Metabolic correlates of cognitive function in children with unilateral Sturge–Weber syndrome: Evidence for regional functional reorganization and crowding

Abstract

To evaluate metabolic changes in the ipsi‐ and contralateral hemisphere in children showing a cognitive profile consistent with early reorganization of cognitive function, we evaluated the regional glucose uptake, interhemispheric metabolic connectivity, and cognitive function in children with unilateral SWS. Interictal 2‐deoxy‐2[¹⁸F]fluoro‐D‐glucose (FDG)‐PET scans of 27 children with unilateral SWS and mild epilepsy and 27 age‐matched control (non‐SWS children with epilepsy and normal FDG‐PET) were compared using statistical parametric mapping (SPM). Regional FDG‐PET abnormalities calculated as SPM(t) scores in the SWS group were correlated with cognitive function (IQ) in left‐ and right‐hemispheric subgroups. Interhemispheric metabolic connectivity between homotopic cortical regions was also calculated. Verbal IQ was substantially (≥10 points difference) higher than non‐verbal IQ in 61% of the right‐ and 71% of the left‐hemispheric SWS group. FDG SPM(t) scores in the affected hemisphere showed strong positive correlations with IQ in the left‐hemispheric, but not in right‐hemispheric SWS group in several frontal, parietal, and temporal cortical regions. Significant positive interhemispheric metabolic connectivity, present in controls, was diminished in the SWS group. In addition, the left‐hemispheric SWS group showed inverse metabolic interhemispheric correlations in specific parietal, temporal, and occipital regions. FDG SPM(t) scores in the same regions of the right (unaffected) hemisphere showed inverse correlations with IQ. These findings suggest that left‐hemispheric lesions in SWS often result in early reorganization of verbal functions while interfering with (“crowding”) their non‐verbal cognitive abilities. These cognitive changes are associated with specific metabolic abnormalities in the contralateral hemisphere not directly affected by SWS.

1. INTRODUCTION

Neuronal plasticity and the capacity for functional reorganization in the developing immature brain have been suggested as reasonable explanations for limited impact on the pattern of specific cognitive deficits in children compared to adults (Vicari et al., 2000; Max, 2004; Warrington et al., 1986; Goodman & Yude, 1996). Functional reorganization can be studied by neuroimaging, which may show increased neuronal activity in areas ipsilateral to a lesion (consistent with intrahemispheric reorganization) or in contralateral homotopic areas (suggesting interhemispheric reorganization) or (Seghier et al., 2001; Staudt et al., 2002; Liégeois et al., 2004; Lidzba et al., 2006). Interhemispheric transfer of language functions is the best‐studied model of functional reorganization in humans (Dick et al., 2013). Effective transfer of language functions to the right hemisphere may also interfere with non‐verbal abilities, consistent with the classic concept of functional “crowding” (Teuber, 1974; Strauss et al., 1990).

Sturge–Weber syndrome (SWS) is a sporadic neurocutaneous disorder characterized by a facial port‐wine stain and leptomeningeal vascular malformation (Bodensteiner & Roach, 2010). Due to the early unilateral progression of damage in the affected hemisphere, unilateral SWS is a unique model to study brain reorganization in children (Chugani et al., 1996). Unlike motor or visual functions that show imaging signs of reorganization in SWS (Batista et al., 2007; Jeong et al., 2015), human cognition is believed to arise from an interplay among widespread interconnected cortical areas (Dehaene & Changeux, 2011). The effects of functional reorganization in the contralateral hemisphere on cognitive functioning and their metabolic correlates are poorly understood.

In order to address this issue, in the present study we utilized a recently established approach of voxel‐based PET analysis by using a pseudo‐normal pediatric age‐matched control group enabling us to objectively map glucose metabolic abnormalities in children above 1 year of age (Jeong et al., 2017). The main goal was to evaluate the impact of unilateral brain injury (as indicated by regional cortical metabolic abnormalities) on cognitive abilities and assess cognitive and metabolic patterns consistent with functional reorganization and crowding in children with left‐ vs. right‐hemispheric SWS.

2. METHODS

2.1. Subjects

From our database of children with SWS who participated in a prospective, longitudinal neuroimaging study, we selected 27 patients (Table 1) who fulfilled the following inclusion criteria: (a) unilateral SWS diagnosis defined by the presence of a facial port‐wine birthmark and evidence of brain involvement on imaging suggesting leptomeningeal venous malformation in the temporal, parietal and/or occipital lobe in one hemisphere, detected by contrast enhanced MRI; (b) age 2.5–13 years; (c) average seizure frequency less than weekly, in order to diminish the adverse effect of severe seizures on cognitive function. Seizure frequency was evaluated by parent interviews and categorized by three different scores, “1”: <1 seizure per year, “2”: 2–11 seizures per year, “3”: 1–3 seizures per month, as adapted from our previous study (Bosnyák et al., 2016). Patients with pure frontal lobe involvement (without posterior involvement) were not included.

