«Published in final edited form as: Brain and Language, 2013 DOI: 10.1016/j.bandl.2013.03.001 Dorsal and ventral pathways in language development Jens ...»
Anatomical connectivity and fiber orientation in brain white matter was investigated by fiber tracking based on the diffusion tensor maps (Anwander et al., 2007; Conturo et al., 1999). Mean DTI data averaged for each group were examined by a whole brain deterministic fiber tracking. Therefore, the preprocessed diffusion images for each group were aligned to a template brain by nonlinear registration (Thirion, 1998) implemented in LIPSIA (Lohmann et al., 2001) and averaged to one dataset. A diffusion tensor was ﬁtted to the combined data, resulting in one averaged diffusion tensor of each voxel in each group, and FA was computed for each voxel. In this way, the averaging was integrated implicitly into the tensor ﬁtting procedure to avoid averaging of diffusion tensors. The fiber tracking algorithm used the entire diffusion tensor to deflect the estimated streamline trajectory corresponding to the fiber tract (Lazar et al., 2003) as implemented in MedINRIA according to Fillard et al. (2007).
Trajectories were started in all voxels (voxel size: 1 mm3) with a FA 0.13 for the children and adults and FA 0.1 for the newborn infants. We applied the same tracking procedure to all three data sets, i.e. for newborns, 7-year-old children and adults. All streamlines crossing two volumes of interest were selected as white matter connections between the two regions. Following Perani et al. (2011), we selected the streamlines connecting ventral precentral gyrus (BA6) with the temporal lobe via the dorsal route as dorsal pathway D1 and the streamlines connecting the IFG (BA44) with the temporal lobe via the dorsal route as the dorsal pathway D2.
The selection of the streamlines of the ventral IFOF was done in the same way with one region-of-interest (ROI) covering the EmC and EC and a second ROI covering the posterior temporal and occipital lobe following the selection criteria proposed by Catani and Thiebaut de Schotten (2008). In a second step, we selected the superficial component of the IFOF connecting to the lateral part of the pars triangularis (BA45) following the method of Sarubbo et al. (2011). The remaining deep component V2 of the IFOF was subsequently subdivided into an orbital component connecting to the frontal pole and the orbito-frontal cortex and a dorsal component connecting to the middle and superior frontal gyrus.
2.4 Quantitative analysis
We measured FA of the diffusion tensor in the different sub-components of the fiber tracts as a microstructural biomarker of structural maturation for the different age groups. Therefore, we computed the individual FA maps for all subjects and normalized them for each group to the template brain that was used for fiber tracking.
The group average FA image was skeletonized with the TBSS (Smith et al., 2006) method. All voxels of the skeleton with a mean FA value smaller than the FA threshold were removed from the skeleton. In a second step, TBSS provided for each point on the skeleton the locally maximal FA value in a small neighborhood perpendicular to the 3D skeleton and copied this value for each subject on the skeleton. For each subject and each tract, all voxels that were crossed by at least one streamline of the selected bundle were selected. In the next step, all voxels of the white matter skeleton within this volume of the fiber bundle were used compute the mean FA values of the skeleton voxel within this tract. In this way, the mean FA value for each bundle was less influenced by partial volume effects at the borders of the fiber bundle compared to a method that uses all FA values within the bundle. In addition, the values were less influenced by the FA values at the noisier fanning endpoints of the fiber bundles. Statistical comparison indicated whether FA values differed significantly. Greenhouse-Geisser correction for degrees of freedom was applied as required (Greenhouse and Geisser, 1959).
Fiber tracking results were compared across the three age groups: newborn infants, 7-year-old children and adults for the dorsal (D1, D2) and the ventral tracts (V1, V2).
We knew that newborns only show dorsal tract D1 terminating in the premotor cortex, and not tract D2, which in adults runs further into the IFG. Adults, on the other hand show both pathways (Fig. 1A and 1C) (after Perani et al., 2011). In contrast to infants, for children we were able to track D2 connecting to the IFG (Fig. 1B, in blue), showing that 7-year-old children possess a connectivity pattern of the dorsal route that is similar to the one observed for adults.
