Nontask-Related Brain Lateralization Biomarkers in Children: The Asymmetry of Language Areas on Functional Connectivity Functional Magnetic Resonance Imaging

Abstract

In this work, we will test the hypothesis that the connectivity of language areas in normal children is asymmetric between the hemispheres. Intrahemispheric region of interest (ROI)-to-ROI connectivity was assessed in 40 normal right-handed children. Asymmetries were assessed (1) between the hemispheres (global language connectivity); (2) between Brodmann areas (BAs) pairs (pairwise connectivity); and (3) between two homotopic BA (Global BA connectivity). Sixteen BAs were selected: 6, 7, 9, 19, 21, 22, 37, 38, 39, 40, 41, 42, 44, 45, 46, and 47. T scores for connectivity of each BA pair were ascertained using the MATLAB toolbox CONN. Lateralization index (LI) scores based on T-values were obtained. Only LIs with 2SD above the mean were considered as significant. Comparisons between T-value groups (per side and per BA) were performed utilizing double-sided T-tests. Null hypothesis was rejected for p < 0.05. There was not a statistical difference between global left and right connectivity strength (p = 0.40). There was significant pairwise connectivity asymmetry for the following pairs: BA7-BA44 (LI = 0.662); BA21-BA42 (LI = −0.616); BA21-BA40 (LI = −0.595); BA38-BA44 (LI = 0.470); BA39-BA44 (LI = −0.903); and BA42-BA47 (LI = −0.445). Language-related brain connectivity asymmetries have been demonstrated in a group of children and young adolescents. Two pairs related to Broca's area were left dominant (BA44-BA38 and BA44-BA7) and four pairs right dominant (BA42-BA47, BA39-BA44, BA21-BA40, and BA21-BA42).

Introduction

Clinical determination of language lateralization is of critical importance in the work-up of patients with pharmacologically refractory epilepsy who are candidates for epilepsy surgery, in which protocols differ for resections performed in the dominant or nondominant hemisphere.

Currently available techniques include intracarotid amobarbital test, electrocortical mapping, functional magnetic resonance imaging (fMRI), and magneto-encephalography. All of these methods, however, require cooperative patients who should perform cognitive tasks while holding still.

Resting-state-based functional connectivity (rsfc)-fMRI is a task-free procedure that provides an insight into the spontaneous brain oscillations. This technique may be utilized in noncooperative patients since spontaneous neural network oscillations are preserved, up to a significant extent, in patients under natural sleep (Manning et al., 2013) and light sedated patients (Greicius et al., 2008; Shimony et al., 2009). region of interest (ROI)-based rsfc-fMRI can quantify temporal correlations between specific brain regions, allowing the ascertainment of the asymmetry between local connectivity strength and that of contralateral homotopic areas.

A set of meta-analyses of fMRI coactivations on language tasks have recently revealed many ancillary language areas (Ardila et al., 2015, 2016a,b,c; Bernal et al., 2015; Rosselli et al., 2015). Functional connectivity of the canonical and ancillary language areas relative to their homotopic regions has not been deeply studied, and thus is not yet well understood. Therefore, investigation may provide a new tool for greater understanding of language networks.

In this investigation, we sought to demonstrate asymmetries in brain connectivity between language areas in a group of strongly right-handed normal subjects. Based on prior well-established knowledge (Benson and Ardilla, 1996; Whitaker, 2010), our hypotheses included expectations of (1) asymmetric global language connectivity between the hemispheres; (2) asymmetric pairwise connectivity among language-related areas; and (3) this pairwise connectivity should be stronger for left Brodmann areas (BAs) 44, 45, 21, and 22 (i.e., the canonical Broca and Wernicke language areas) than for homotopic regions in the right hemisphere.

Methods

Subjects

Data were selected from a publicly released rs-fMRI dataset known as the “ADHD-200,” which includes normative data. This dataset is part of the “1000 Functional Connectomes Project” available at http://fcon_1000.projects.nitrc.org/indi/adhd200. Of the data series from New York University (NYU), we selected 40 subjects satisfying the following criteria: (1) age 18 or below; (2) right handed; (3) minimum Verbal IQ of 85; (4) minimum Full-Scale IQ of 80; and (5) absence of neurological conditions. The NYU dataset has a large sample of typically developing English-speaking subjects who were recruited as controls for the Attention-Deficit Hyperactivity Disorder (ADHD) project, and their examination protocols are very similar to our own. All anatomical datasets were visually inspected by one of the authors (B.B.), and any cases presenting with signs of overt head motion or irregular/asymmetric skull shape were discarded and replaced with other subjects from the pool. Right handedness was evaluated using the Edinburgh Handedness Inventory (Oldfield, 1971), only subjects with scores above 0.30 were included. Verbal and Full-Scale IQ were assessed with the Wechsler Abbreviated Scales of Intelligence-Second Edition (Wechsler, 2014). As shown in Table 1, the dataset was comprised of 19 male and 21 female subjects, from 7 to 18 years of age.