Table 1.

Clinical and imaging data of the 27 children with SWS, listed from the youngest to the oldest, separately in left and right hemispheric subgroups

Affected side	Age (year)	Gender	Lobe(s) involved on MRI	Lobe(s) involved on PET	Epilepsy duration (year)	Seizure frequency score	FIQ	VIQ	PIQ
Lt	2.5	F	O	OP	2.2	3	111	96	100
Lt	2.5	F	P	P	0.5	2	102	110	93
Lt	2.5	F	FrTPO	FrTPO	2.2	2	89	100	82
Lt	3.1	F	TPO	TP	1.1	2	112	123	97
Lt	3.1	M	P	FrTP	0	1	96	102	90
Lt	3.4	F	PO	TPO	0	1	104	113	93
Lt	3.4	F	TPO	TPO	3.3	2	78	90	66
Lt	4.3	M	FrTPO	FrTPO	3.9	2	78	78	75
Lt	6	F	FrTPO	FrTPO	5.7	1	69	76	66
Lt	7.5	F	FrT	TPO	7	1	107	114	102
Lt	8.2	M	PO	FrTP	7.6	2	100	95	106
Lt	9	F	FrTPO	TPO	2	2	87	89	100
Lt	10.7	F	FrTPO	FrTPO	10.5	1	52	67	57
Lt	12.7	M	FrPO	FrTPO	11.2	3	77	83	71
Rt	3	M	TPO	TP	2.5	1	92	90	97
Rt	3	F	TPO	FrTPO	2.6	2	88	110	76
Rt	3.3	F	FrTPO	FrTPO	0.8	2	96	107	84
Rt	3.7	M	TPO	TPO	3.1	2	86	97	79
Rt	3.8	M	TPO	TPO	2.2	1	92	102	84
Rt	4.3	M	P	FrP	3.4	3	95	102	88
Rt	4.4	F	FrTPO	FP	3.4	2	93	102	81
Rt	4.6	F	P	P	3.8	3	116	112	114
Rt	5.5	F	FrTPO	FrTPO	5.2	1	90	88	93
Rt	6.2	F	O	TPO	3.7	2	101	112	100
Rt	8.7	M	TPO	TPO	7.4	3	60	55	65
Rt	8.9	M	PO	TPO	7.1	2	76	88	69
Rt	10	M	TPO	TPO	3	1	101	104	98

Lt: left. Rt: right. M: male. F: female. Fr/T/P/O: frontal/temporal/parietal/occipital. FIQ: Full scale IQ. VIQ: verbal IQ. PIQ: performance IQ.

To improve the accuracy using an objective SPM analysis to detect glucose metabolic abnormalities in children with SWS, the present study utilized an age‐matched pseudo‐control group of 27 children who underwent 2‐deoxy‐2‐[¹⁸F]fluoro‐D‐glucose (FDG)‐PET due to their history of epilepsy. The use of such a pseudo‐control group for objective PET analysis of children above 1 year of has been demonstrated by our group recently (Jeong et al., 2017). The PET scans for the pseudo‐control pediatric group were selected from a large pediatric PET database of 1,500 children, all performed on the same scanner using the same acquisition protocol, with EEG monitored during the uptake period. First, PET scans described as normal (no focal findings or diffuse metabolic abnormalities) were selected based on the original reports generated by one of the co‐authors (H.T.C.). The selected PET scans were digitally retrieved and re‐reviewed by two additional investigators (A.K. and C.J.), followed by a session where the investigators came to a consensus and selected a total of 64 scans for children 1–18 years of age. The 27 PET scans for the present study were selected from these 64 scans to match the ages of the 27 SWS children as much as possible (mean age: SWS: 5.5 years vs. pseudo‐controls: 5.6 years [range: 2–11.3 years]). All patients in the pseudo‐control group fulfilled the following criteria: (a) No structural lesion on MRI, (b) normal glucose metabolic pattern including lack of significant asymmetries, focal decreases or increases, or global cortical changes by expert visual assessment on interictal FDG‐PET, (c) seizure frequency less than weekly, (d) No significant developmental delay or psychiatric diagnoses. Duration of epilepsy (SWS: 4.0 vs. control: 3.1 years) and age at seizure onset (SWS: 1.5 vs. control: 2.5 years) were not significantly different (p ≥ .1) between the two patient groups. Children in the control group had variable seizure types, but the majority (n = 19) had focal (mostly complex partial) seizures, similar to the SWS group; 7 patients had generalized onset seizures (5 with generalized tonic‐clonic seizures, 1 with myoclonic seizure, 1 epileptic spasms). Twenty‐three of the 27 children had less than monthly seizures. The Human Investigation Committee (HIC) of Wayne State University granted permission for the longitudinal clinical and neuroimaging study and multimodal comparisons of the children with SWS, and parents signed an informed consent form. We also had permission from the HIC to use de‐identified clinically acquired FDG PET scans from children with epilepsy.

