The development of noninvasive techniques for the assessment of functional brain maturation is critical. The present study analyzed 63 babies' and children's (27 days to 5.5 years) cerebral responses to a pattern-reversal visual stimulation using high-density (128 electrodes) electrophysiological recordings. Developmental data were further compared with those of young adults (n = 16). Tremendous changes in pattern visual evoked potentials (pVEPs) morphology were observed between 7 and 24 months characterized by the emergence of negative components labeled “N70” and “N145” and the reduction of the P100 amplitude. The adult pattern of response appears from 24 months onward. Spectral density values show an increase of higher frequencies with age. Coherence values show a reduction between 3 and 23 months of age as well as a further increase toward adulthood between areas implicated in visual processing. These results are discussed in light of developmental features such as synaptogenesis, myelination, and neuronal networks refinement.
The human brain is known to undergo major changes during the prenatal period and in the 1st years of postnatal life. Important modifications occur at the anatomical and functional levels as a result of both endogenously coded sequences of events and the organism's interaction with its physical environment (Simon and O'Leary 1992; Colonnese and Constantine-Paton 2001). Cycles of accelerated development of nerve cells, axonal arborization, and synaptic links are followed by global gradual elimination (pruning) of exuberant neurons and connections to allow for the emergence of greater specificity and fine-tuning of functions in the nervous system. The aim of the present study was to find a reliable, noninvasive technique for the in vivo investigation of different functional stages of brain development in infants, in light of known developmental features.
Synaptogenesis is an important neurodevelopmental process in our understanding of neurological development and plasticity in the 1st years of life. Studies of synaptic density in animals and humans (Rakic and others 1986; Bourgeois and others 1994; Yamada and others 1997) indicate that synaptic overproduction in the 1st 2 years of life is followed by a reduction of synapses that extends into adolescence. These processes follow specific timetables that differ among brain regions. For instance, according to studies by Huttenlocher and Dabholkar (1997), the synaptic density of the visual cortex reaches its peak at approximately 9–15 months of age and declines afterward to attain the adult level in early adolescence. In parallel, a decrease in cerebral metabolic (Chugani and others 1987) and neurophysiologic indices (Courchesne 1977) occurs at a functional level.
Myelinogenesis is another important feature of brain development. Recent magnetic resonance imaging studies suggest a sudden increase in whole brain white matter between 29 and 41 weeks after conception (Huppi and others 1998). This increase in white matter keeps on from birth to 20 years of age (Courchesne and others 2000).
Visual evoked potentials (VEPs) have proven to be of great value in the study of brain development. Parameters such as latencies and amplitudes of the response components have been linked to myelination and synaptic transmission (Scherg and Picton 1991; Tsuneishi and Casaer 1997). The approach we describe here consists of high-density electrophysiological recordings of VEPs. We used this technique to study the electrophysiological responses at various stages of brain maturation in children aged from 27 days to 5.5 years, as well as in young adults. VEP peaks (N70, P100, N145) were investigated by means of amplitudes, latencies, and brain topographies. In addition, spectral density analysis was performed to provide cues about possible maturational patterns of predominant frequency bands and their implication in the development of visual processing. Coherence of the visual evoked electrophysiological signals was also examined as it is thought to underlie some developmental features, such as cortical synaptic proliferation and “pruning” (Thatcher 1992).
Sixty-three infants aged from 27 days to 5 years 6 months were separated in 6 age groups (Table 1). These subjects were chosen from 120 tested infants and children; the data from 57 children or infants were excluded from the original sample because of their state of arousal (crying or sleeping babies), movement artifacts, and other nonneurological activity (i.e., electrical noise, cardiac rhythm artifacts). In addition, 17 adults aged between 20 and 30 years of age were tested. The data from 1 adult were rejected because of excessive artifacts.
