SEARCH

SEARCH BY CITATION

Keywords:

  • EEG;
  • morphometry;
  • slow wave sleep;
  • structural MRI

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

Sleep studies often observe differences in slow wave activity (SWA) during non-rapid eye movement sleep between subjects. This study investigates to what extent these absolute differences in SWA can be explained with differences in grey matter volume, white matter volume or the thickness of skull and outer liquor rooms. To do this, we selected the 10-min interval showing maximal SWA of 20 young adult subjects and correlated these values lobe-wise with grey matter, skull and liquor thickness and globally with white matter as well as segments of the corpus callosum. Whereas grey matter, skull thickness and liquor did not correlate significantly with maximal slow wave activity, there were significant correlations with the anterior parts of the corpus callosum and with one other white matter region. In contrast, electroencephalogram power of higher frequencies correlates positively with grey matter volumes and cortical surface area. We discuss the possible role of white matter tracts on the synchronization of slow waves across the cortex.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

Slow wave activity [SWA, electroencephalogram (EEG) spectral power between 0.75 and 4.5 Hz] during non-rapid eye movement (NREM)] sleep is a well-established marker of the homeostatic regulation of sleep (Borbély, 1982). SWA increases in proportion to the time spent awake and decreases during sleep and, thus, reflects an electrophysiological measure of ‘sleep pressure’. Over the past decades, many aspects of SWA have been investigated and have yielded important insights into sleep regulatory processes. For instance, the dynamics of SWA was quantified by mathematical models (Achermann and Borbely, 2003), the topographical SWA distribution was suggested as individual traits of functional anatomy (Finelli et al., 2001) and several authors reported the use-dependent regulation of SWA (Huber et al., 2004; Kattler et al., 1994). Most of these analyses were performed on relative or normalized values. However, it is still largely unclear which factors determine the absolute level of SWA under baseline conditions. Several polymorphisms have been associated with the absolute SWA level (Landolt, 2008). For example, a genetic variant of the adenosine deaminase associated with a reduced metabolism of adenosine to inosine was related to increased sleep SWA (Retey et al., 2005).

SWA also shows remarkable maturational changes from childhood through adolescence (Campbell and Feinberg, 2009): the activity of slow waves during sleep increases in the first years of life, reaches its maximum shortly before puberty and declines thereafter throughout adolescence. A recent study demonstrated that these changes are associated with cortical grey matter thickness (Buchmann et al., 2011). Although the maturation of the cortex reaches an asymptote in adults, absolute values of SWA still vary substantially between individuals.

In this paper we examined the relationship between SWA in healthy adults and different aspects of brain anatomy [accessible by structural magnetic resonance images (MRI)] and determined the amount of variance of SWA explained by several anatomic variables. We particularly considered thickness of the skull, external liquor space as well as cortical grey and white matter volumes.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

Subjects

We investigated 20 healthy adult subjects [nine females, age range 18–35 years, mean age 25.2 years (SD 4.1 years)]. The study protocol was approved by the local ethics committee. Subjects were either recruited for pilot experiments (= 12) or served as adult control subjects for a cohort of children and adolescents (= 8, three of whom were already included in Buchmann et al., 2011). Informed consent was obtained from all study participants.