2.2. Neuropsychology assessment

All SWS children underwent a formal neuropsychological evaluation at the time of the PET scan to assess cognitive, language, and motor functions. According to age, Wechsler Pre‐primary and Pre‐school Scale of Intelligence (Third Edition) and Wechsler Intelligence Scales for Children (Third Edition) provide quantitative metrics for verbal (VIQ), non‐verbal (PIQ) and global cognitive functioning (full‐scale IQ, FIQ) that have been used in our previous studies in children with SWS (Behen et al., 2011; Bosnyák et al., 2016).

2.3. FDG‐PET acquisition

All participants underwent interictal FDG‐PET scanning using a GE Discovery STE PET/CT scanner (GE Medical Systems, Milwaukee, WI). The scanner has a 15 cm field‐of‐view and generates 47 image planes with a slice thickness of 3.1 mm. The reconstructed image resolution was 5.5 ± 0.35 mm at full width at half‐maximum (FWHM) in‐plane and 6.0 ± 0.49 mm at FWHM in the axial direction. Intravenous injection of 5.29 MBq/kg of FDG was followed by a 30‐min uptake period. EEG was monitored throughout the tracer uptake period to ensure interictal images. Forty minutes post injection, a static 20‐min emission scan was acquired parallel to the canthomeatal plane.

2.4. Data analysis

SPM Diffeomorphic Anatomical Registration Through Exponential Lie Algebra (DARTEL, http://www.fil.ion.ucl.ac.uk) procedure (Ashburner, 2007) was applied to create three age‐specific FDG PET templates from FDG PET images of the pseudo‐control group for children < 3 years, 3–6 years, and 6–12 years old. The use of such age‐specific pediatric PET templates has been demonstrated by our group to detect hypo‐ and hypermetabolic cortical regions in non‐SWS children with epilepsy between age 1 and 18 years (Jeong et al., 2017).

The resulting age‐specific templates were then used to normalize PET images of individual age‐matched SWS patients and pseudo‐normal controls using SPM 12 DARTEL deformation toolbox. For each subject, the normalized FDG‐PET image was compared at voxel level to the age‐matched control data using a two‐sample t‐test of SPM 8 package, with age as a covariate. Statistical thresholds of extent k > 50 and uncorrected p‐value < .001 corresponding to SPM(t) score >3.5 for hypermetabolism and SPM(t) score <–3.5 for hypometabolism were applied to identify significant clusters of SPM(t) maps in individual subjects. This threshold was shown to be optimal to lateralize and localize metabolic abnormalities in previous epilepsy studies with FDG‐PET SPM analysis, both when using adult (Kumar et al., 2010) and pseudo‐normal pediatric control data (Archambaud et al., 2013).

To parcellate bilateral, homotopic regions of interest (ROIs) in each age‐specific FDG‐PET template, we utilized an SPM automated anatomical labeling atlas (AAL, http://www.gin.cnrs.fr/spip.php) consisting of 90 cerebral parcellations in standard template space that make homotopic ROIs available in whole cerebral cortex as well as subcortical structures such as thalamus and basal ganglia (Tzourio‐Mazoyer et al., 2002; Ashburner et al., 2013). The ROIs of this AAL template were spatially warped to each of three age‐specific FDG‐PET templates by using the symmetric diffeomorphic image normalization algorithm provided through the Advanced Normalization Tools (ANTs) (Avants & Gee, 2004). The resulting age‐specific AAL ROIs were applied to the spatially normalized FDG‐PET images of individual patients in order to evaluate SPM(t) scores in the voxels of individual ROIs.