|Groups||N||Mean age (months)||Min–Max|
|1||9||1.30||27 days–1 month 23 days|
|2||13||2.17||2 months–2 months 14 days|
|3||13||4.58||4 months–6 months 16 days|
|Groups||N||Mean age (months)||Min–Max|
|1||9||1.30||27 days–1 month 23 days|
|2||13||2.17||2 months–2 months 14 days|
|3||13||4.58||4 months–6 months 16 days|
Developmental information was gathered from interviews and a developmental questionnaire completed by the parents. All subjects were born at term; they had no history of psychiatric or neurological illnesses and had normal or corrected vision. All children were subjected to an intellectual developmental scale administered by 1 experimenter and 1 observer. An interjudge agreement was established. All subjects 0–24 months of age scored in the normal range (within a standard deviation of 15 from 100) on the mental development index of the Bayley scale of infant development (Bayley 1993). All tested subjects from 2 years to 5 years 6 months obtained a global score in the normal range (within a standard deviation of 15 from 100) on the Stanford–Binet intelligence scale (Thorndike and others 1994). Parents and adult participants signed the consent form authorized by the ethics, administrative, and scientific committees from the Ste-Justine's Hospital research centre and the University of Montreal.
Apparatus and Stimuli
Pattern-reversal VEPs were generated using a black and white checkerboard stimulus subtending a visual angle of 2 degrees and presented at a reversal rate of 1 Hz. Stimuli had a luminance of 40 cd/m2 and were generated by a Dell GX 150 PC using E-Prime 2000 software (from Psychology Software Tools Inc. Pittsburgh, PA). Pattern reversals were presented binocularly at a distance of 70 cm from the subject's eyes and subtended 38.5 × 38.5 degrees of visual angle. Young infants were seated on their parent's lap. Their attention to the screen was driven by small objects held in the lower middle part of the screen by the experimenter. Following a procedure widely used in developmental electroencephalogram experiment (Roy and others 1995), the EEG was recorded only when the children were still and their gaze was focused on the screen center. Recording was done with the 128 electrodes (Electrical Geodesics System Inc., Eugene, OR); the reference was at the vertex, and the impedances were maintained below 40 kOhms, as suggested by Tucker (1993). The EEG signal was amplified and analog band-pass filtered from 0.1 to 100 Hz. The signal was digitized at 250 Hz in 1024 ms epochs. A G4 Macintosh computer controlled data acquisition.
Off-line analyses were performed with Brain Vision Analyser software. Data were digitally filtered with a 0.5- to 50-Hz band-pass filter and re-referenced to an averaged reference. The EEG was subjected to algorithmic artifact rejection of voltage exceeding ±100 μV. Eye movement artifacts were corrected using the Gratton and Coles algorithm (Gratton and others 1983). Visual examination of the segmented (0–1000 ms) data was also performed, and segments with artifacts were manually rejected.
Artifact-free segments were averaged and baseline corrected. Due to the nonstationary nature of EEG signals in infants, standard peak to baseline analysis was substituted by a new technique of VEPs peak detection. Rather than defining positive and negative peaks as global minima and maxima of each lobe, we characterized their timings as the latencies required to reach a 50% proportion of the summed activity in each positive and negative lobe. The amplitudes of each component were defined by the amplitude value reached at that latency with respect to the baseline (100 ms prestimulus). Specific time windows were visually selected for the N70, P100, and N145 peaks at electrode 76, the closest electrode from Oz of the Electrical Geodesic system. Although the component morphologies were generally comparable within the respective age groups, in rare cases some low amplitude components were so close to the baseline that they showed an opposite sign value (i.e., a negative P100). In these cases, a missing value was attributed. Consequently, samples of each evoked potentials component per age group vary slightly (see Fig. 1).
Brain topographies were done using Brain Vision Analyser software version 1.05 (Brain Products, Munich, Germany). They were peaked at the maximal amplitude of each component on the groups' grand averaged evoked potentials. “All in one” maps (90 degrees of maximal angles) and back view maps were used with interpolation by spherical splines and automatic scaling.