Table 1.

Demographic Characteristics of the Sample

n = 40
Male = 19	Range	Mean	SD
Age	7–18	11.0	2.64
Handedness^a	0.31–1.0	0.68	0.21
Verbal IQ^b	85–141	111.3	13.13
Full-scale IQ^b	80–142	109.3	14.66

Edinburgh Handedness Inventory.

Wechsler Abbreviated Scale.

SD, standard deviation.

MRI technique

Magnetic resonance imaging (MRI) data consisted of anatomical volumes and blood oxygenation level dependent (BOLD)-sensitive echo-planar images obtained with a 3.0-Tesla Siemens Magnetom Allegra Syngo (Erlangen, Germany). Anatomical T1-weighted MRI 3D-volumes were obtained in the sagittal plane with a field of view (FOV) of 256 × 256 mm, 128 slices per slab, and slice thickness of 1.33 mm. Acquisition settings: TR: 2530 ms, TE: 3.25 ms, and flip angle of 7°. The echo-planar sequence sensitive to the BOLD effect was obtained with the following parameters: 180 timepoints (scan time 6:00 min), voxel size 3.0 × 3.0 × 4.0 mm, 33 axial interleaved slices with no gap, FOV 240 mm, slice thickness 4.0 mm, TR: 2000 ms, TE: 15 ms, 1 average, flip angle: 90°, standard shim mode.

Preprocessing

For each subject, the 180 timepoints of the rs-fMRI data were visually inspected in a dynamic presentation. Those with overt motion were discarded, since motion correction algorithms have demonstrated poor correction of gross movement. Preprocessing was performed utilizing SMP12 (www.fil.ion.ucl.ac.uk/spm) and MATLAB 8.6, R2015b. Data were realigned, normalized, and spatially smoothed using a Gaussian kernel of full width at half maximum of 8 mm. Anatomical volumes were segmented into gray and white matter, and cerebrospinal fluid (CSF) areas and the resulting masks eroded in one voxel dimension to minimize partial volume effects. Three temporal covariates were removed from the BOLD functional data using linear regression methods: (1) The temporal time series characterizing the estimated subject motion (three translations, three rotations, and six parameters representing their first-order temporal derivatives); (2) the BOLD effect of the white matter mask; and (3) the CSF pulsation. The resulting residual BOLD time series were then bandpass filtered (0.008 < f < 0.09). The rs-fMRI was first coregistered to the structural T1 anatomical image, and then to Montreal Neurological Institute 152 standard space.

Processing

First level

Intrahemispheric ROI-to-ROI connectivity between specific functional areas was assessed. (Note: Interhemispheric connectivities were not considered in this study.) Seeding sources (ROIs) were derived from the atlas of BAs provided by SPM12. Based on prior published meta-analyses of coactivations present in language tasks (Ardila et al., 2015, 2016a,b,c; Bernal et al., 2015; Rosselli et al., 2015), 32 BAs were selected—16 per hemisphere. These included the following: BA6, premotor; BA7, dorso-parietal; BA9, prefrontal; BA19, secondary visual; BA21 and BA22, Wernicke's area; BA37, fusiform gyrus; BA38, temporal pole; BA39, angular gyrus; BA40, supramarginal gyrus; BA41 and BA42, primary auditory area; BA44 and BA45, Broca's area; BA46, prefrontal; and BA47, pars orbitalis of the inferior frontal gyrus (IFG). With these areas, 120 possible intrahemispheric pairs may be considered. Functional connectivity was performed utilizing the MATLAB toolbox “CONN” (Whitfield-Gabrieli and Nieto-Castanon, 2012) version 15.g available at www.nitrc.org/projects/conn For each seed-to-target-ROI pair, statistical correlation was computed utilizing a weighted general linear model (ROI-to-ROI analyses). This provided a correlation/anticorrelation map between the average time series of the hemodynamic response function of the BA sources enumerated above, and the rest, as target ROIs. The generated automatic display was initially thresholded at R = 0.3 for visual inspection (Fig. 1).

FIG. 1.

First-level analysis (subject level). Exemplary image of left BA44 connectivity map in a subject. Clusters are color coded (from low to high) in warm (yellow to red) for positive correlations and cold colors (cyan to blue) for negative correlations. The background is a T1-weighted brain axial slice, oriented in neurological convention. Threshold has been set at R = 0.30. BA, Brodmann area.

Second level

For the second-level analysis, ROI-to-ROI connectivity maps for all subjects were pooled with the following settings: Between-subject contrast, 1 (Effect of All subjects, one group); between-conditions contrast, 1 (Effect of Rest); Between-sources contrast, 1 (Effect of each BA against all brain BA).