2.5. Statistical analysis

For each ROI, SPM(t) scores of individual SWS children in both affected and unaffected hemispheres were averaged to quantify the degree of regional metabolic differences compared to the age‐matched pseudo‐normal controls. In addition to individual t‐scores, we also calculated mean SPM(t)scores for each region. We also performed two different correlation analyses: (a) to investigate the association between regional metabolic abnormalities and IQ measures, IQ was correlated with SPM(t) scores of individual ROIs; in a secondary analysis, we also correlated the IQ with SPM(t) score of individual ROIs (i.e., regressor: β) using a linear regression model, where a covariate of epilepsy duration [a known risk factor of poor IQ]) or age were also entered; (b) to investigate interhemispheric interactions, cross‐hemispheric correlation of glucose metabolism in homotopic regions of the two hemispheres was performed, where SPM(t) score of each ROI in the affected hemisphere was correlated with that of corresponding homotopic ROI in the unaffected hemisphere. For the correlation analyses, Pearson's correlation coefficients (R) were calculated, using SPSS 23.0 (IBM Corp., Armonk, NY). These Pearson's correlation coefficients were compared using Fisher's Z transformation method between R values of IQ and SPM (t) scores in right and left hemispheres. Finally, all p‐values from the ROI comparisons were additionally corrected for multiple comparisons at α = .05 (p _β) using the Benjamini–Hochberg procedure when multiple significant p‐values were reported (Benjamini & Hochberg, 1995).

3. RESULTS

3.1. Clinical data

Fourteen patients had left and 13 patients had right‐sided SWS brain involvement (Table 1). In the comparison of left‐ vs. right hemispheric subgroups, there was no significant difference in age (left/right: 5.6 ± 3.4/5.3 ± 2.4 years, p = .80), epilepsy duration (4.8 ± 3.6/3.7 ± 1.8 years, p = .36), age at seizure onset (1.2 ± 1.7/1.8 ± 1.9 years, p = .41), seizure frequency score (1.8 ± 0.7/1.9 ± 0.8, p = .63), FIQ (90 ± 18/91 ± 13, p = .86), VIQ (95 ± 16/98 ± 15, p = .72) or PIQ (86 ± 16/87 ± 14, p = .83). Interestingly, VIQ was substantially higher (i.e., ≥10 point difference) than PIQ in the majority of children both in the left and right hemispheric subgroups (10/14 [71%] in the left and 8/13 [61%] in the right hemispheric group). As expected, a strong negative correlation was found between duration of epilepsy and IQ (FIQ/VIQ/PIQ, R = −0.6/–0.6/–0.4, p = <.001/<.001/.04).

3.2. Regional FDG uptake abnormalities in the SWS group

Extent of hypometabolic cortex detected by SPM was slightly larger in the left hemispheric subgroup (Figure 1), and mean SPM(t) scores were lower in this subgroup as compared to right‐hemispheric SWS patients in 35 of the 45 automated anatomical labeling (AAL) regions (Supporting Information Table 1). However, mean SPM(t) scores did not differ between left affected and right affected groups (p > .45 in all comparisons). In the affected hemispheres, the whole‐group analysis found the lowest mean SPM(t) scores distributed in multiple lobes located in temporal, parietal and frontal regions, with the lowest values in a group of peri‐Sylvian regions (including the transverse temporal gyrus, superior temporal gyrus, insula, and rolandic operculum) (Supporting Information Table 1). In the unaffected hemisphere, average t‐scores were negative (consistent with slight decreases) in anterior regions and positive (suggesting slight increases in several patients) in some occipital, parietal, and temporal regions (Supporting Information Table 1).

Figure 1.

Spatial overlap of hypometabolic clusters (p‐value < .001, extent > 50 voxels) objectively identified by SPM in children with Sturge–Weber syndrome with left and right hemispheric involvement. The color bar indicates the number of patients showing hypometabolic abnormality at every vertex of the cortical template utilized for SPM analysis

3.3. Correlation of IQ with regional FDG uptake abnormalities (SPM(t) scores)

There were no significant correlations between FIQ, VIQ, or PIQ and SPM(t) scores of individual ROIs in the whole group or the right hemispheric group.

In contrast, FDG SPM(t) scores in the affected hemisphere showed strong positive correlations with IQ measures in the left hemispheric SWS group in widely distributed brain regions including frontal, parietal, temporal, and limbic regions in linear regression analysis (see Figure 2 and Table 2 for FIQ; PIQ and VIQ showed similar results, and detailed data are shown in Supporting Information Table 2); the strongest correlations were in the frontal lobe (medial orbito‐frontal, and orbital part of the superior frontal gyrus). The subsequent linear regression analysis with age as a covariate also showed significant positive correlations in a similar group of widely distributed ROIs (Supporting Information Table 3). When duration of epilepsy was entered as a covariate, correlations were again reproduced in similar regions, while correlation coefficients were lower (Supporting Information Table 4).

Figure 2.

AAL regions showing positive correlations with full‐scale IQ in the ipsilateral left hemisphere of the left affected subgroup of children with Sturge–Weber syndrome. (a) pars triangularis; (b) middle frontal, orbital part; (c) pars orbitalis; (d) rectus; (e) medial orbitofrontal; (f) post‐central; (g) angular; (h) supramarginal; (i) inferior parietal; (j) inferior temporal [Color figure can be viewed at http://wileyonlinelibrary.com]

Table 2.