Spectral Density Analyses and Coherence Analyses
Spectral density and coherence analyses were performed on the VEP artifact-free EEG data. Using Brain Vision Analyser software (1.05), a mean of 60 (27–105) 1000-ms artifact-free EEG epochs, time locked to the stimulus onset, was chosen for each subject. All epochs were resampled at 256 Hz and submitted to fast Fourier transformation (FFT) with a 10% Hanning window, giving a frequency resolution of 1 Hz. The sum of spectral amplitude values (square root of power) was averaged across segments and extracted for the following EEG bands: delta, 0.5–2.5 Hz; theta, 2.5–7.5 Hz; alpha, 7.5–13.5 Hz; beta1, 13.5–22.5 Hz; beta2, 22.5–32.5 Hz; gamma, 32.5–50 Hz. Data were averaged across 5 electrodes over the occipital, parietal, central, and frontal regions.
In parallel, a complex FFT was calculated, and all EEG bands in uV2 were then subjected to a cross-spectrum/autospectrum coherence analysis over chosen pairs of electrodes (occipital interhemispheric leads, O1–O2; left occipitoparietal leads, O1–P3; right occipitoparietal leads, O2–P4; left occipitotemporal leads, O1–T7; right occipitotemporal leads, O2–T8).
All data were transferred to SPSS software. A logarithmic transformation was applied when data were not normally distributed. Analyses of variance (ANOVAs) and a T2 Tamhane correction on post hoc analyses were used, assuming inequality of variance. Analyses of regression were applied on coherence values.
VEPs: Morphology and Topography
The VEPs responses and topographies shown in Figure 1 consist of averages from 7 to 16 normal individuals per age group (N = 79, see Table 1). Overall, we observe patterns of maturation that reveal a turning point period from 7 to 24 months, where the morphology of the occipital VEPs responses becomes clearly triphasic with significant changes in latencies and amplitudes in the 3 components N70, P100, and N145 (see Fig. 1).
The 0- to 1-month (30 days) group shows a prominent positive response in the absence of marked negative deflections. Except for its greater amplitude, the P100 topography shows an adult-like voltage distribution, with positive and negative activity, respectively, distributed in posterior and anterior regions. As revealed by both VEPs responses at electrode 76 (Oz) and topographical maps, negative components are not well developed at 1 month of age. The N70 can first be visualized around 2 months of age, but the topographical maps show an adult-like distribution only from 3 months onward. Even though the N145 can be identified at 2 months of age, it is only in the oldest children's group (24–66 months) that the topography clearly resembles that of the adults with a distribution over inferior–posterior regions.
The maturation of VEPs polarities can be appreciated from Figure 2, which represents the total sum of 125 data points (uV) over a 500-ms poststimulus interval. It clearly exhibits shifts from initially marked overall positive amplitudes to prominently negative values during the period of 3–24 months, imputable to the emergence of the 2 negative components. A negative/positive balance is established after 2 years of age.
VEPs: Amplitudes and Latencies
One-factor ANOVAs were performed separately on the latency and amplitude data for each of the 3 components (N70, P100, N145). To ensure a normal distribution of the amplitude and latency data, a logarithmic transformation was applied when needed. Significant group effects (P < 0.0005) were found for all computed analyses and a post hoc Tamhane correction analysis demonstrated developmental trends for both positive and negative components (see Fig. 3a,b).
Important variations in P100 amplitude were found across ages. Indeed, the P100 amplitude increases from birth to 2 months and then decreases to attain its minimal value at 13–23 months of age (F6,70 = 6.650, P < 0.0005) (Fig. 3a). A further increase is observed among the children aged 24–66 months (P < 0.05) and adults (P < 0.005). Indeed, the 13- to 23-month group's P100 amplitude differs significantly from both the youngest and the oldest (adult) groups (P < 0.0005).
Both negative components (N70 and N145) appear around 2 months of age and increase gradually in amplitude as their morphology sharpens. One-factor ANOVAs applied on log corrected data show significant group effects (N70: F6,67 = 8.628, P < 0.0005; N145: F6,62 = 6.007, P < 0.0005), which mainly reflect marked gains in amplitude until 13–23 months. In fact, the 13- to 23-month N70 amplitude is significantly different from both the youngest and the adult groups (P < 0.01) (see Fig. 3a). Both negative components show a significant reduction toward adulthood (P < 0.01).
