication of the direction in which water diffusion is highest (typically parallel to white matter fiber fascicles). The first eigenvalue, often referred to as axial diffusivity, was shown to be affected by axonal integrity (Kim et al. 2006). The second and third eigenvectors describe diffusivity perpendicular to the axial diffusivity. Radial diffusivity, defined by the average of the second and third eigenvectors, was shown to be affected in adult animals by changes in myelin ensheathment (Song et al. 2005). From these three eigenvalues is calculated the apparent diffusion coefficient (ADC) and mathematical measures of anisotropy describing the degree to which water diffusion is restricted in one direction relative to all others, referred to as fractional anisotropy (FA) and relative anisotropy (RA) (Fig. 5). FA and RA are indicators of the degree of water diffusion anisotropy with a value that is equal to zero for diffusion equal in all directions and increases with anisotropy.

Fig. 5.

(A) Healthy preterm infant born at 29 weeks of gestation imaged at term equivalent, coronal FA map at the level of the posterior limb of the internal capsule. (B) P5 live rat pup, coronal FA map with sufficient in plane resolution to identify major white matter bundles such as the corpus callosum or the internal capsule. Image acquired at 11.7 Tesla in collaboration with Gregory Lodygensky and Jeffrey J Neil (Washington University, St Louis, Missouri).

Fiber tracking: diffusion tensor imaging

Fiber tracking is able to delineate specific cerebral white matter tracts by following local vector orientation from the 3D vector field (Fig. 6). Different tractography algorithms have been developed such as the popular streamline deterministic fiber tracking (Mori et al. 1999). The 3D fiber track is allowed to continue unless it enters a region of FA less than a predefined value of FA or turns at an angle greater than a predefined angle between two consecutive voxels. Fiber crossing causes one of the major downfalls of fiber tracking. Several algorithms have been proposed to solve this issue such as the probabilistic index of connectivity (Parker et al. 2003) or, more recently, the Gibbs tracking model (Kreher et al. 2008).

Fig. 6.

(A) 3D representation of the cortical surface of an infant at term with superimposed fiber tracking through the corpus callosum. Courtesy Jessica Dubois (CEA/SAC/DSV/DRM/NeuroSpin/Cognitive Neuroimaging Unit, Gif-sur-Yvette, France) (B) Relative anisotropy map of a fixed rat brain at 21 days of age with the superimposed fiber tracking through the corpus callosum. Image acquired at 9.4 Tesla in collaboration with Gregory Lodygensky and Rolf Gruetter (CIBM, Ecole polytechnique fédérale de Lausanne, Switzerland).

Diffusion tensor imaging: analysis strategies

The four main approaches to analyzing DTI data are region-of-interest quantification, voxel-based approaches, histogram analyses and tract-based analysis. Manual region-of-interest selection and quantification is the most frequently used. This method is time-consuming and is characterized by a large inter-rater variability due to a bias in the selection of the regions of interest. Voxel-based morphometry is an alternative method initially designed to quantify regional changes in volume on conventional imaging (Ashburner & Friston, 2001; Kesler et al. 2008; Nosarti et al. 2008). Its automated analysis eliminates a-priori knowledge and user bias but it is not immune to error due to registration and normalization. Tract-based spatial statistics is an alternative method (Smith et al. 2006) developed to address errors in registration by the tools built for conventional imaging used on DTI data. It is based on the realignment of the FA maps of all subjects onto a FA map skeleton.

Diffusion tensor imaging studies

During normal brain development, the ADC of the white matter was shown to decrease together with a steady increase in anisotropy (Huppi et al. 1998a; Neil et al. 1998; Dudink et al. 2007). Following a hypoxic–ischemic injury, ADC also decreases, resulting in values that are much reduced compared with the age- and region-dependent normal values (Rutherford et al. 2004; McKinstry et al. 2002b). This ADC restriction is clinically used to detect brain injury as it appears very early on, preceding conventional T2 hyperintensity. During white matter development, a gradual decrease in diffusion is observed principally in λ2 and λ3 (and much less in λ1), which reflect changes in water diffusion perpendicular to white matter fibers in single coherently ordered fiber bundles and may indicate changes due to premyelination (change of axonal width) and myelination (Mukherjee et al. 2002; Partridge et al. 2004; Giorgio et al. 2008). The increase in white matter anisotropy values during development appears to take place in two stages. The first increase takes place before the histologic appearance of myelin (Huppi et al. 1998a; Neil et al. 1998). This increase has been attributed to changes in white matter structure that accompany the ‘premyelinating state’ (Wimberger et al. 1995). This state is characterized by an increase in the number of microtubule-associated proteins in axons, a change in axon caliber, and the maturation and organization of oligodendrocytes. It is also associated with changes in the axonal membrane, such as an increase in conduction velocity and changes in Na⁺/K⁺-ATPase activity. The increase in anisotropy associated with premyelination is notable in that it takes place in the absence of changes in T1- or T2-weighted imaging as well as before the histologic appearance of myelin. Regional anisotropy is not only clearly influenced by myelination alone, but also by factors such as axon packing, relative membrane permeability to water, internal axonal structure and intra-axonal space as outlined in recent studies of animal spinal cord using AxCaliber, a model of water diffusion that estimates the axonal diameter distribution within a nerve bundle (Assaf et al. 2008).