Correlations between regional SPM(t) scores and FIQ in the left hemispheric SWS group (n = 14). Twenty‐nine regions with significant correlations are shown

		Affected side (Left)			Unaffected side (Right)
Region	AAL	R	p α‐value	p β‐value	R	p α‐value	p β‐value
Lateral frontal	Precentral	0.61	<.05*	.05	−0.18	.80	.76
	Superior frontal	0.60	<.05*	.05	−0.07	.90	.64
	Superior frontal ‐orbital part	0.79	.01*	.01*	0.12	.88	.75
	Middle frontal	0.61	<.05*	<.05*	−0.15	.85	.54
	Middle frontal ‐orbital part	0.65	<.05*	.02*	0.01	.98	.73
Inferior frontal	Pars opercularis	0.59	<.05*	<.05*	−0.29	.62	.44
	Pars triangularis	0.62	<.05*	.04*	−0.21	.77	.44
	Pars orbitalis	0.59	<.05*	.02*	−0.29	.90	.92
	Rolandic operculum	0.58	<.05*	.07	−0.32	.60	.37
Orbito‐frontal	Olfactory	0.59	<.05*	.27	0.36	.57	.91
	Medial orbitofrontal	0.87	<.001*	<.001*	0.35	.57	.84
	Rectus	0.79	.01*	.02*	0.23	.77	.82
Medial frontal	Anterior cingulate	0.69	<.05*	.08	0.55	.22	.32
	Mid cingulate	0.59	<.05*	.10	0.43	.43	.31
Limbic	Insula	0.59	<.05*	.09	0.20	.77	.63
	Parahippocampal	0.63	<.05*	<.05*	0.20	.77	.66
	Amygdala	0.57	<.05*	.11	0.31	.77	.35
Occipital	Middle occipital	0.56	.06	.10	−0.52	.12	.01*
	Fusiform	0.63	<.05*	.05	−0.12	.88	.80
Parietal	Postcentral	0.65	<.05*	.03*	−0.34	.57	.47
	Inferior parietal	0.60	<.05*	.05	−0.73	<.05*	.03*
	Supramarginal	0.68	<.05*	.04*	−0.72	<.05*	.01*
	Angular	0.70	<.05*	.03*	−0.77	<.05*	<.001*
Temporal	Transverse temporal	0.65	<.05*	.07	−0.35	.57	.22
	Superior temporal	0.62	<.05*	.06	−0.63	.12	.02*
	Superior temporal pole	0.64	<.05*	.05	−0.47	.41	.28
	Middle temporal	0.64	<.05*	.06	−0.66	.10	.04*
	Middle temporal pole	0.62	<.05*	.07	−0.38	.57	.31
	Inferior temporal	0.63	<.05*	<.05*	−0.60	.14	.18

R: Pearson's correlation coefficient. pα‐value: p‐value with no correction for multiple correaltions, p _β: p‐value corrected for multiple correaltions using Benjamini–Hochberg procedure; <.05:.045 ≤ p _α or p _β < .049, *p _α or p _β < .05.

Furthermore, SPM(t) scores showed an inverse linear correlation with FIQ in four parietal ROIs of the right (unaffected) hemisphere of the left‐hemispheric subgroup (p < .05 for inferior parietal/supramarginal gyrus/angular gyrus; Table 2, Figure 3), indicating that higher FDG uptake in these regions was significantly associated with lower IQ; a quadratic fit showed moderately improved R ² and p values particularly in the angular gyrus, indicating relatively preserved IQ in some patients with increased FDG uptake in this region (Figure 3). These inverse correlations with FIQ were also shown in additional ROIs (i.e., superior/middle occipital, p _a = .04/.04, respectively, Table 2). Similar inverse correlations were found in an additional analysis with age and epilepsy duration included as covariates (Supporting Information Tables 3 and 4).

Figure 3.

AAL regions showing negative correlations with IQ in the right hemisphere of the left affected group. In addition to linear regressions, quadratic fits are also shown demonstrating a non‐linear component, due to a relatively preserved IQ in some patients with increased metabolism in some regions of the unaffected hemisphere. (a) inferior parietal, (b) angular, (c) supramarginal, (d) superior temporal, (e) middle temporal

3.4. Interhemispheric correlation analysis

Table 3 presents the interhemispheric interaction of FDG SPM(t) scores across each pair of homotopic regions in the whole group of SWS children (n = 27) and age‐matched pseudo‐controls (n = 27). In the pseudo‐control group, bilateral homotopic regions showed a strong, positive interhemispheric correlation (p _β < .001 in all regions). In contrast, children with SWS showed regional differences in these correlations. Positive interhemispheric correlations of SPM(t) scores were found mostly in regions not directly affected by SWS pathology in most cases (including medial/orbito‐frontal cortex, insula, hippocampus/amygdala, and deep nuclei), although the correlation coefficients were lower than those in the control group (Table 3), suggesting preserved but diminished interhemispheric metabolic coupling in these regions. In contrast, several regions in the parietal and temporal lobes showed a significant inverse interhemispheric correlation (R values ranging between −0.45 and −0.82, Table 3), indicating that lower metabolism in these ipsilateral regions was associated with higher metabolism in the contralateral homotopic region.

