153 successful datasets were obtained from 159 infants who underwent scanning, representing a successful scanning rate of approx. 96%. However, not all children were placed in the scanner on the first visit. 52 infants and toddlers (between the ages of 16 and 60 months) either failed to fall asleep on their first visit or were hesitant about going in the scanner. Of these, 28 chose not to return. The other 24 returned for a second visit and all were successfully scanned. These rates compare favorably to other infant imaging studies (Courchesne et al., 2011; Dubois et al., 2006; Lenroot and Giedd, 2006
), and show that imaging across this challenging age-range is possible with appropriate imaging sequences and research team. Time required for each visit (from arrival, falling asleep and scanning) ranged from less than 45 min to more than 3 h.
Corresponding matched axial slices through the MWF, T1
maps of a representative infant from each age-group are shown in . For data under 9 months, T2
values were calculated for each imaging voxel, however, beyond 9 months T2
was only calculated for voxels with T1
values less than 3500 ms. This differential processing is the reason behind the differing presentation of the 3 and 6 month mean T2
maps (showing values within the ventricles and surrounding cerebral spinal fluid) and the remaining T2
maps in which these areas appear masked. The MWF data illustrate the progressive advancement of myelination throughout the brain. The spatio-temporal pattern of myelination demonstrated by these data closely resembles post-mortem studies (Yakovlev and Lecours, 1967
). We observe myelination beginning in the cerebellum and internal capsule prior to 3 months. Myelination then proceeds to the splenium of the corpus callosum and optic radiations; the occipital and parietal lobes and body of the corpus callosum; and genu of the corpus callosum. The last regions to myelinate, as predicted by histology, are the frontal and temporal lobes. Analogous trends of maturation are seen in the quantitative T1
data, which also show progressive reductions in white and gray matter values across the age span.
Fig. 5 Matched axially-oriented slices through the mean MWF, T1 and T2 maps from each age-group. For the 3 and 6 month data, T2 values were calculated for all voxels. For 9 months and above, T2 was only calculated in voxels with a corresponding (more ...)
Mean male and female MWF trajectories for each of the investigated white matter regions are shown in . Superimposed on the mean data are the logarithmic curves calculated for each genders data. Using an F-test to discriminate regions with gender differences, we found significant (p < 0.05 corrected for multiple comparisons) male/female growth differences in the genu of the corpus callosum, left frontal white matter, left temporal white matter and right optic radiation. In each case, females showed an increased developmental rate compared to males. A summary of the fit curve equations and F-stat values is shown in . No hemispheric differences in growth rate were found ().
Fig. 7 Myelination trajectories, separated by gender, for each left hemisphere and midline white matter region and pathway spanning 83 through 2040 days of age. Points represent the mean value obtained from each region. Female values are denoted by gray (more ...)
Fig. 8 Myelination trajectories, separated by gender, for each right hemisphere white matter region and pathway spanning 83 through 2040 days of age. Points represent the mean value obtained from each region. Female values are denoted by gray circles. (more ...)
Table 6 Calculated logarithmic fits to each brain region for each gender. An F-test was used to determine if the data justified modeling the data independently by gender. Values in bold type denote regions were the male and female data were significantly different (more ...)
Table 7 Calculated logarithmic fits to each brain region for each hemisphere. An F-test was used to determine if the data justified modeling the data independently. Values in bold type denote regions were the right and left hemisphere data were significantly (more ...)
Overall, the myelination trajectories follow a sigmoidal shape, with a lag period followed by exponential growth over the first 12–16 months and slower growth from 2 through 5 years of age. By 60 months, the white matter regions are approaching myelin water fraction values close to 0.2, approx. 80% of values measured in adult white matter (Deoni et al., 2011
). Previous studies of white matter development (Bartzokis et al., 2010; Cho et al., 1997
) show white matter and overall brain growth continues into the second and, in some brain regions, third decade of life.
Comparison of MWF and R1 (1/T1) vs. age curves for each region is shown in . MWF and R2 (1/T2) vs. age curves are shown in . Both R1 and R2 follow approximately logarithmically shaped curves, whereas the MWF curves are sigmoidal. This reflects the differential sensitivity of these measures. Myelin is not histologically present at birth (except in the cerebellum and brainstem) and, thus, the MWF values are zero in all other brain regions. In contrast, as T1 and T2 reflect the presence of water, measurable values are always present.
Comparison of MWF (gray circles) and R1 (1/T1) (black squares) trajectories for each white matter region and pathway across the age-span.
Comparison of MWF (gray circles) and R2 (1/T2) (black triangles) trajectories for each white matter region and pathway across the age-span.