Within-subject effects of the between-sources contrasts correspond to multivariate/repeated-measures analyses of the selected effects modeled using a general linear model. The output is a within-subject linear combination of effects specified by the “between-conditions” and “between-sources” contrasts, applied to the first-level connectivity-measure matrix (the ROI-to-ROI analyses).

ROI-to-ROI connectivity was thresholded at p < 0.05 (two-sided) and corrected for multiple comparisons with the False Discovery Rate (FDR) technique. From the report provided by CONN, independent tables of intrahemispheric connectivity were obtained, selecting the language-related BAs previously referred to as targets. The following parameters were computed for each ROI pair: T-value, as well as uncorrected and corrected (FDR) p-values. Only pairs with p < 0.05 FDR corrected were accepted, which resulted in a minimum T-value of 2.03. Pairs were sorted by the T-values obtained from the left hemisphere. For each of these pairs, values for the right-hemisphere homotopic pairs were found and annotated. Post hoc analysis was performed in those cases in which the counterpart did not reach the threshold value.

Representations of ROI-to-full intrahemispheric connectivity were graphically produced as radial plots. In these graphs, the source ROI appears at one point of a ring, with target ROIs distributed around the rest of the circle. Lines are color coded to represent connectivity strength and correlation type (positive vs. negative), so that correlation (warm colors) versus anticorrelation (cold colors) may be readily identified.

Comparisons

BA pairwise connectivity strength values were obtained in T scores per each BA pair (PWCS) and per each hemisphere as shown in Table 2 (T-score columns). All T-values belonging to pairs related to a given BA were considered as a group and statistically compared with the values in the opposite hemisphere. This provides a global interhemispheric BA asymmetry score (GIBA). Likewise, all T-pairs belonging to a hemisphere were summed, averaged, and statistically compared with the opposite hemisphere to find an interhemispheric global connectivity asymmetry (IGCA). These comparisons are shown in Tables 3 and 4. GBA and IGC asymmetries were assessed as differences between left and right hemisphere, and tested for statistical significance utilizing two-tailed T-test, thresholded at p < 0.05. PWCS, lateralization index (PWCS-LI) based on connectivity T-statistics was computed using the following formula: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}{\rm LI}= \frac{{T_{left}}-{T_{right}}}{{T_{left}}+{T_{right}}} \tag{1} \end{align*} \end{document}

where positive values indicate left lateralization, whereas negative values mean rightward lateralization. Usually, values <−0.2 are considered rightward, values >0.2 leftward; values in-between are not considered to be significantly lateralized. For this study, we accepted as lateralized those values exceeding 2SD from the LI distribution of the sample. For the LI calculations, we utilized absolute values of T to focus on strength of connection. The anticorrelations indicated by these negative T values are related to negative BOLD and will be addressed in the discussion.

Table 2.

PWCS-Lateralization Index Values with Their Statistical Significance and Mean Comparison