Table 3.

Interhemispheric correlation of SPM(t) scores in children with unilateral SWS and pseudo‐controls. Only the 18 regions with significant correlations (p _β < 0.05) in the SWS group are listed. Negative interhemispheric correlations were found in the angular gyrus and several temporal lobe regions

Region	AAL	SWS (n = 27)		Pseudo‐controls (n = 27)
		R	p β‐value	R	p β‐value
Orbito‐frontal	Medial orbitofrontal	0.53	.02*	0.96	<.001*
	Rectus	0.52	.02*	0.96	<.001*
Medial frontal	Anterior cingulate gyrus	0.81	<.001*	0.98	<.001*
	Midcingulate area	0.77	<.001*	0.97	<.001*
	Posterior cingulate gyrus	0.64	<.001*	0.95	<.001*
Limbic	Insula	0.51	.02*	0.96	<.001*
	Hippocampus	0.52	.02*	0.94	<.001*
	Amygdala	0.77	<.001*	0.90	<.001*
Deep gray matter	Caudate nucleus	0.61	<.001*	0.99	<.001*
	Putamen	0.87	<.001*	0.97	<.001*
	Globus pallidus	0.82	<.001*	0.84	<.001*
	Thalamus	0.60	<.001*	0.99	<.001*
Parietal	Superior parietal lobule	−0.50	.02*	0.89	<.001*
	Inferior parietal lobule	−0.54	.01*	0.90	<.001*
	Angular gyrus	−0.74	<.001*	0.89	<.001*
Temporal	Transverse temporal gyrus	−0.45	.05	0.86	<.001*
	Superior temporal gyrus	−0.82	<.001*	0.99	<.001*
	Middle temporal gyrus	−0.72	<.001*	0.93	<.001*

R: Pearson's correlation coefficient. p _β: p‐value with correction by Benjamini–Hochberg procedure, *p _β < .05.

Correlation analyses in both left and right hemispheric SWS subgroups found significant positive interhemispheric correlations in five ROIs, typically not directly affected by SWS pathology, such as the anterior and mid cingulate gyrus, amygdala and putamen/globus pallidus (Supporting Information Table 5). However, only the left hemispheric group showed negative correlations, found in six homotopic ROIs localized in the parietal, temporal and occipital regions (inferior parietal lobule, supramarginal, angular, middle occipital, superior temporal, and middle temporal cortex, Supporting Information Table 5; Figure 4). Interestingly, these same six ROIs showed a significant negative correlations between SPM(t) score of right (unaffected) hemisphere and IQ as reported in Table 2. In the right hemispheric group, no significant negative interhemispheric correlation was found.

4. DISCUSSION

This study has several novel findings that provide some new insights in the effects of early unilateral brain lesions, associated with SWS, on neurocognitive development. First, the majority of our SWS patients showed markedly better verbal than nonverbal IQ, regardless of the side of the brain affected by the SWS lesion. Indeed, the left and right‐hemispheric subgroups showed similar mean verbal and non‐verbal IQs, and >70% of the left‐hemispheric subgroup showed a relative non‐verbal weakness, despite an apparently intact right hemisphere. This finding replicates previous studies comparing children with (non‐SWS related) unilateral lesions (Levine et al., 1987; St. James‐Roberts, 1981) and is consistent with the notion of a “competitive edge” of verbal over nonverbal (i.e., spatial) functions. Further, our results revealed that only the left‐hemispheric group showed inverse inter‐hemispheric regional metabolic correlations, which were present in the same posterior brain regions in the unaffected (non‐lesional) hemisphere that also correlated with IQ. Altogether, these findings strongly suggest that early left‐hemispheric SWS lesions may have a more robust functional effect on homotopic contralateral brain regions than right‐hemispheric lesions do; this effect may contribute to the relative preservation of speech function while interfering with (i.e., crowding) visuo‐spatial abilities in unilateral left‐hemispheric SWS.