The MWF and R1 curves appear to have similar shape and slope between 2 and 5 years of age, with the curves approximately parallel to each other. This may suggest that they reflect similar changes over this period. In contrast, the R2 curves quickly plateau at 2 years, showing only subtle increases beyond this age point. This contrasts with the MWF curves, which continue increasing throughout childhood.
To examine the associations between MWF, T1 (R1) and T2 (R2) more quantitatively, Pearson R values were calculated (and converted to T statistics) across the full age range; as well as across more defined developmental periods (0 and 6 months of age; 6–12 months;12–24 months; 24–36 months; 36–48 months; 48–60 months). A summary of these results is provided in . Statistical significance was defined at p < 0.05 (uncorrected for multiple comparisons). When all data was included, statistically significant correlations between MWF and R1 were found in all regions except left occipital white matter and right optic radiation. Between MWF and R2, statistically significant correlations were found in all regions except the left cerebellum.
Summary of associations (T-statistics) between MWF and R1 (1/T1) over different brain white matter regions and pathways. Bold type denotes significant (p < 0.05, uncorrected) correlations.
Summary of associations (T-statistics) between MWF and R2 (1/T2) over different brain white matter regions and pathways. Bold type denotes significant (p < 0.05, uncorrected) correlations.
When the relationships between MWF, R1 and R2 were investigated over more discrete age periods, a less coherent picture is noted. For example, focusing on R2 measures, between 3 and 6 months, MWF and R2 are well correlated in all investigated regions except the right cerebellum. Between 6 and 12 months, only the corpus callosum, bilateral internal capsule and corona radiata, and right optic radiation and right cingulum show significant correlation. Between 12 and 24 months, the body and genu of the corpus callosum, bilateral frontal, cingulum, corona radiata and left internal capsule and temporal white matter show significant correlation. By 36 months, there is no correlation between MWF and R2.
Investigating R1 vs. MWF, we note different regions with significant correlations than observed with R2. For example, under 6 months, body and splenium of the corpus callosum, bilateral cingulum, optic radiation, corona radiata, parietal white matter, and left frontal and internal capsule are significantly correlated with MWF. Between 6 and 12 months, values within the optic radiations, parietal white or splenium of the corpus callosum are no longer correlated, but the genu, temporal white matter and bilateral internal capsule are. As with R2, by 36 months few areas show significant correlations between R1 and MWF, and those that do are surprisingly negative correlations (increased MWF corresponds to decreased R1—or increased T1).
Investigating these relationships further, we also sought to determine the direct relationships between MWF and R1 and R2, accounting for the effect of age. Age-corrected partial correlations (Pearson R) were calculated for the defined developmental periods (0 and 6 months of age; 6–12 months;12–24 months; 24–36 months; 36–48 months; 48–60 months). These age groups were chosen because the data could be approximated as linear within these regions. Results of this analysis are shown in , with statistical significance defined at p < 0.05 (uncorrected for multiple comparisons). These results yield sporadic associations, mainly between MWF and R1 between 12 and 24 months of age, suggesting that MWF is a complementary, but distinct, measure of maturation from relaxometry values.
Summary of age-corrected partial correlations (T-statistics) between MWF and R1 (1/T1) over different brain white matter regions and pathways. Bold type denotes significant (p < 0.05, uncorrected) correlations.
Summary of age-corrected partial correlations (T-statistics) between MWF and R2 (1/T2) over different brain white matter regions and pathways. Bold type denotes significant (p < 0.05, uncorrected) correlations.
Cumulatively, the results shown demonstrate the ability to reliably acquire high quality data over the early childhood age-range. Though optimized imaging protocols were used for each age group, the signal-to-noise ratio (SNR) of the calculated MWF maps (myelin water fraction to noise ratio, MWFNR) was not consistent across the age-spectrum, ranging from a low of 7 at 6 months of age, to a high of 20 by 2 years of age (and remaining at this value throughout the rest of the age range). This value was obtained by calculating the mean MWF value/standard deviation for each of the brain regions investigated, and the averaging across the regions. This variable MWFNR may have future implications for modeling, necessitating a weighted least squares approach, as well as in the ability to accurately discriminate subtle MWF differences at early ages. Of note, however, quality and cross-sectional agreement of the data contradict a recent theoretical analysis of mcDESPOT (Lamkford and Does, 2012
), which cautioned that the method was incapable of producing reproducible results. The qualitative agreement with prior histologically-determined patterns of myelination and white matter development further underscore the ability of mcDESPOT to provide salient information related to myelin content through the quantification of myelin water fraction. It is, however, possible that mcDESPOT is influenced by additional effects, such as magnetization transfer.