Fig. 1: The dorsal pathways of the language network in newborn infants, 7-year-old children, and adults. For newborns (A), no connection between the IFG and temporal regions is observed. Rather they only show a connection terminating in the premotor cortex. In addition to this dorsal pathway D1 (in yellow), children (B) also show a connection to the dorsal portion of BA 44 in the IFG, thus that dorsal pathway D2 (in blue) is present from at least childhood on. This connectivity pattern for children already closely resembles the one in the mature brain of adults (C) which show both dorsal pathways. The dorsal pathway D2 is assumed to play a crucial role in the processing of more complex linguistic functions, an ability that develops at about 7 years of age.
Figures (A) and (C) after Perani et al. (2011).
For the ventral pathway, we aimed to separate the superficial tract (V1) and the deep tract (V2). The superficial tract (V1) as characterized by its connection to BA 45 in the IFG, was clearly apparent and easy to identify in adults and in children. The parieto-occipital ending of this component appeared more concentrated in the group of children compared to adults which might be related to an overall reduced number of streamlines in this component. This tract was also observed in infants and clearly distinguished from other components of the IFOF but it was weaker than in children and adults (Fig. 2, A-C, in green). Comparison of the number of streamlines for V1 showed 2592 streamlines for adults, 330 for children and 58 for infants. The deep tract (V2) was also identified in all three groups. In adults and children it was clearly advancing into the further frontal regions connecting to the SFG, MFG, DLPFC, and OFC. For infants, its middle and anterior connections were much shorter and did not reach as far as was observed in children and adults.
Fig 2: The ventral pathway of the language network in newborn infants, 7-year-old children, and adults. In newborns (A), the superficial tract (V1) of ventral pathway directed to BA 45 in the IFG (IFOFlat, in green) is already in place, though much weaker than in children (B) who show only small differences in this connectivity pattern compared to adults (C).
Thus, the superficial tract (V1) is the only fronto-temporal connection from the temporal cortex to the IFG in the language network of infants.
For all three groups, the deep tract (V2) of the ventral pathway can be separated into at least two subcomponents (D-F): an anterior component connecting to the orbito-frontal cortex and frontal pole (IFOForb, in blue) and a posterior component (IFOFdors, in red). For the deep tract (V2) there are no differences between the three groups.
We also aimed to differentiate the three subcomponents of the deep tract (V2), as described in Sarubbo et al. (2011). In all three groups, the anterior component directing to orbito-frontal cortex and frontal pole was clearly separated (IFOForb in blue, Fig. 2 D-F) from the middle and dorsal components. However, we were unable to unambiguously differentiate between the latter two components in any of the three groups. Thus, we treated them as one subcomponent directed to SFG, MFG, DLPFC and PMC (IFOFdors in red, Fig. 2 D-F). The resulting two observable components of the deep tract V2 were clearly differentiated in all three groups. Although we were not able to unambiguously differentiate between the middle and the posterior subcomponents of V2, it was evident that in infants the tracked connection to the MFG was shorter than the connection to the posterior SFG and PMC.
Statistical comparison was conducted in order to test for significant differences in mean FA between the three groups and the four tracts observable in all groups.
Corresponding means are displayed in Fig. 3. ANOVA revealed significant main effects for the factor tract: F(3,108) = 15.9, p.001, and group F(2,36) = 47.5, p.001 and a significant interaction F(6,108) = 7.0, p.001. Children showed significantly higher FA values across all tracts (FA =.43, SD =.019) than infants (FA =.26, SD =.018) and significantly lower values than adults (FA =.48, SD =.020).
The lowest FA was observed in the dorsal pathway D1 of infants (FA =.23, SD =.015) and highest FA was observed in the dorsal and orbital portions of adults’ IFOF (FA =.49, SD =.020). Dorsal pathway D2 was, unlike in infants, evident in children and adults. Statistical comparison between the two latter groups reveal that although this tract is in place in children (FA =.42, SD =.024) similarly to adults (FA =.48, SD =.020), it is still not fully mature in children: t(1,18) = 5.8, p.001.
Fig. 3: Mean FA values for the dorsal pathway D1 (AFpmc) and the
three components of the ventral pathway (the lateral portion of the IFOF:
IFOFlat (V1), the dorsal portion of the IFOF: IFOFdors (V2), and the orbital portion of the IFOF: IFOForb (V2) in infants, children, and adults.
Infants show lowest FA values for all tracts, while adults show highest.