Sleep EEG

Eight subjects underwent EEG recordings during two nights; in 12 subjects only one night was available (pilot experiments). Sleep EEG recordings were performed using a 128-channel EEG amplifier (Electrical Geodesics Inc., Eugene, OR, USA), sampled at 500 Hz and band-pass filtered between 0.5 and 50 Hz. Sleep stages were scored visually for 20-s epochs according to the American Academy of Sleep Medicine (AASM) criteria (Iber et al., 2007). After visual and semiautomatic artefact removal, spectral analysis of consecutive 20-s epochs was performed for the first four NREM sleep episodes [fast-Fourier transform (FFT) routine, Hanning window, averages of five 4-s epochs, software package matlab; The Math Works, Inc., Natick, MA, USA]. Maximal SWA (0.5–4.5 Hz) during NREM sleep (stages N2 and N3) was estimated by (1) calculating SWA over 10-min intervals for the first four NREM sleep episodes for each cluster of electrodes and (2) of those intervals, selecting the one with the maximal SWA value. EEG electrodes were assigned to clusters overlaying the main lobes of the brain as follows: all electrodes more anterior than the line Cz–C3–C4 were assigned to the frontal electrode cluster (60 electrodes); a circle around Cz including P3, Pz and P4 (24 electrodes) to the parietal electrode cluster; the occipital electrode cluster included 17 electrodes around O1, Oz and O2 (close to, but excluding Pz); and the temporal electrode cluster contained 26 electrodes around T3–T6. Power values were then averaged for the electrodes we had assigned to each electrode cluster (over both hemispheres). When more than one night was recorded (= 8), values were averaged across the two nights. In these subjects the intraclass coefficients of logarithmized maximal SWA were between 0.760 (parietal cluster of electrodes) and 0.879 (occipital cluster); for the frontal cluster it equalled 0.867. Power values were log-transformed to approximate normal distribution. The same procedure was used for the theta (5–7 Hz), alpha (8–11 Hz) and sigma (12–15 Hz) frequency ranges for examining the frequency specificity of the findings.

Magnetic resonance (MR) images

All images were obtained with a 3T General Electrics Signa HDx scanner (Milwaukee, WI, USA). We used T1-weighted gradient-echo whole brain images, TR 8.928 ms, TE 3.496 ms, flip angle 13°; image resolution in the x–y–z direction was 256 × 256 × 40 voxels, resulting in a resolution of 0.938 × 0.938 × 1.2 mm.

MR anatomy

We divided the brain into the four main lobes for each hemisphere and then averaged the surface area and volume values for both hemispheres and calculated cortical thicknesses as means weighted with the volumes of the respective structures. This division allowed estimation of the influence of the skull and the enhancement of the signal by the grey matter volume (and the surface area) separately for frontal, parietal, temporal and occipital lobes. The white matter was investigated globally, because a division into lobes would be unreliable.

Measurement of skull thicknesses, liquor space, head size and shape

The thickness of the skull and the size of the liquor space were measured manually using Mricron (Chris Rorden; http://www.mricro.com). For this purpose, the original T1 images were overlaid with the skull segment obtained by the ‘new segment’ procedure in SPM8. To increase accuracy of this relatively simple technique, we used average measurements of four (parietal and occipital lobe), six (temporal lobe) and seven predefined landmarks (frontal lobe) per lobe. The aim was to choose the points as representatively as possible for the direction of the electrodes, but always measured along coordinate axes of the aligned brain scans. The landmarks used for the frontal lobes (F1, F2) were the points over to the outermost borders of the putamina left and right, in the xz-plane of the anterior cingulate (measured in parallel to the z-axis); (F3) over the most anterior point of the callosum (z-axis); (F4, F5) from left and right frontal poles anteriorly (y-axis); and (F6, F7) from the most anterior point of the callosum left and right (x-axis). For the parietal lobes (P1, P2) we used the outmost (left and right) points of the lateral ventricles in the xz-plane of the anterior limit of the cuneus, measured upwards (z-axis) and (P3, P4) in the same plane, the points measured outwards (along the x-axis) from the anterior limit of the cuneus. For the temporal lobes, all points were measured outwards (x-axis) on both sides, namely from the anterior cingulate (T1, T2), the upper limit of the mesencephalon (T3, T4) and the lowest point of the lateral ventricles in the xz-plane of the posterior limit of the callosum (T5, T6). For the occipital lobes, all thicknesses were measured backwards (y-axis), namely from the most posterior point of the callosum (middle; = 0) (O1), from the outmost points of the lateral ventricles left and right in the xz-plane of the most posterior part of the callosum (O2, O3) and from the dorsal–posterior limit of the cuneus (adjacent to the precuneus) (O4). For the determination of head size, brain length, width and height was measured on the levels crossing the origin of the Montreal Neurological Institute (MNI) coordinate system (marked by the anterior cingulate) using Mricron. In order to assess if brains have a spherical or elongated form, an eccentricity index was defined by the following formula: (brain length0.5 × brain width0.5 × brain height)/brain length.