This is different with intracortical maturation, where maturation is associated with a progressive decrease in FA during development (Deipolyi et al. 2005), confirmed by studies in the developing rat brain (Huang et al. 2008; Sizonenko et al. 2007b) with sufficient resolution to identify clearly the microstructural organization in the cortex (Fig. 5). During rat brain development, ADC and FA changes were detected in cortical layers between postnatal day 3 and 6. A distinct radial organization of the cortical layers with the eigenvectors perpendicular to the pial surface was observed at both ages. This organization was most prominent in the external cortical layers 1–3 compared with the deep layers 4–6. With cortical maturation, ADC was reduced, whereas FA was decreased only in the deep layers of the cortex. Histology revealed maturational differences in the cortical architecture with increased neurodendritic density and reduction in the radial glia scaffolding (Sizonenko et al. 2007b). Similar developmental changes in anisotropy within the cortex have been shown in primates at different gestational ages. At 90 days of gestation, diffusion exhibits prolate symmetry and high anisotropy, reflecting apical neuronal dendrites and radial glia. Later in gestation, with maturation of basal neuronal dendrites, connections with thalamo-cortical afferents, branching of interneurons and reduction of radial glia, the water diffusion was modified and reflected by a reduction of diffusion along the principal eigenvalue λ1 and an increase in the eigenvalues λ2 and λ3 (Kroenke et al. 2005).

The histological correlate of DTI modifications after cerebral hypoxic–ischemic injury at 3 days of life in rat pups (equivalent to a preterm infant born at 24–28 weeks of gestation) showed that DTI was indeed sensitive to microstructural changes with an alteration of the radial organization of the cortex (Fig. 7). The decrease in ADC and FA at 24 and 72 h after injury corresponded to neuronal cell death, astrogliosis, axonal degeneration, and alteration of the radial glia with early transformation into astrocytes (Sizonenko et al. 2005, 2007b). Long-term assessment with DTI showed an altered cortical structure and FA reduction in the corpus callosum and external capsule. The damage resulting from the hypoxic-ischemic injury correlated with functional alteration of the whisker-elicited somatosensory response (Quairiaux et al. 2010). Lipopolysaccharide administration in the developing animal mimics the inflammation-induced white matter injury seen in preterm infants. Intracallosal injection of lipopolysaccharide in 5-day-old rat pups has shown a similar pattern on magnetic resonance-DTI to that seen in a hypoxic–ischemic injury with an initial decrease of the ADC corresponding to the peak of maximal injury. At 4 days after lipopolysaccharide exposure, the ADC was increased together with radial diffusivity corresponding to a decrease in cell density and an increase in callosal thickness and a diffuse astrogliosis. The increase in ADC of the white matter at 4 days after injury reflected diffuse gliosis with increased extracellular space rather than an arrest in development (Lodygensky et al. 2009). Translating this knowledge to white matter assessment in preterm infants with a higher ADC and lower RA worsened by white matter injury would imply a significant injury rather than a delayed development. The acute changes seen in different animal models of preterm brain injury reflect the initial damage on brain structures. These initial tissue and cellular alterations will influence the subsequent brain development in the long term in balance with the intrinsic protective response of the brain (Sizonenko et al. 2007a).

Fig. 7.

Principal eigenvector plot representing a disruption in the parietal cortex of the radially organized cortical eigenvectors 24 h after hypoxia-ischemia. (B) Fluoro-Jade B stain showing degenerating neurons in the same area. Courtesy of Sizonenko et al. (Sizonenko et al., 2007b).

Preterm infants imaged at term even without evidence of white matter injury were shown to have a decrease of the RA in the internal capsule when compared with term infants and a decrease of FA in the splenium of the corpus callosum and coronal radiata (Rose et al. 2008). White matter injury in preterm infants was shown to have an even stronger impact on the microstructural organization with an arrest in the normally decreasing ADC and increasing anisotropy of the white matter together with a decrease of RA in the internal capsule (Huppi et al. 2001; Miller et al. 2002). Tract-based spatial statistics found significant changes in regions within the centrum semiovale, frontal white matter and the genu of the corpus callosum that had a significantly lower FA in preterm infants imaged at term-equivalent age compared with term-born controls (Anjari et al. 2007), thus assessing alterations of brain development in ex-preterm infants. Further experience-related changes in diffusion characteristics have been shown in practicing piano players (Bengtsson et al. 2005) or in working memory training (Takeuchi et al. 2010) and provide evidence in support of the experience-based structural plasticity in the brain. Recently, DTI performed before and after a 2-h task was able to detect a decrease in ADC in the hippocampus with an increase in FA in the right amygdala, bilateral entorhinal cortex and left insula (Tavor et al. 2010). The recent notion that DTI not only can explore microstructural foundations of the brain, but also appears to be sensitive to its functional adaptation opens a new domain of research.