		Left			Right
No.	BA pair	T score	p-uncorrected	p-FDR	T score	p-uncorrected	p-FDR	LI
1	6-9	16.79	0	0	19.1	0	0	−0.064
2	6-40	13.64	0	0	13.49	0	0	0.006
3	6-37	13.51	0	0	9.37	0	0	0.181
4	6-47	12.47	0	0	9.99	0	0	0.110
5	6-45	11.55	0	0	7.08	0	0	0.240
6	6-46	11.3	0	0	6.78	0	0	0.250
7	6-44	10.79	0	0	11.09	0	0	−0.014
8	6-22	9.06	0	0	9.84	0	0	−0.041
9	6-21	8.7	0	0	6.58	0	0	0.139
10	6-7	8.39	0	0	8.29	0	0	0.006
11	6-39	8.37	0	0	7.79	0	0	0.036
12	6-19	7.03	0	0	7.57	0	0	−0.037
13	6-38	6.45	0	0	4.62	0	0	0.165
14	6-41	5.01	0	0	7.68	0	0	−0.210
15	6-42	4.58	0	0	5.17	0	0	−0.061
16	7-19	14.14	0	0	11.73	0	0	0.093
17	7-37	12.24	0	0	19.9	0	0	−0.238
18	7-40	9.57	0	0	12.74	0	0	−0.142
19	7-39	7.08	0	0	6.45	0	0	0.047
20	7-46	6.15	0	0	6.99	0	0	−0.064
21	7-9	5.81	0	0	6.43	0	0	−0.051
22	7-21	3.32	0.002	0.0038	1.97	^a	^a	0.255
23	7-44	−2.31	0.026	0.0429	0.47	^a	^a	0.662
24	9-47	17.25	0	0	12.29	0	0	0.168
25	9-21	14.55	0	0	9.04	0	0	0.234
26	9-39	13.17	0	0	7.15	0	0	0.296
27	9-38	12.07	0	0	6.61	0	0	0.292
28	9-45	11.6	0	0	11.7	0	0	−0.004
29	9-46	9.11	0	0	11.36	0	0	−0.110
30	9-44	8.12	0	0	7.52	0	0	0.038
31	9-40	8.08	0	0	10.78	0	0	−0.143
32	9-37	7.82	0	0	6.44	0	0	0.097
33	9-22	6.62	0	0	9.75	0	0	−0.191
34	9-19	5.05	0	0	5.12	0	0	−0.007
35	9-41	1.64	^a	^a	2.05	0.047	0.0486	−0.111
36	19-37	14.17	0	0	13.7	0	0	0.017
37	19-39	5.78	0	0	6.51	0	0	−0.059
38	19-21	4.32	0.0001	0.0002	3.84	0.0004	0.0007	0.059
39	19-46	3.87	0.0004	0.0009	3.44	0.0014	0.0023	0.059
40	19-45	3.77	0.0005	0.0011	2.05	^a	^a	0.296
41	19-47	3.31	0.002	0.0039	4.12	0.0002	0.0003	−0.109
42	19-38	2.9	0.006	0.01	4.69	0	0.0001	−0.236
43	21-39	19.74	0	0	16.88	0	0	0.078
44	21-38	16.93	0	0	15.38	0	0	0.048
45	21-47	13.42	0	0	12.39	0	0	0.040
46	21-45	11.42	0	0	7.28	0	0	0.221
47	21-22	9.97	0	0	12.28	0	0	−0.104
48	21-41	3.83	0.0005	0.0006	3.34	0.0019	0.0026	0.068
49	21-37	3.82	0.0005	0.0006	4.18	0.0002	0.0003	−0.045
50	21-44	3.7	0.0007	0.0009	−3.49	0.0012	0.0018	0.029
51	21-42	0.95	^a	^a	4	0.0003	0.0005	−0.616
52	21-40	−0.53	^a	^a	2.09	0.0436	0.0466	−0.595
53	22-41	25.16	0	0	16.98	0	0	0.194
54	22-42	18.57	0	0	17.47	0	0	0.031
55	22-47	11.21	0	0	10.78	0	0	0.020
56	22-44	10.76	0	0	9.04	0	0	0.087
57	22-45	8.81	0	0	8.49	0	0	0.018
58	22-38	8.68	0	0	10.62	0	0	−0.101
59	22-40	6.58	0	0	8.55	0	0	−0.130
60	22-37	3.9	0.0004	0.0005	4.91	0	0	−0.115
61	22-46	3.61	0.0009	0.0011	4.92	0	0	−0.154
62	22-39	3.22	0.003	0.0031	5.24	0	0	−0.239
63	37-46	11.64	0	0	7.85	0	0	0.194
64	37-40	11.33	0	0	9.88	0	0	0.068
65	37-45	6.01	0	0	4.02	0.0003	0.0003	0.198
66	37-47	5.53	0	0	4.73	0	0	0.078
67	37-39	4.33	0.0001	0.0002	3.46	0.0013	0.0015	0.112
68	37-44	4.3	0.0001	0.0002	5.39	0	0	−0.112
69	37-42	3.21	0.003	0.003	2.91	0.006	0.0064	0.049
70	37-38	2.91	0.006	0.0062	3.1	0.0036	0.004	−0.032
71	37-41	2.01			4.89	0	0	−0.417
72	38-47	14.5	0	0	13.33	0	0	0.042
73	38-45	9.47	0	0	5.23	0	0	0.288
74	38-39	8.74	0	0	7.12	0	0	0.102
75	38-41	5.04	0	0	3.72	0.0006	0.001	0.151
76	38-44	4.16	0.0002	0.0003	1.5	^a	^a	0.470
77	38-42	2.94	0.006	0.0078	4.5	0.0001	0.0001	−0.210
78	39-47	7.33	0	0	5.64	0	0	0.130
79	39-45	6.42	0	0	2.69	0.0105	0.0142	0.409
80	39-42	−4.67	0	0.0001	−3.55	0.001	0.0015	0.136
81	39-44	−0.36	^a	^a	−7.06	0	0	−0.903
82	40-46	10.51	0	0	10.39	0	0	0.006
83	40-42	9.23	0	0	7.65	0	0	0.094
84	40-44	8.95	0	0	12.19	0	0	−0.153
85	40-47	5.82	0	0	5.34	0	0	0.043
86	40-41	4.78	0	0	6.36	0	0	−0.142
87	40-45	3.7	0.0007	0.0009	4.81	0	0	−0.130
88	41-42	20.73	0	0	18.19	0	0	0.065
89	41-44	5.4	0	0	5.9	0	0	−0.044
90	41-38	5.04	0	0	3.72	0.0006	0.0012	0.151
91	41-47	3.72	0.0006	0.0014	3.91	0.0004	0.0007	−0.025
92	41-45	3.01	0.005	0.0078	2.05	^a	^a	0.190
93	41-46	1.38	^a	^a	2.87	0.0066	0.0093	−0.351
94	42-44	7.52	0	0	7.58	0	0	−0.004
95	42-47	1.74	^a	^a	4.53	0.0001	0.0001	−0.445
96	42-45	1.86	^a	^a	2.64	0.0117	0.0151	−0.173
97	42-46	1.98	^a	^a	2.56	0.0146	0.0181	−0.128
98	44-45	14.02	0	0	11.92	0	0	0.081
99	44-47	11.04	0	0	8.51	0	0	0.129
100	44-46	7.88	0	0	8.11	0	0	−0.014
101	45-47	18.26	0	0	15.81	0	0	0.072
102	45-46	11.1	0	0	11.4	0	0	−0.013
103	46-47	5.85	0	0	6.53	0	0	−0.055
	Mean	7.85388			7.31			0.0048
	SD	5.21111			4.74505			0.2120
	Mean + 2SD	18.2761			16.8001			0.4288