Voxel-based morphometry (VBM analysis)

Images were analysed with SPM8 for Matlab (http://www.fil.ion.ucl.ac.uk/spm/software/spm8) (Ashburner and Friston, 2000). All images were transformed from GE dicom to nifti-format, realigned manually along the AC–PC line and then normalized, segmented, bias-corrected and hidden Markov random field (HMRF)-corrected using a unified model, which optimizes the errors of all these processes, as implemented in the new VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm; see also Bitter et al., 2010). This sophisticated method models image noise and mixed tissue classes for each volume element (voxel). For the normalization we used the variant basing on diffeomorphic warping (Dartel; Ashburner, 2007) included in VBM8 and standard a priori maps. Warped images were smoothed with an 8-mm Gaussian kernel to increase the signal-to-noise ratio and reach a Gaussian distribution of the data.

Image statistics were calculated on a voxel × voxel basis. Grey matter images were masked explicitly using a relative threshold of 0.8 to exclude voxels that are unlikely to belong to the grey or white matter segments. We used multiple regressions to assess the volume–EEG power relationship with sex and age as covariates of no interest. The statistical threshold was < 0.05 with familywise error (FWE) correction.

Surface-based analysis

Local grey matter volumes (cortical and subcortical), the surface areas of a predefined set of cortical gyri/sulci (i.e. area of a gyrus/sulcus in a spherical model of the cortex) and mean thicknesses of the cortex in these gyri/sulci were calculated using Freesurfer version 4.5.0 for Mac OS 10.5.2 [http://surfer.nmr.mgh.harvard.edu; see also (Dale et al., 1999; Fischl et al., 1999)]. Freesurfer statistics were evaluated with spss version 16.0 for Windows (SPSS Inc., Chicago, IL, USA), using multiple regressions with sex and age as covariates of no interest. Effect sizes were assessed on standardized beta weights of the multiple regressions. Statistics for the brain lobes were obtained by summing up all regions of each lobe (areas, volumes) and averaging over lobes (for thickness) using weighted means corrected for the grey matter volumes of each structure. The statistical threshold was < 0.05 uncorrected [the number of independent comparisons was limited to five due to the highly correlated nature of the EEG and structural magnetic resonance imaging (MRI) data; see Results].

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

The visually scored sleep variables indicated good sleep quality of the subjects (Table 1, e.g. mean sleep efficiency of 88%). Sleep latencies and percentage of sleep stages were in agreement with the current sleep laboratory data in the literature (e.g. Finelli et al., 2001; Retey et al., 2005).

Table 1.   Sleep variables. Eight subjects slept two nights in the lab, 12 individuals only one night. If applicable, data were averaged over the two nights
VariableMeanSEM
  1. Sleep latency to non-rapid eye-movement sleep stage N2, WASO waking after sleep onset. Sleep stages are expressed as a percentage of total sleep time. Sleep efficiency indicates total sleep time expressed as a percentage of time in bed.

Sleep latency (min)21.73.7
REM latency (min)113.210.4
Waking after sleep onset (min)32.76.1
Stage 1 (%)9.10.9
Stage 2 (%)58.01.7
Stage 3 (%)13.91.5
REM (%)19.01.1
Total sleep time (min)356.69.0
Sleep efficiency (%)88.11.9

First, we investigated the relationship of SWA across the electrode clusters. SWA values correlated highly between electrode clusters (from 0.934 between electrodes over frontal and temporal lobes to 0.969 between electrodes over temporal and occipital lobes; all < 0.001). Therefore, unless stated otherwise, in the following analysis we report SWA data measured over the frontal lobe. For the higher-frequency bands the correlations were weaker (theta: between 0.836 and 0.947; alpha: between 0.854 and 0.947; sigma: between 0.792 and 0.972). Grey matter volumes also correlated significantly between lobes (between 0.463 and 0.799), surface areas between 0.475 and 0.761 and average cortical thickness between 0.606 and 0.893 (all < 0.05).