Our previous studies in children with unilateral SWS provided preliminary imaging and neuropsychology evidence for contralateral hemispheric functional reorganization (Lee et al., 2001; Behen et al., 2011; Batista et al., 2007; Jeong et al., 2015); however, those studies did not address the effect of lesion lateralization and gave limited insights in the regional distribution of these effects. In one those studies, we reported imaging markers of contralateral functional and structural plasticity in the visual system in a small group of SWS children with unilateral occipital lobe damage using FDG‐PET and diffusion tensor imaging tractography (Batista et al., 2007; Jeong et al., 2015). Our current results demonstrate that the interhemispheric metabolic interaction (in the form of inverse correlations) extends beyond the occipital cortex to other posterior brain regions in the parietal and temporal lobe. In addition, we now demonstrate that metabolic activity in these contralateral posterior brain regions are also associated with IQ: while this correlation is inverse, there appears to be a subgroup where substantial contralateral regional metabolic increases are associated with relatively preserved cognitive functions. Intriguingly, these functional interactions were strictly confined to patients with left‐hemispheric SWS brain involvement, a novel finding that will need further studies.

The reason for the robust differences between the left and right hemispheric subgroups observed in our study is not clear, although there are some potential explanations. For example, some of these differences may be related to the known functional asymmetries and different maturational rates of the two hemispheres and associated developmental trajectories of various cognitive functions. While left hemispheric language dominance is typical in adults and school age children, infants and young children show less functional lateralization; this may explain the less lateralized cognitive dysfunction profile in children with congenital unilateral (left or right) lesions (Kolk et al., 2001). During normal brain development, the two cerebral hemispheres undergo differential region‐specific maturation (Corballis et al., 1978; Chiron et al., 1997; Saugstad, 1998; Casey et al., 2005). A seminal SPECT study in young children demonstrated an early, transient right‐hemispheric dominance of posterior (parieto‐temporal) associative brain regions (Chiron et al., 1997). This right > left functional asymmetry emerged during the first year of life, coinciding with development of visuo‐spatial abilities. This was then followed by evolution of a reverse, left > right asymmetry by age 3 years, coinciding with speech development. This reversal of functional asymmetry was specific for the posterior associative cortex (Brodmann's areas 39 [angular gyrus] and 40 [encompassing the supramarginal gyrus]), i.e. regions commonly affected in SWS (Figure 1) and showing the strongest inverse metabolic interaction in our study (Figure 4).

Figure 4.

AAL regions showing the most robust negative interhemispheric correlations in the left hemispheric SWS group. (a) superior temporal; (b) supramarginal; (c) inferior parietal; (d) angular gyrus

Differences in the capacity and patterns of functional reorganization between left and right hemispheres could be another possible attributing factor to the observed side differences in our study. It is well known that early and severe unilateral hemispheric injury can facilitate shifting function to the unaffected hemisphere (Liégeois et al., 2004; Behen et al., 2011; Adcock et al., 2003). However, the majority of the data demonstrate effective left‐to‐right language function shift (Liégeois et al., 2004; Strauss et al., 1990; Müller et al., 1998; Rosenberger et al., 2009), while less robust evidence exists for right‐to‐left interhemispheric reorganization of non‐verbal functions. For example, a functional MRI study of a child with a right hemispheric congenital lesion and right language dominance before surgery found only intrahemispheric reorganization after right frontal surgery (Seghier et al., 2001). Such findings, along with our data, may be consistent with speculations that the right (or nondominant) hemisphere, particularly early in life, may contain language circuits that exist in “an inactive or inhibited state” (Corballis & Morgan, 1978). Such cortical areas may exist as nonspecified or “silent areas”, which can be recruited for a given function associated with environmental demand or focal neural insult. The same authors also suggested that the right hemisphere may have more “silent areas” than the left hemisphere, and this may help to explain the apparent interhemispheric reorganization of language functions in children with left‐hemisphere lesion, along with the lack (or scarcity) of interhemisperic reorganization of visual spatial functions (to the left hemisphere).

With regard to “crowding”, Powell et al. (2012) defined the phenomenon as co‐existence of verbal and non‐verbal functions in the same hemisphere, competing for the same set of neuronal populations. Typically, the observed deficits involve reduced non‐verbal abilities following early left hemispheric injury, i.e., a cognitive pattern also seen with right hemispheric damage (Vargha‐Khadem et al., 1992). This is consistent with our results, where pronounced (≥10) VIQ > PIQ difference (suggesting right hemispheric crowding) was associated with 10 of 14 children with left hemispheric SWS lesions. There is less evidence for right‐to‐left interhemispheric reorganization that would involve a cognitive profile with impaired language abilities despite apparently intact dominant (left) hemisphere. Of note was that a robust PIQ > VIQ split was seen in one child with right hemispheric lesion (consistent with left hemispheric crowding); it is important to note, however, that in this one patient both VIQ and PIQ were measured in the impaired range, suggesting a lack of effective reorganization. One could speculate that the apparent lack of interhemispheric functional transfer in most of our right‐hemispheric patients may be also be due to the less severe injury (i.e., less metabolic involvement) in these children, as compared to their left‐hemispheric counterparts, thus allowing more room for intra‐hemispheric reorganization. However, the metabolic differences between the two (left and right) subgroups were modest and unlikely to fully explain the robust observed differences in neuro‐cognitive profiles.