All components follow a significant decrease in latency with age, reaching their adult values around 7–12 months (F6,72 = 33.940, P < 0.0005) (see Fig. 3b). Indeed, the P100 latency significantly shortens from the 1st week of life until 7–12 months (P < 0.0005). In spite of their later appearance, from 2 months onward, the negative deflections show gradual reductions in latencies. Whereas a significant latency decrease of the N145 component occurs around 7–12 months, where it reaches its adult value (F6,72 = 30.272, P < 0.0005), the N70 adult latency level is reached at 24–66 months (F6,72 = 4.604, P < 0.001).
Spectral Analysis: Spectral Density
An analysis of the topographic distributions of spectral density was first performed and revealed that frontal, central, parietal, and occipital regions followed the same developmental patterns. Larger amplitudes of densities in all frequency bands were nevertheless observed in posterior regions, where VEP responses are greatest. Therefore, for data reduction purposes, one-factor ANOVAs were applied to the occipital region only. All frequency bands were found to change significantly with age (P < 0.0005). As expected, absolute spectral amplitudes (uV) decrease with increasing frequency (Fig. 4).
Spectral Analysis: Frequency Bands Ratios
To further investigate the developmental gain in the density of frequency bands and to fairly compare children with adults in whom amplitudes differ partly because of physical conduction reasons, ratios of higher over lower frequency bands were calculated for each age group.
Both high-frequency band ratios over delta and theta bands were calculated, but they yielded similar results. For data reduction purposes, only the ratios of high frequencies over theta bands are reported here. One-factor ANOVAs showed a between-group effect for all ratios examined (P < 0.0005).
The alpha (uV)/theta (uV) ratio tends to increase steadily with age (Fig. 5a) (F6,72 = 44.109, P < 0.0005) with a critical period at 13–23 months where it becomes significantly different from the youngest as well as from the adults. The beta1 (uV)/theta (uV) bands ratio (Fig. 5a) (F6,72 = 46.793, P < 0.0005) shows to some extent a different pattern of development; although increasing slightly until 24 months of age, it displays an abrupt gain at 24–66 months (P < 0.0005) and a significant decrease, thereafter, at adulthood (P < 0.05). Similar to the alpha/theta ratio, the 13- to 23-month group shows a significant difference from both the youngest infants and the older groups.
High bands log-transformed ratio values demonstrate an increase in beta2 (F6,72 = 8.984, P < 0.0005) and gamma (F6,72 = 7.039, P < 0.0005) frequency bands relative to theta, reaching significance at adulthood (P < 0.0005) (Fig. 5a). Again, the 13- to 23-month group seems to represent a turning point in terms of changes in high-frequency amplitude, being different from both the youngest and adult groups.
Because of its suspected similarity to the adult occipital alpha rhythm (Stroganova and others 1999), theta development was further investigated. Theta amplitude was thus calculated in relation to delta spectra (Fig. 5b). Results are congruent with resting state EEG studies, showing a between-group effect (F6,72 = 7.978, P < 0.0005) that is reflected, according to Tamhane corrected post hoc analysis, by an obvious increase at 24–66 months of age, which further continues until adulthood where it becomes significantly different from the youngest group.
Coherence analysis is thought to measure the degree of correlated changes between different brain regions signals. Coherence analyses were performed on interhemispheric occipital leads (O1 and O2, Fig. 6a), as well as intrahemispherically, on the right and left occipitoparietal leads (O1–P3, O2–P4, Fig. 6b) and the right and left occipitotemporal leads (O1–T7, O2–T8, Fig. 6c) for each subject. Coherence values were then averaged per frequency bands. After careful examination of the individual bands' developmental trends, their similarity enabled averaging over homologous sites of both hemispheres as well as a global band averaging. Analyses of regression were then performed on averaged coherence values for the electrode pairs under study.
Figure 6 displays the averaged coherence values per age group. A significant quadratic regression function was found for the interhemispheric coherence between the occipital electrode pair O1–O2 (F2,74 = 13.50, P < 0.0005), the occipitoparietal (F2,74 = 12.96, P < 0.0005) as well as the occipitotemporal (F2,74 = 14.33, P < 0.0005) coherence values. Thus, all coherence values for each electrode pairs were found to follow a U-shaped pattern of development.