We then investigated the association between maximal SWA of the frontal lobe and specific anatomical measures. In a first step, we addressed the question of whether global brain size and shape is related to maximal SWA. Multiple regressions (corrected for sex and age) did not show a significant relationship between SWA and brain eccentricity (see Methods; for frontal electrodes: standardized beta = 0.116, = 0.640), nor did we find a significant effect for brain size (beta −0.022, = 0.927).

In a second step, we investigated the tissue between the brain lobes and the electrodes including skull and liquor space. Skull thickness ranged from 3 to 5.5 mm [e.g. frontal lobe from 3.29 to 5.86 mm; mean 4.65 mm, standard deviation (SD) 0.68 mm]. The liquor space was much more variable (e.g. frontal lobe from 0.86 to 5.29 mm; mean 3.01 mm, SD 1.34 mm). Multiple regressions (correcting for sex) did not show any significant relationship between maximal SWA and thickness of the tissues around the brain, either for the skull, or for the liquor, or for the sum of both (beta weights were distributed around zero for the different lobes; see Table 3 for mean effect sizes).

Table 3.   Effect sizes of the relationship between different anatomical variables and power in the slow wave, theta, alpha and sigma (spindle) frequency ranges. Effect sizes were calculated with multiple regressions corrected for sex and age respectively for sex only
VariableSlow wavesThetaAlphaSigma
Betan minBetan minBetan minBetan min
  1. Beta denotes standardized weights of the multiple regression; n min denotes the minimum sample theoretically needed to find a significant correlation, if our sample was representative for the population (a large n means that the effect is probably small or nonexistent). The beta values are averaged (using Fisher’s Z values) over betas for frontal, parietal, temporal and occipital electrodes.

  2. CSF, cerebrospinal fluid.

Corrected for sex and age
 Brain eccentricity0.1451440.0226299−0.146142−0.061818
 Brain size−0.0983160.1152290.247490.28536
 Skull thickness−0.27340−0.063767−0.0953370.080475
 Thickness outer CSF0.0481322−0.04118130.1271880.20869
 Gray matter
  Total volume−0.0598750.178950.397180.5549
  Volume right cortex0.01984450.313300.502110.5629
  Surface area−0.0569710.281380.468130.5519
  Average thickness0.20969−0.0013 049 000−0.002762 200−0.126191
 White matter
  Total volume0.205720.57780.520100.39918
  Volume right hemispheres0.1541280.465130.503110.47113
  Callosal volume0.58180.332270.255460.20075
  Volume anterior callosum0.537100.341250.252470.129182
Corrected for sex
 Brain eccentricity0.071604−0.01221 173−0.142150−0.077513
 Brain size−0.277390.00684 6930.213660.19480
 Skull thickness−0.35623−0.083442−0.0923590.0491269
 Thickness outer CSF−0.161117−0.1321740.0963300.119214
 Gray matter
  Total volume−0.1421500.1341690.428160.52110
  Volume right cortex0.1701040.335260.494110.5708
  Surface area−0.231560.1631140.415170.43815
  Average thickness0.396180.109256−0.00762 223−0.0401905
 White matter
  Total volume−0.1122420.294340.382200.21764
  Volume right hemispheres−0.197780.1671080.392190.23554
  Callosal volume0.325280.203730.225590.129182
  Volume anterior callosum0.404180.275390.242510.120211

In a third step, we assessed the association between lobar grey matter and maximal frontal SWA. Multiple regressions (corrected for sex and age) did not show any relationship between maximal SWA and grey matter, neither for volumes, nor surface areas, nor thickness, for any of the four brain lobes (e.g. the beta weight for prefrontal lobe volume was 0.195, = 0.383, for parietal lobe volume −0.142, = 0.520).

