We evaluated power and phase-based FC using routine clinical resting-state EEG data in children with DD compared to controls to test for resting-state power and FC differences on global and local levels.
We have confirmed our hypothesis that there are differences in regional power between children with DD and controls. Specifically, we observed differences in the delta and theta bands at both global and local levels, with a particular emphasis on the parieto-occipital regions of the left hemisphere. These regions correspond to the left dorsal attention and visual RSNs. While previous sensor-level studies did not detect power changes in children with DD33,34,75, a study that used LORETA source-reconstruction found lower alpha power within temporo-parieto-occipital sources35. Since sensor-level analyses can be susceptible to volume conduction and allow indirect anatomical localization76, our source reconstructed results offer improved anatomical precision of brain sources showing power and FC differences in children with DD. We found significantly lower delta-theta band power in the left parieto-occipital regions, which can be interpreted as a sign of parieto-occipital hypoactivation as demonstrated by previous fMRI studies of DD14–16, 77. These findings are potentially consistent with the magnocellular theory of dyslexia, which proposes abnormal interactions between magnocellular and parvocellular cells underlying reading-related visual processing deficits, although this theory still requires confirmation25. In this direction, motion perception and VWFA processing deficits have been found previously in children with DD26,27,78, and others have proposed the existence of temporal sampling deficits in the visual system in addition to the temporal sampling deficits for speech/auditory stimuli28,29.
We confirmed our hypothesis that there are resting-state connectivity changes between children with DD and controls. Both ImCoh and wPLI indicate significantly lower FC in theta for widespread left hemisphere vertices and numerous right parieto-occipital vertices. The ImCoh metric showed significantly higher FC predominantly within left fronto-temporo-parietal vertices, but also in multiple right fronto-temporal vertices. These differences correspond to RSNs in both hemispheres, namely, the left dorsal attention, frontoparietal, DMN, somatomotor, and ventral attention, but also the right dorsal attention, DMN, and visual RSNs. The reliability of our results is supported by the fact that two phase-based connectivity metrics take different properties of the EEG signal into account55, and that previous resting-state EEG studies already found differences using the phase lag index (PLI)33,34. Therefore, our data suggests that children with DD show significant differences in FC not restricted to the reading and language networks, but extended to other RSNs. Although the Yeo atlas43 does not provide annotations for the language network, it is an atlas derived from rs-fMRI data and there is evidence from whole-brain rs-fMRI studies of children with DD indicating that their reading-related regions are distributed among different RSNs30,31. Furthermore, fMRI investigations demonstrated decreased FC between the left VWFA and inferior frontal gyrus, and increased FC to the right hemisphere in participants with DD31,78,79. Other rs-fMRI studies have found abnormal FC patterns between multiple regions of the language network and the DMN in children with DD, corresponding to multiple regions where we observed greater alpha or lower theta FC14,20,32. In addition, our results of reduced theta FC in widespread left hemisphere regions are compatible with those from resting-state EEG and MEG studies that pointed to reduced network integration and impaired dynamic information flow in regions recruited during visual word processing and letter-speech sound associations35,36,80, as well as reduced temporal correlations in left temporo-parietal sensors in beta band (local efficiency)37–39. Both findings of increased alpha FC and decreased theta FC within left frontal to occipital regions suggest that abnormal power and FC patterns may underlie temporal sampling deficits for separated or integrated phonological and visual processing5,22,24,29,81. To this extent, they may also underlie the atypical phase synchronization patterns during phonological and speech processing18,21. At the clinical level, the atypical FC in multiple RSNs can be a neurophysiological constraint that may partially explain their phonological processing and reading speed deficits1, 82–85 involving right hemisphere regions as a potential compensatory mechanism to the constraints at the phonological, sublexical, and lexical linguistic processing levels1,16,17,77,86.
The age-stratified comparisons revealed FC differences between the 1st -4th and 5th -8th school grade subgroups, which can be explained by previous research on normal developmental trajectories of EEG power and FC. Specifically, delta and theta power decrease throughout childhood and adolescence, while alpha to beta power increases87–89. Moreover, coherency measures increase globally between 8 and 12 years of age in all frequency bands except theta90,91, whereas wPLI increases in theta, alpha, and beta bands, in fronto-parietal, frontal-occipital, and frontal-postcentral regions, respectively88. The lower wPLI in the theta band observed in our 1st -4th graders is consistent with the finding of logarithmically increasing wPLI values in the left parietal cortex in the theta band across normal development88, which may be a neurodevelopmental marker of difficulties in writing skills during early literacy acquisition, as this region is involved in meaning processing and writing1,88,92,93. Moreover, the increased wPLI in alpha in the left prefrontal regions of 5th -8th graders with DD may reflect increased demands for cognitive control, possibly as a coping strategy in early adolescence88. However, changes in power and FC across development with and without DD are matters outside the scope of the current cross-sectional design. Future investigations with larger samples and a wider age range are needed to test these hypotheses.
Lastly, we found variable correlations between power and FC and cognitive metrics. Increased theta power within the left parieto-occipital regions was correlated with increased spelling scores, whereas a decrease in theta power was associated with improved reading scores. We suggest that the difference in correlation is due to a clear floor effect in the reading scores (Fig. 5A). This makes the spelling data the most robust and useful to interpretation (Fig. 5B). It is emphasized that 20 children with DD assessed with SLRT-II or ZLT-II were unable to read more items than the equivalent to the first percentile rank and, as a consequence, the group showed particularly low z-scores and high standard deviations in reading subtests (Table 2), pointing to a tendency to severe dyslexia in our sample. Therefore, after controlling for age, a significant partial rank correlation between lower theta power in five left parieto-occipital regions and poor spelling performance (Table 4) can be explained by a decreased efficiency in sublexical processing in children with DD, i.e. the ability to use grapheme-phoneme correspondence rules either in the reading or spelling direction94,95. Furthermore, it is possible that decreased power in the left parieto-occipital regions relates to reduced long-range communication with other language-related regions, a finding of the aforementioned resting-state EEG/MEG studies of DD33,34,39. Finally, this finding may be related to parieto-occipital hypoactivation in fMRI14–16, 77,96, disrupted FC between left parieto-occipital regions (including VWFA) and inferior frontal regions32,79, and poorer visual letter-word recognition, meaning processing, and decreased reading time when children with DD are compared to controls35,36,80.