Table rows are organized by BA and T-values. p-Values <1E-4 have been considered as 0. Numbers in bold are considered statistically significant (above mean + 2SD).

p-Values are below the cutting threshold.

BA, Brodmann area; FDR, False Discovery Rate; LI, lateralization index.

Table 3.

Mean Comparisons for Global Interhemispheric Brodmann Area Scores (T-Test Results)

	T score left		T score right
BA	Mean	SD	Mean	SD	p
6	9.84267	3.43971	8.96267	3.60971	0.500
7	7.15444	4.86214	8.33000	5.87414	0.650
9	12.30444	3.47278	10.61667	3.81881	0.252
19	6.82667	4.30753	6.45333	3.93905	0.931
21	11.43111	5.37235	9.66778	4.85430	0.572
22	12.09333	5.93023	11.69444	3.31968	0.919
37	9.62000	3.71967	8.81667	5.28098	0.935
38	9.56000	4.26817	7.57000	4.61581	0.521
39	7.99556	6.46616	6.29778	5.31634	0.490
40	9.30111	2.38794	10.11222	2.60974	0.483
41	8.52000	8.28796	7.57000	5.93906	0.897
42	5.22077	7.27795	5.49538	6.34063	0.919
44	7.25615	4.23118	6.36385	4.99686	0.521
45	8.65643	4.70927	6.92429	4.39604	0.324
46	7.03242	3.74642	6.93333	3.10136	0.944
47	9.38929	5.33773	8.42143	3.97943	0.591

Table 4.

Interhemispheric Global Connectivity Asymmetry

Left			Right			T-score-based LI
Sum (T)	Mean	SD	Sum (T)	Mean	SD	Sum LI	Mean	SD
808.950	7.854	5.211	752.930	7.310	4.745	0.49545	0.00481	0.21201
p-Value difference of means (2s–T-test)
Left vs. Right
p = 0.4048
95% CI: −0.7843 < 0.58369 < 1.9516

CI, confidence interval.

Statistical methods

At the subject level, pairwise connectivity statistics were provided by CONN, as previously explained. These were generated from contrasts between the source ROI, and each of the target ROIs represented by the areas of Brodmann. The lateralization indices (LIs) obtained from all comparisons were averaged, and the group standard deviation (SD) was found. Those PWCS-LIs exceeding 2SD above the mean, each side considered apart, were accepted as statistically significant. No directionality in the connectivity was considered, thus pair BA6-BA9 is the same as BA9-BA6. All group comparisons were performed utilizing GNU-PSPP 0.7.9, 2012 (available at www.gnu.org/software/pspp/pspp.html).

Results

From the 120 possible BA pairs, 103 had at least one side with T-values above the threshold. The summary of results is displayed in tabulated values of PWCS which include connectivity strength in terms of T-values, their respective corrected p-values (FDR) and their LIs (Table 2). T-test comparisons for GIBA and IGCA are presented in Tables 3 and 4, respectively.

There was not a statistical difference between left and right IGCA means (mean left = 7.85; mean right = 7.31; p = 0.40; 95% CI: −0.784 < −0.5837 < 1.9516). Neither did IGCA-LI show significant asymmetry (LI = 0.004) (Table 4).

No statistical significant asymmetry was found in any of the GIBA group comparisons. T means differences in p-values ranged from 0.252 for BA9 to 0.944 for BA46 (Table 3).