We investigated whether global and callosal white matter is related to maximal SWA. Multiple regressions (corrected for sex and age) between maximal SWA and whole brain white matter (as calculated from the SPM8 segmentation) failed to show an association (frontal electrodes: beta = 0.269, = 0.306; occipital electrodes: beta = 0.182, = 0.464). Volumetric analysis of the corpus callosum, however, showed robust positive correlations between the volume of the corpus callosum and maximal SWA, both in the VBM8 analysis (Fig. 1) and the Freesurfer analysis (Table 2). In addition, all five segments of the corpus callosum showed positive beta weights with significant results for the anterior and the posterior segment (Table 2). As a stability check for both this result and the maximal power values, we correlated callosal volumes with the maximal power values in both the first and second night and found a tendency towards a respectively significant result only for the mid-posterior corpus callosum (first night: beta = 0.668, = 0.094; second night: beta = 0.709, = 0.039). This subsample contained an outlier in the mid-anterior callosal volume, which probably explains the lack of significance in the anterior part of the callosum, because in this case Freesurfer could have shifted the border between these two segments.

image

Figure 1.  Localization of the strongest correlation in the anterior corpus callosum. The colors code the T-values which are a measure of the effect size (df = 18; thresholded at = 4.3, = 0.0002, uncorrected for multiple comparisons). The MNI coordinates of the maximum are −8, 12, 19.

Download figure to PowerPoint

Table 2.   Effect sizes of the multiple regressions between callosal volumes and maximal log transformed slow wave activity (SWA) during NREM sleep; separately for average SWA across the electrodes of the four main brain lobes (see Methods for electrode details). Effect sizes are given in terms of standardized beta weights of the multiple regressions, corrected for sex and age
Region of CCFrontal electrodesParietal electrodesOccipital electrodesTemporal electrodes
  1. 0.05 < P < 0.10; *P < 0.05; **P < 0.01 (corresponds to P < 0.05 corrected for multiple comparisons); ***P < 0.001.

Posterior0.483*0.469*0.4210.558*
Mid-posterior0.2440.2390.1920.320
Central0.2180.2160.1860.242
Mid-anterior0.3810.3640.2960.390
Anterior0.606***0.500**0.498**0.539**
Total0.617**0.561**0.509*0.631**

The voxel-wise volumetric analysis allowed for the localization of one additional correlation between white matter volume and maximal SWA in the posterior temporal lobe (MNI coordinates 56, −49, −8). There were no regions correlating negatively with SWA at the statistical threshold chosen (see Methods).

We also calculated multiple correlations between maximal power in the theta, alpha and sigma frequency ranges (see Methods) with white matter. Whereas correlations with the volume of the whole corpus callosum failed to be significant for the other frequency ranges and showed decreasing effect sizes with increasing frequency (theta: beta = 0.416, = 0.080; alpha: beta = 0.428, = 0.067; sigma: beta = 0.336, = 0.149), we found significant correlations with white matter volume for both theta (beta = 0.613, = 0.018) and alpha (beta = 0.616, = 0.016) and a trend for sigma (beta = 0.492, = 0.057).

We finally estimated effect sizes of macroscopic anatomical variables on absolute SWA, theta, alpha and sigma measured on the scalp and the theoretical number of subjects needed to obtain significant effects (Table 3). The variables to explain most of the individual variability in maximal SWA were situated in the white matter. The relationship between white matter and SWA was most pronounced for the anterior and posterior regions of the corpus callosum (Table 2). The relatively small sample size of this study only allowed showing statistically robust differences for white matter volume, but not for other variables, such as grey matter volumes, thickness of the skull or the size of the outer liquor rooms (Table 3). For higher frequencies, however, there were increasing effect sizes of the multiple correlations between several grey matter variables (volume, surface area) and EEG power, reaching statistical significance at least for the sigma power range (e.g. frontal derivations: whole brain grey matter volume: beta = 0.592, = 0.011; right hemispheric grey matter volume: beta = 0.610, = 0.009; for the alpha frequency range, right hemispheric grey matter volume (beta = 0.546, = 0.027) as well as prefrontal and temporal grey matter volumes (0.582, = 0.006, respectively, 0.496, = 0.030); see also effect sizes estimated from the data of all electrodes in Table 3). Total surface area was not significant for alpha (effect size beta = 0.464, = 0.051), but was significant for sigma (beta = 0.575, = 0.013).