This study investigated brain development through pVEPs and their brain topographies, the spectral density of the signal and differences in coherence values from 0 to 5.5 years of age. All results confirm the presence of enormous changes throughout infancy that may be triggered by a number of developmental features.
PVEPs results show latency, amplitude, and morphological changes. As previously reported (Crognale and others 2001), all components' (N70, P100, N145) latencies decrease from 0 to 7 months of age. Indeed, the time window showing the greatest spurts of latency reductions (1 to 3–7 months) corresponds to the myelination period of the geniculate–occipital pathway, namely, from birth to 5 months of age (Yakovlev and Lecours 1967). Interestingly, this period of increased myelination is paralleled by a rapid development of visual competences such as visual discrimination and the use infants make of the visual information (e.g., Eimas and Quinn 1994; Slater 1998).
From a single dominant positive wave at the youngest age, pVEPs responses become clearly triphasic from 3 months onward. Thereafter, morphological modifications are characterized by the appearance of negative waves (N70 and N145) and a transient reduction of the P100 amplitude. The underlying mechanisms leading to the transitory predominance of negative components from 3 to 24 months remain unclear. Nonneuronal reasons, such as the alteration of electrophysiological field potentials by changes in cortical geometry in postnatal months (Regis 1994) or modification in tissue resistance, could affect the morphology of the signal. It is noteworthy, however, that it is at this point in time, namely, between 7 and 24 months, that the occipital region's synaptic density is enhanced. Although the mechanistic of the changes in pVEPs morphology is unknown, it may reflect the activity of these new, transient connections. It is only after 24 months, when occipital synaptic density diminishes as the circuitry is refined, that the adult-like pattern of responses becomes apparent, being characterized by a reduction in negative waves and a gain in P100 amplitude.
Spectral Density Data
Traditionally, slow bands (delta, theta) activity have been linked to drowsiness and lack of attention (Willekens and others 1984; Fisch 2000). In contrast, higher frequency bands have been related to cognitive processing (Tallon-Baudry and others 1999; Bertrand and Tallon-Baudry 2000; Csibra and others 2000; Bazhenov and others 2002; Tamas and others 2004). Most developmental spectral studies have been performed on resting state EEG (Gasser, Verleger, and others 1988; Stroganova and others 1999; Marshall and others 2002). The special interest of this work is the possible new insights it provides into the functional dynamics of different EEG bands in visual processing across infancy, early childhood, and adulthood. In line with resting state EEG studies (Gasser, Jennen-Steinmetz, and others 1988; Martinovic and others 1998), our results show a general amplitude reduction of the spectral density with age, mostly imputable to a decrease in the magnitude of the delta band as the theta/delta ratio demonstrates (see again Figs. 4 and 5b).
The pattern of maturation observed in spectral density of the brain activity seems to mimic the development of visual attention/regulation capacities. In this context, our study has demonstrated a steady increase of alpha absolute and alpha/theta relative values. This result may express the shift in EEG frequency dominance with development measured in resting states studies (Matthis and others 1980; Gasser, Verleger, and others 1988; Clarke and others 2001). The developmental pattern of alpha activity, which is clearly present within our pVEP protocol, could also reflect greater attentional and expectancy capacities (Klimesch and others 1998; Yamagishi and others 2003), which start developing only around 3 months of age (Kostovic and others 1995). Furthermore a drastic beta1/theta ratio gain is observed in the 24- to 66-month group. In continuity with the alpha band relevance to attentional processes, beta activity has been related to voluntary attention enhancement (von Stein and others 1999). The period of strengthened beta1 activity may thus mark the appearance of voluntary attentional behaviors.
Brain rhythms occur through the synchronization and entrainment of neurons (Maex and De Schutter 2003). Intact inhibitory synaptic transmission through subclasses of interneurons is thought to play crucial roles in high-frequency rhythms generations (Whittington and Traub 2003). Indeed, higher frequency activity development arises at the end of the crucial period of synaptic exuberance in the visual cortex (24–66 months and adulthood). Refined inhibitory and excitatory cell assemblies and networks might thus be required for mature brain activity patterns.