Functional magnetic resonance imaging and functional magnetic resonance imaging connectivity

Functional MRI (fMRI) is based on the detection of regional changes in signals that are correlated with brain functional activity. Neuronal activation is fuelled by an increase in-loco regional blood flow resulting in a decrease of deoxygenated hemoglobin. As deoxygenated hemoglobin has paramagnetic properties, a change in its concentration results in a change in signal intensity that can be quantified. This effect is known as the blood oxygenation level-dependent contrast (Ogawa et al. 1990). The high spatial resolution and relatively fast temporal resolution of fMRI has made it an excellent technique to study brain function in combination with structural MRI and diffusion MRI. Functional MRI connectivity (fcMRI) assesses neural processing (Fox et al. 2005; Schafer et al. 2009) using the spontaneous blood oxygenation level-dependent signal intensity time-course's coherent oscillations within a neuronal network.

Functional magnetic resonance imaging in infants

Previous studies have demonstrated the feasibility of fMRI in healthy newborns and infants at different ages using different passive tasks (Seghier & Huppi, 2010; Seghier et al. 2006). These studies have mainly used sensory stimuli in the visual auditory or sensorimotor domain (Born et al. 1998; Martin et al. 1999; Yamada et al. 1997; Muramoto et al. 2002). Using passive somatosensorial stimulation, somatosensory areas have been identified in the precentral and postcentral gyri but showed weak hemispheric dominance as compared with the somatosensory system in adults (Erberich et al. 2006). Remarkably, speech perception in very young infants showed prefrontal, temporo-parietal, precuneus and angular gyri activations with a significant dominance of the left hemisphere as in the adult brain (Dehaene-Lambertz et al. 2002). The follow-up of perinatal brain lesions combining fMRI with DTI has shown remarkable structural and functional recovery in perinatal stroke (Seghier et al. 2004, 2005). Unresolved issues of fMRI in newborns and infants are the equivocal hemodynamic response influenced by sedation and other mechanisms controlling the cerebral blood flow response to brain activity (Seghier et al. 2004; Colonnese et al. 2008).

Impact of prematurity evaluated by functional magnetic resonance imaging connectivity

Preterm infants scanned at term were shown to have consistent active resting-state networks during sleep in the primary visual cortex, bilateral sensorimotor areas, and bilateral auditory cortex, a network including the precuneus area, lateral parietal cortex, and cerebellum as well as an anterior network (Fransson et al. 2007). Serial resting-state fcMRI every 4–5 weeks was performed on preterms during their initial hospitalization and compared with full-term infants (Smyser et al. 2010). Smyser et al. (2010) identified similar networks and defined their regional growth and their connections with homotopic counterparts during development. In addition to the networks described by Fransson et al. (2007), they demonstrated the presence of interhemispheric connectivity as early as 26 weeks of gestation. The comparison with newborn infants born at term revealed a reduction of long-range connectivity worsened by gray and white matter injury. This study underlines the striking changes occurring during the last trimester of gestation and the extraordinary susceptibility of white matter in children born prematurely. More studies are needed to evaluate the specific impact of prematurity-associated brain injuries such as periventricular venous infarction or diffuse white matter injury and to determine its utility in the clinical arena. Although the acquisition itself is performed in a very short time and holds a mine of unexploited valuable information on brain function, it remains a very challenging imaging technique, as the quantified changes in signal are close to the noise level and susceptible to a variety of artifacts such as motion cyclic cardiac and respiratory artifacts.

Functional magnetic resonance imaging connectivity in animal models

The fcMRI has been studied in rats at 4.7 T (Zhang et al. 2010) with a relatively large voxel size of 0.125 × 0.125 × 1 mm limiting its translation to younger animals with much smaller brains without using a higher field strength and stronger gradients. The fcMRI was established between the prefrontal cortex, thalamus, and retrosplenial cortex and their connectivity with very precise anatomical regions such as the visual, auditory, motor, and somatosensory cortices. Interestingly, fcMRI was studied in an adult animal model of stroke, showing that it may be considered as a new approach to quantify brain plasticity. Indeed, the recovery of the contralateral baseline activation correlated with the recovery of the sensorimotor function (van Meer et al. 2010). The fcMRI not only provides the opportunity to study in detail connectivity between different parts of the brain, but also provides information regarding plasticity after injury, thus opening a new era of research in brain injury too often focused on the lesion itself.

Conclusion

The use of advanced MRI techniques has revolutionized the in-vivo visualization and comprehension of human brain development and function. The advance in microstructural and functional evaluation techniques such as DTI and fcMRI will open up ways to study longitudinally important aspects of brain plasticity during development. The implementation of imaging on high-field magnetic resonance systems in animal models of brain development and injury further allows the translation of the study of mechanisms of development and disease based on experimental data to human applications and the testing of emerging hypothesis coming from human brain magnetic resonance research in animal experiments.