The positive correlations with white matter volumes for these higher-frequency ranges shifted from local effects to more global effects, but with a non-significant (> 0.1) tendency towards higher effect sizes for the right hemisphere than for the left hemisphere (left hemispheric white matter volume: for theta beta = 0.302, = 0.256; for alpha 0.360, = 0.165; for sigma 0.316, = 0.217; right hemispheric white matter volume: for theta beta = 0.513, = 0.069; for alpha 0.593, = 0.029; for sigma 0.589, = 0.027; whole brain white matter volume: for theta, beta = 0.613, = 0.018; for alpha 0.616, = 0.016; for sigma 0.492, = 0.057).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

Sleep EEG studies often report individual differences in absolute SWA during NREM sleep. This study estimated how much of interindividual differences may be explained by variables derived from conventional anatomical (T1-weighted) MR images.

Relationship between white matter and slow wave activity

Individual differences in the volume of the corpus callosum explain 38% of the variability of SWA during NREM sleep (Table 2). The relationship seems to be specific for SWA, because no association was found between callosal volumes and power in higher frequencies (theta, alpha, sigma). An interpretation of this finding may be that callosal fibre connections play a role for the synchronization of slow cortical activity. Several studies support the importance of interhemispheric connections for SWA and slow wave synchronicity/coherence, where coherence represents a correlation measure in the frequency domain. A sleep study in healthy human subjects showed strong interhemispheric coherence for frequencies below 10 Hz and progressively weaker interhemispheric coherence for higher frequencies (Achermann and Borbely, 1998a). This finding was observed in all sleep stages including REM sleep. Interestingly, the interhemispheric coherence of slow waves was stable across sleep episodes, whereas SWA showed a marked decrease (Achermann and Borbely, 1998b). The contribution of the corpus callosum for coherent activity is illustrated by the reduced interhemispheric coherence during NREM sleep in congenitally acallosal children (Koeda et al., 1995; Kuks et al., 1987) and adults (Nielsen et al., 1993), as well as in adults with callosotomy (Montplaisir et al., 1990) and mice with total callosal agenesis (Vyazovskiy et al., 2004).

The importance of long white matter tracts for the synchronization of activity in the slow wave frequency range may not be confined to the corpus callosum, but should also include long intrahemispheric tracts (e.g. antero-posterior). It has been reported that slow waves during NREM sleep resemble travelling waves (Massimini et al., 2004). They start predominantly in the anterior-medial prefrontal cortex and then propagate along the midline axis towards the occipital cortex. Longitudinal tracts could accelerate or enable the propagation of these travelling waves.

We found that the volumes of the anterior two segments of the corpus callosum explain more variance of SWA than the three posterior segments. The segmentation of the corpus callosum used by Freesurfer follows Witelson’s scheme (Witelson, 1989), where the anterior segment contains the fibres connecting the prefrontal cortices and the mid-anterior segment the fibres connecting the motor cortices. We note that the anterior regions left and right of the midline, where slow waves are elicited most frequently (‘triggering zones’), are connected by these two anterior segments of the corpus callosum.

Several mechanisms may explain how the corpus callosum could boost SWA: first, the probability and/or the speed of midline-crossings of the travelling waves could be higher with increased corpus callosum volume, which in turn would increase the size of the regions over which electrodes can still detect slow waves. Secondly, reciprocal transcallosal connections between the two triggering zones in each hemisphere could synchronize the neuronal activity within these zones, leading to constructive instead of destructive interference.

It is not clear whether the size of white matter volume corresponds rather to larger fibre calibre or to more fibres. Both larger fibre calibres or thicker myelination may affect conduction velocity. Aboitiz et al. (2003) hypothesize that specialized fibres with high conduction velocity could be responsible for the synchronization of activity. Intracellular recordings of neurones in the cat’s visual cortex have shown that network activity synchronization needs a spike dispersion smaller than about 10 ms (Engel et al., 2001; Singer, 1999), which is about twice as fast as the average conduction delay across the corpus callosum of about 26 ms. The postulated ‘fast track’ for interhemispheric transmission fits well with the travelling wave data by Massimini et al. (2004), who showed small delays between the left and right hemispheres. These issues should be investigated in more detail with more sophisticated methods (e.g. diffusion tensor imaging in humans).