The greatest statistical significant findings were noted on pairwise comparisons. The following pairs showed connectivity strength exceeding 2SD above the mean (left: mean + 2SD = 18.276; right: mean + 2SD = 16.80): left and right BA21-BA39 (left: T = 19.74; right: T = 16.88); left and right BA22-BA41 (left: T = 25.76; right: T = 16.98), left and right BA22-BA42 (left: T = 18.57; right: T = 17.47), left and right BA41-BA42 (left: T = 20.73; right: T = 18.19); and right BA6-BA9 (T = 19.1). We noted that higher connectivity values do not guarantee significant asymmetry, as both sides may have similar values. Pairwise LI values ranged from −0.90 (BA39-BA44) to 0.662 (BA44-BA7). The mean pairwise LI was 0.0048 with SD = 0.21 (mean +2SD = ± 0.428). Six BA pairs had LIs beyond 2SD from the mean: BA7-BA44 (0.662), BA21-BA40 (−0.595), BA21-BA42 (−0.616), BA39-BA44 (−0.903), BA42-BA47 (−0.445), and BA44-BA38 (0.47). Notice that only three BAs suffice to explain all the asymmetries: BA44, BA21, and BA42, as there is no directionality differentiation (Fig. 2). The following pairs had negative T-values (i.e., anticorrelation): left BA21-BA40, left and right BA39-BA44, and right BA21-BA44. Although no significant asymmetries were found for BA45 and BA22, BA45-BA39 was near the cutoff (LI = 0.409). The maximum LI for area BA22 was just slightly right lateralized (LI = −0.239).

FIG. 2.

(A) BA44 connectivity represented in a radial graph. The seeding area is located at the root of the bundle. Each hemisphere is represented independently. Intrahemispheric target BAs are located around the circle. Connectivity is represented utilizing the same code mentioned in Figure 1. Main asymmetries are explained by lack (underthreshold) of connectivity: 44 to 39 in the left side, and 44 to 38 and 7 in the right side. (B) BA21 connectivity represented in a radial graph. Same coding as mentioned in previous figure. Notice the asymmetry of 21L to 42L in the left panel, while 21R to 42R is shown in green on the right side. (C) BA42 connectivity represented in a radial graph. A clear asymmetry is demonstrated in the connectivity of BA42 to BA47, present in the right but absent on the left. Some other interhemispheric discrepancies (e.g., BA42-BA45) that are not reflected as significant in Table 2 are due to the fact that the absent connectivity pair had subthreshold strength values.

Figure 3 illustrates the PWCS pairs with significant LI.

FIG. 3.

Graphic representation of most significant lateralization indices from PWCS analysis. Of note is that not only pairs BA44-BA39 and BA21-BA40 had negative LI values (right-sided dominance) but also the T-values were negative. LI, lateralization index.

Discussion

In summarizing the main results, our first hypothesis was rejected because global intrahemispheric connectivity unexpectedly appears to be quite symmetric. Our second hypothesis was accepted as true, with findings of significant connectivity asymmetry for BA44: (pairs 44-7 [left dominant], 44-38 [left dominant], and 44-39 [right dominant]); BA21: (pairs 21-42 [right dominant] and 21-42 [right dominant]); and pairs 42-47 (right dominant). Several BA pairs show strong negative correlation: left BA21-BA40, left BA39-BA44, right BA21-BA44, and right BA39-BA44; and no significant asymmetries were found for BA45 and BA22.

Some of our results are unexpected, as they have not been previously reported as far as we could find. It is difficult to understand (1) the lack of connectivity asymmetry of all pairs related to BA22 and BA45; (2) the striking right dominance of the BA44-BA39 connectivity; and (3) the negative correlation of BA39 and BA21 with other canonical language areas (suggesting inhibitory input).

Some pioneering studies have utilized rs-fMRI to assess the connectivity between language areas. Interhemispheric differences between the connectivity of the IFG, the left frontal middle gyrus, and the anterior cingulate gyrus (pre-SMA) have been investigated (Lou et al., 2017; Waites et al., 2006). The first group only contrasted normal controls versus adult patients with temporal lobe epilepsy (TLE); while the second sought to demonstrate connectivity differences between left and right pre-SMA to the ipsilateral frontal operculum in a group of 30 normal adult subjects. Language networks have also been obtained from independent component analysis (ICA), another resting-state postprocessing connectivity method, in a group of 23 epilepsy patients (age range 10–59 years) (DeSalvo et al., 2016). It is worthy of note, however, that language network appears in only 57.5% of cases in a normal pediatric population, utilizing ICA (see: https://sites.google.com/site/independentcomponentatlas/home/language). Since this group consists of subjects performing rs-fMR during wakefulness, we presume that the spontaneous language network will show up in a significantly lower number in sedated subjects.

Closer to our approach, Doucet and colleagues (2015) studied 55 TLE patients and 23 healthy controls who underwent rs-fMRI. In this work, connectivity was sought to correlate LI values previously obtained from an expressive language task (verb generation) to functional connectivity T-values between the pars orbitalis and opercularis, and different areas of the brain. Interestingly, within the group of normal controls a positive correlation with the LI was found involving connectivity with the right superior frontal cortex and the left SMA. Our methodology differs primarily by correlating the signal time course from function-related ROIs rather than anatomical regions, as well as inclusion of analyses of ancillary language areas. Despite differences in methodology and scope, our findings confirm the asymmetries of BA44 (pars opercularis) and BA47 (pars orbitalis) previously found by Doucet's group.