In addition to side, timing, and extent of early unilateral lesions, additional factors such as severity of epilepsy can also affect neurocognitive outcome. In the study of Vargha‐Khadem et al. (1992), non‐epileptic children with left and right hemispheric lesions showed normal average verbal and nonverbal IQ, while epileptic children with similar lesions showed ≥10 point IQ deficit in both cognitive domains. In the present study, we have excluded a priori SWS patients with severe epilepsy in order to diminish the effect of seizures. However, estimation of seizure frequency is difficult and unreliable in SWS children because of subclinical seizures and a common seizure clustering pattern (Kossoff et al., 2009; Sugano et al., 2009). Nevertheless, long duration of epilepsy in SWS is strongly associated with poor outcome (Bosnyák et al., 2016). To take this effect into account, we have added epilepsy duration as a covariate in our metabolic correlation analysis to confirm that the observed effects are still present when this important clinical factor is also taken into account.

In addition to epilepsy duration and seizure frequency, antiepileptic medications and age of onset are all factors that can affect cognitive outcome. We tried to minimize these confounding factors by excluding patients with severe epilepsy. Seizure frequency and age of onset were not contributing factors in this cohort, while medication effect could not be included, because the drugs used were highly variable in type and numbers. In regard to the effect of the extent of ipsilateral SWS brain abnormality, the significant correlation of ipsilateral SPM{t} and FIQ in the left SWS group (reported in Table 2) was not confounded by the extent of PET abnormality in the affected hemisphere, as only a fraction (14%) of the obtained Pearson's correlation coefficients were reduced (and they all remained significant) when the extent of ipsilateral PET abnormality was added as a covariate (no detailed data shown).

This is the first study using an objective voxel‐by‐voxel analysis to assess regional metabolic analysis for whole‐brain in children with SWS. As such, there are some important methodological considerations. First, our patient group included children as young as 2.5 years of age, whose overall brain volume is lower than the volume of older children or adults. In order to address this issue, we have recently created age‐specific PET templates and pseudo‐normal control groups for SPM analyses and demonstrated the feasibility of this approach in analyzing PET scans of children with epilepsy above 1 year of age (Jeong et al., 2017). We demonstrated that SPM performed similarly in younger and older pediatric age groups (below and above 4 years of age, respectively) for detection of hypometabolic regions. Thus, by using the same approach in the present study, age differences are unlikely to have had a major impact on our analyses. Second, the affected hemisphere of children with SWS often shows some degree of brain atrophy that may distort normal brain anatomy in the affected brain region. This may lead to inaccuracies during the image normalization process to the age‐specific templates. Such atrophies are typically more pronounced in older children as the disease progresses; however, most of our subjects had no severe or extensive atrophic changes as demonstrated by MRI. In our previous pediatric PET study using SPM, 60% of the patients had an MRI lesion, and SPM performed similarly in lesional vs. non‐lesional subgroups (Kumar et al., 2010). To ensure that SPM results provide similar, reasonable detection of hypometabolism in the present study, we have inspected individual SPM(t) score maps to confirm that the detected areas are consistent with the visual assessment. Based on this, we were confident that minor distortions did not introduce a major error on the overall detection of regional hypometabolism in our SWS group. Nonetheless, registering labeled regions from standard atlas to native brain is obviously challenging with severe tissue loss in the affected hemisphere. We evaluated three nonlinear registration algorithms: FNIRT (FMRIB Software Library v5.0, https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSL), SyN (Advanced Normalization Tools, http://stnava.github.io/ANTs/), and SPM‐DARTEL. Indeed, these three methods derived consistent accuracy across subjects and label sets thus suggesting that the findings of this study might be generalizable to new subject populations that are evaluated using different labeling protocols.

In summary, these findings suggest that left‐hemispheric lesions in SWS often induce an early reorganization of verbal functions in affected children while “crowding” their non‐verbal cognitive abilities. This reorganized cognitive pattern is associated with metabolic abnormalities in specific posterior brain regions of the contralateral hemisphere not directly affected by SWS, resulting in the reversal of normal positive interhemispheric metabolic correlations.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article.

Supporting Information

Kim J‐A, Jeong J‐W, Behen ME, et al. Metabolic correlates of cognitive function in children with unilateral Sturge–Weber syndrome: Evidence for regional functional reorganization and crowding. Hum Brain Mapp. 2018;39:1596–1606. 10.1002/hbm.23937