In order to test whether the observed smaller anisotropy for fiber tracts in newborns is specific for fronto-temporal fiber pathways or whether this is a reflection of the general immaturity of the infant brain white matter, FA in fronto-temporal white matter (averaged across all four tracts present in all age groups) was compared to the mean FA across the whole cerebral white matter skeleton. In a 2 (white matter, WM) × 3 (Groups) ANOVA, newborns’ FA in fronto-temporal pathways (FA =.255, SD =.018) was slightly but not significantly smaller than white matter mean FA (FA =.260, SD =.012). For children and for adults, fronto-temporal FA (children: FA =.429, SD =.019; adults: FA =.484, SD =.020) was significantly higher than mean FA (children: FA =.347, SD =.019; adults: FA =.390, SD =.014). This was reflected in a Group × WM interaction (F(2,35) = 43.7, p.001).
Fiber tracking and group comparison can be affected by individual variability, measurement noise, or crossing fibers. However, group-level tracking results were confirmed on the individual level for the present comparison. Sample data for the dorsal connection from three randomly selected individuals from each group are presented as Suppl. Content. Individual data confirmed that tract D2 was not present in any of the newborns, while it was clearly distinguishable in all children and adults.
Moreover, we added 3D views for the group tracking results which can be rotated and allow a closer and detailed inspection of the data (see Suppl. Content).
There are two main fiber connections in the language network connecting temporal and inferior frontal language-relevant areas in the brain, a dorsal one via the AF/SLF and a ventral one via the IFOF/ECFS. In the present study, we analyzed the maturation patterns of these pathways across three age groups, infants at birth, 7year-old children, and adults.
Both streams consist of pathways and tracts with different structural terminations and functional roles. Dorsal pathway D1 terminates in the PMC and has been proposed to be relevant for auditory-motor integration which is important for early language learning (Friederici, 2011). Dorsal pathway D2 terminates in the dorsal IFG in BA 44 of Broca’s area and was proposed to be relevant for more complex linguistic functions (Brauer et al., 2011).
The two dorsal pathways show a distinct pattern of maturation during ontogeny.
Dorsal pathway D1 is evident from very early on in life and it was observable in newborn infants, unlike dorsal pathway D2 which was not observable in this age group (Perani et al., 2011). In our analysis, however, we showed that pathway D2 is present in children at age 7. The tracking result for children was clearly different from that of newborn infants and rather resembled that of adults.
Nevertheless, it is important to note that although the termination of the dorsal pathway D2 in the dorsal IFG appears complete in children, it is still not yet fully mature as reflected in significantly lower FA for this tract at age 7 compared to adults.
It is at around this age that children start to successfully process more complex and also passive sentence structures (Hahne et al., 2004) or object-first constructions (Dittmar et al., 2008; Knoll et al., 2012), functions suggested to be supported by the dorsal connection to BA 44 (Friederici, 2011).
The ventral pathway contains two tracts, a superficial tract (V1) terminating in BA 44 in the IFG and a deep tract (V2) terminating in the prefrontal cortex in the DLPFC, PMC, MFG and OFC. Unlike for the dorsal pathways, for the ventral pathway the tract connecting the language-relevant cortices, i.e., tract V1 terminating in BA 45, reveals a fast maturation and is already evident in newborns but still immature as indexed by FA. Also deep tract V2 can be identified in newborn and already shows the same connections as evident in adults. However, the tracking shows that it is the most posterior part of the frontal terminations, connecting to the PMC, SFG and DLPFC that show the longest terminations, while the terminations to the MFG and OFC are shorter. For children and for adults these terminations reach farther into the more anterior part of the frontal lobe in fiber tracking.
The middle and the posterior component of the IFOF’s deep tract V2 were not unambiguously separable in our analysis. Therefore, we treated them as one component.
Nevertheless, based on their terminations, they can be roughly differentiated. Note that the posterior component terminating in the PMC, DLPFC, MFG and SFG as shown in our data is not always included as part of the IFOF in some studies (Vandermosten et al., 2012 and Voineskos et al., 2010) while it is included in others (Lebel & Beaulieu, 2011; Lebel et al., 2012; Sarubbo et al., 2011 and Thiebaut de Schotten et al., 2012). This discrepancy is probable a result of different tracking protocols and procedures (for a discussion on tracking protocols, see Gierhan, 2013).