Relationship between grey matter and sleep EEG power

This study failed to show a significant relationship between grey matter volumes or cortical surface area and maximal SWA. This finding is somewhat surprising, because the surface area of the brain is approximately proportional to the number of cortical columns (Rakic, 1995) which modulate the number of pyramidal neurones available to display the characteristic up-states and down-states during NREM sleep (Amzica and Steriade, 1998; Steriade et al., 1993). The synchronized up- and down-states influence the macroscopic potentials measured on the scalp (Vyazovskiy et al., 2009). The geometry of the two cortical sheets within the skull, however, complicates an association between grey matter volume and SWA. Thus, in general, a larger area leads to higher convolution of that sheet which influences the orientation of the cortical columns towards the electrodes. Such changes in the orientation of the cortical columns have unclear consequences for the power measured on the scalp EEG. Nevertheless, the negative result regarding the correlation between grey matter and SWA does not necessarily mean that grey matter is not involved in the generation of slow waves, but rather that simple macroscopic measures such as the surface area are not linked to SWA in a straightforward manner. Evidence is accumulating that slow waves are connected tightly to the functional properties of cortical synapses (Tononi and Cirelli, 2003; Vyazovskiy et al., 2009), which are not easily accessible with structural MRI. However, in a study with children and adolescents, we found significant correlations between cortical thickness and SWA (Buchmann et al., 2011). This relationship might be an effect of microscopic changes in the neurophil. However, in adults, such microscopic grey matter changes are smaller compared to the changes during childhood and adolescence, which might be a reason for the lack of a relationship with grey matter in adults.

Note that the grey matter correlates we are reporting here for the power in faster frequency ranges are qualitatively different from those reported recently (Buchmann et al., 2011), in that (1) the power depends on the cortical surface, but not on cortical thickness and (2) the effect was found in adults and is not age-dependent, is therefore unlikely to show developmental differences, but rather stable differences between adult subjects.

Effects of skull and liquor layer thickness

We did not find any significant correlations of skull thickness or liquor space and SWA in our sample (Table 3). Thicker layers of skull and liquor between electrodes and brain should lead to lower SWA depending on the dielectric properties of these layers. The skull acts as a low-pass filter (Gabriel et al., 1996). Accordingly, the effects of the skull on SWA are low. However, electromagnetic properties of the human skull are difficult to determine, because in vivo the skull is very different compared to the skull ex vivo. The conductivity in the human skull is about 20 times lower than in soft tissue [about 0.015 S/m (Oostendorp et al., 2000)] and depends on the subject age; that is, the low water content in older subjects leads to higher resistance (Hoekema et al., 2003). Overall, the skull does not seem to be a good insulator, because of its high water content in vivo. Furthermore, the variability in skull thickness between subjects is small (see Results). Conversely, the variability of the size of the liquor space is high between subjects. However, the influence of liquor on SWA is limited due to higher conductivity of liquor compared to the skull [about 1.8 S/m; (Oostendorp et al., 2000)].

Taken together, these results indicate that the electrical properties of skull and liquor space do not affect SWA values significantly in young adults. However, it is not clear whether this conclusion also holds for subjects at other ages.

We have to note that the size of the sample was relatively small. Thus, we estimated sample sizes needed to find a significant relationship for all variables based on our results, with or without age correction (Table 3). A larger sample may have revealed additional anatomical variables contributing to interindividual differences of SWA.

Summary and Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

Interindividual differences in corpus callosum volume predicted up to 38% of SWA variability across subjects. The significance of this finding is augmented when the large day-to-day variability of SWA, due presumably to differences in sleep–wake history, is taken into account. In the future, anatomical white matter differences may have to be considered when reporting group differences in SWA.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References

This work was supported by the SNF Professorship Grant PP00A–114923 to R. H., a research grant of the Children’s Hospital Zurich to R. H., a research grant of the University of Zurich (Nachwuchsförderungskredit) to A. B. and a grant from the University Research Priority Program (URPP) of the University of Zurich to R. H. and O. G. J.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Summary and Conclusions
  8. Acknowledgements
  9. Conflicts of Interest
  10. References