It is noteworthy that the highest frequencies (beta2, gamma) were best seen at adulthood. Previous studies have, however, provided indications of a relative independence of high frequencies (20–55 Hz) from pattern VEPs (De Carli and others 2001). High-frequency bands are in fact better correlated with cognitive visual processes (Tallon-Baudry and others 1996) than with early perceptual treatment. In this context, it is quite plausible that the greater proportion of higher frequencies seen in adulthood simply reflects the fact that our adult participants invested high levels of attention in this simple perceptual task.
Coherence analysis is interpreted as a quantitative measure of functional relationship between paired locations (Nunez and others 1999). Our coherence study of visual processing suggests similar patterns of development between networks. Coherence levels follow a U-shaped pattern from birth to adulthood that appears inversely related to the time course of synaptogenesis in the visual cortex (Huttenlocher and Dabholkar 1997). In fact, we found a marked decrease in coherence occurring during the period of 3–23 months in occipito-occipital, occipitoparietal, and occipitotemporal regions. This period of low coherence coexists with a period of high density of synaptic connections and yet, it would seem that in spite of their higher level of connectivity, unrefined connections elicit lower functional relationships between networks during that period of synaptic exuberance.
At occipital interhemispheric sites, the progressive rise in coherence exceeds 5 years of age. This result may reflect the gradual myelination process of posterior callosal fibers visible from 3 to 6 months (Barkovich and others 1992) and extending until early adulthood (Yakovlev and Lecours 1967; Pujol and others 1993; Giedd and others 1996; Giedd 2004). Refinement of visual cortical connections may in turn, allow for the generation of synchronous ensembles of neuronal activity and promote coherence in the 2 hemispheres.
The intrahemispheric increase in coherence also seen until adulthood is more pronounced in the occipitoparietal than in the occipitotemporal leads. Functional differences between the 2 major visual information–processing pathways, namely, the dorsal and ventral streams, could account for this dichotomy. These coherence results may indeed reflect the preferential response of the dorsal stream to our pattern-reversal achromatic stimulus, characterized by low spatial frequency and perceived motion effects.
Visual electrophysiological responses investigated by pVEPs, spectral, and coherence analyses revealed important changes throughout infancy, most pronounced between 7 and 24 months. In spite of the challenges the evaluation of infants of that age represent, the coherence between our results and the present understanding of developmental events confirms that this age span is a period of marked functional and anatomical changes in the developing brain.
Thus, the 7- to 24-month period shows a transient triphasic pattern of pVEPs responses characterized by a dominance of negative components and a decline in P100 amplitude. Interhemispheric and intrahemispheric reductions of the occipito-occipital, occipitoparietal, and occipitotemporal coherences values might reflect a decreased brain efficiency to respond synchronously during this age of synaptic overproduction (Huttenlocher and de Court 1987). Following this period of nonoptimal functionality, the increase of coherence values until adulthood seems to mimic the development of visual networks in their growing capacity to produce constant and phase coherent responses to a visual stimulus.
In the 7- to 23-month period of presumably reduced brain specialization, the particular nature of the evoked responses could also reflect disorganized top–down signaling and feedback connections from higher order to lower order areas implicated in active perception (Hupe and others 2001). Thus, from 24 months onward, the mature pattern of pVEPs, the quantitative decline in low-frequency bands with the quantitative increase in high-frequency bands' density as well as the increases in coherence of the occipital interhemispheric and intrahemispheric “dorsal” connections may all mark the end of a critical period for visual processing development. The underlying mechanisms of this maturation process may include some developmental features such as synaptic refinement, myelination, and inhibitory interneurons development. Our study, however, may not only reflect the development of early visual perception but also the growing attentional capacities allowed by cortical maturation.
This work was supported by the Canadian Institutes of Health Research (postgraduate scholarships to LS and grants awarded to RMS and LM), the Fonds de la Recherche de la Santé du Quebec and INSERM collaboration program (Postdoctoral fellowship to PC), the Canada Research Chair Program (LM), and the Canadian Foundation for Innovation (LM).