The present findings are also concordant with a similar study recently published by Joliot et al. (2016) that utilized rs-fMRI to create complex correlation matrices for each hemisphere based on the Atlas of Intrinsic Connectivity of 192 Homotopic Areas, as described previously by the same group (Joliot et al., 2015). Their approach also utilized pairwise connectivity, while evaluating intrahemispheric intrinsic connectivity asymmetry. However, our methods differ substantially. Instead of ascertaining LIs or contrasting left and right mean scalars, Joliet's group assessed hemispheric asymmetries by categorizing a matrix of the differences “between the left and right intra-hemispheric matrices of intrinsic correlation computed for each pair” as either typical or atypical, and correlating them with language LIs obtained from standard fMRI.

Although we found greater asymmetry in canonical language areas, this was not universally true across all of them. The leftmost asymmetry was found for pairs BA44-BA7 (LI = 0.662) and BA44-BA38 (LI = 0.47). Results emphasize the importance of left BA44 (Broca's putative area). It is striking that the most significant BA44 pairs have been formed with language ancillary areas and not with the canonical pairs of the receptive areas (BA21 and BA22). The involvement of left BA7 in language has been previously established: It has been found activating in attention to phonological relations (McDermott et al., 2003), imageability in word comprehension tasks (Bedny and Thompson-Schill, 2006), literal sentence comprehension (Shibata et al., 2007), and phonology and semantic processing (Seghier et al., 2004). Left BA7 is more frequently reported as having involvement in verbal working memory tasks (Shivde and Thompson-Schill, 2004; Tsukiura et al., 2001); hence, the high connectivity noted across all mentioned studies may be explained by verbal working memory demands. Moreover, a recent study published by Bernal et al. (2015), utilizing meta-analytic connectivity model, found BA7 as the third main cluster connecting to the left BA44 (the seeding ROI). The two other clusters were BA6, 13, 9, 46, 47 (first cluster) and areas of the pre-SMA (second cluster).

One unreported finding is the asymmetry of the pair BA44-BA38, although the role of left temporal pole (BA38) in language has been reported in numerous studies (Bottini et al., 1994; Giraud et al., 2004; Vandenberghe et al., 2002). Despite this, significant language impairment is seldom seen after anterior temporal lobe resection in epilepsy patients. We speculate that this could be the result of a faulty neuropsychological battery with insufficient sensitivity, as a number of recent publications have shown the importance of the left temporal pole in retrieving category-specific names (Belfi and Tranel, 2014; Mehta et al., 2016; Tranel, 2009).

The right side lateralization of BA21-related pairs was another unexpected finding. Left BA21 is the core of the receptive language area, along with BA22. Numerous studies have shown left BA21 activation in semantic processing (Chou et al., 2006; Duzel et al., 2001; McDermott et al., 2003), as well as word and sentence production (Brown et al., 2006; Friedman et al., 1998). However, language usually (at least oral declarative language) conveys prosody, and right BA21 has been implicated in prosody analysis (Ethofer et al., 2006; Hesling et al., 2005). Current understanding of how the cortex processes prosody is far less sophisticated than for declarative language; while the latter is more clinically and functionally important, results of this study suggest that stronger connectivity of BA21 with areas of auditory input and analysis (BA42 and BA40) likely involved with prosodic aspects of language on the right side. The same appears to happen with pair BA42-BA47, also with right side dominance. To our knowledge, no studies have focused on BA42 because this area is always included with BA41 as a primary auditory area. An ad hoc PubMed search (March 2, 2017) using key words “Prosody Brodmann area 42” yielded zero results.

It remains unclear why significant asymmetry of any BA45 pairs was not demonstrated. The absence of significant asymmetrical connectivity of BA45 with the remainder of each hemisphere or with other BA pairs may merely represent a consequence of the high statistical threshold. In fact, BA45-BA39 shows a LI of 0.409, which is left lateralized. However, this number is below our cutoff value, and was therefore discarded. We want to emphasize the effect of the stringent values we chose, and it should be noted that in the neurofunctional radiology literature a LI greater than 0.2 is an acceptable score to define any activation as left lateralized (Guillen et al., 2009; Ruff et al., 2008). The involvement of BA45 in language processing has long been established, producing activation across a wide range of fMRI language tasks (i.e., word generation, phonological and semantic processes, metaphoric and grammatical processing, lexical search, semantic memory retrieval, etc.; see www.fmriconsulting.com/brodmann/BA45.html for review).

Another completely unexpected finding was the lack of asymmetry between BA22 pairs, particularly since this is the core part of Wernicke's area. It has been demonstrated that a consistent structural connectivity asymmetry exists between receptive language areas (which includes BA22) and frontal expressive areas through the arcuate fasciculus (Matsumoto et al., 2008; Nucifora et al., 2005). Our unanticipated finding seems to be indicative of equally strong involvement of the right hemisphere beyond that of verbal-linguistic functions. Areas for melodic and prosodic processing of verbal communication within the right hemisphere have been previously suggested (Lindell, 2006), as well as for comprehension of reading and language with higher complexity (Xu et al., 2005). Within the present data, the most salient finding with regard to BA22 was strong bilateral connectivity with the ipsilateral BA42 (primary auditory area). While this strengthens support for the role of the posterior most superior temporal gyrus in receptive language processing, because of clearly asymmetrical propositional vs. prosodic language functions within the cortex, it remains curious why these associations are so comparable across hemispheres.

The rightward asymmetry between BA44 and BA39 deserves special attention. Although the LIs are based on strength of activation given by positive T-values, they may have a negative value also, which is expressed mathematically as an anticorrelation between the pair. Anticorrelation, in this context, refers to negative BOLD, occurring relative to the signal in the seed ROI. Experimental findings have found sustained negative BOLD associated with decreased neural response due to inhibitory input (Shmuel et al., 2002). We favor this explanation to blood steal as an alternative explanation since the anticorrelated pairs are located in remote brain areas.

In this study, BA44 was negatively correlated with BA39 in both hemispheres but at quite different strengths (Tleft = −0.36; Tright = −7.06) suggesting an inhibitory input from one area to the other in both hemispheres. We acknowledge that the technique used does not allow discrimination of directionality in the connectivity. However, it does seem reasonable that this BA pair follows the principles of general brain organization proposed by Luria (1972), in which information derived from afferent input is being relayed from posterior receptive areas to anterior regions responsible for execution of behavior. If this were the case, the right BA44 would be more strongly inhibited by BA39 (the higher negative correlation) than its left counterpart, leaving the left BA44 with some advantage as it has less inhibitory input. The inhibitory effect of BA39 seems to make sense as it has been established that BA39 (the angular gyrus/inferior parietal lobule) is part of the default mode network, which is known to have a negative correlation with task-related (i.e., salience and executive) networks. However, the interplay between these two areas is not well understood as other studies have found positive input from BA39 to BA44. For instance, Hampson and colleagues (2006) found a positive correlation between left Broca's area and the ipsilateral angular gyrus in reading skills. It is important to note that both BA44 and BA39 on the right side have been less studied and therefore less understood than their left homotopic counterparts. In sum, the significance of the functional relationship between BA44 and BA39 and its role in cognition warrants further investigation.

Limitations and Future Directions

The main limitation of this study is the lack of a gold standard for language lateralization in the sample group. However, this limitation is likely to have a limited impact upon interpretation of the results since language lateralization in strongly right-handed normal subjects has been found in 96% of cases (Knecht et al., 2000). We feel that our effort to select only strong right-handed subjects should add confidence that, as a group, our sample has a strong left hemisphere language lateralization.

The age range of this sample could be considered either a limitation or strength depending on the perspective. On one hand, these findings may not fully apply to adults, as the relationships reported may represent an incomplete maturational stage. On the other, imaging data for children are far less common; therefore, this study has a greater potential upside to understanding pediatric brain networks. The natural developmental trajectory of cortical connectivity was represented in the work of Xiao and colleagues (2016), who tracked longitudinal changes in resting-state connectivity from 5 to 6 years old. Despite looking at only a narrow developmental snapshot, progressive changes in connectivity across the whole brain were correlated with increasing age and language improvements. Higher level cognitive and reasoning skills are also associated with maturation of functional networks, particularly between the left rostrolateral prefrontal cortex (BA10) and angular (BA39) and supramarginal gyri (BA40) (Uddin et al., 2011; Wendelken et al., 2016).

The use of absolute values for the LI to maintain the scalar as a ratio between −1 and 1 may obscure more significant differences. For instance, a BA pair having T-values −7.0 on the left and 7.0 on the right will have the same LI than other BA pair having T-values 7.0 left and 7.0 right. Although the connectivity strength is the same, the functional effect may be completely different, as negative values suggest inhibitory input. Ad hoc analysis of plain T differences subtracting right T-value from left T-value in the PWCS analysis showed five values 2SD over the mean. Four of them with leftward advantage: BA9-BA39: 6.02; BA21-BA44: 7.19; BA22-BA41: 8.18; and BA39-BA44: 6.7. BA7-BA37 had a rightward advantage: −7.66. Of note is the relevance of the canonical language areas obtained with this simple approach. Among these pairs, only BA39-BA44 showed up in our analysis based on connectivity strength.

The current findings are presented at an aggregated group level; although all participants were neurologically normal, individual variability may lead to conclusions that are not applicable at the individual level. Therefore, these results should be interpreted with caution, as they require replication and additional studies before it could be utilized in the clinical arena where normative patterns may not apply. For example, atypical resting-state connectivity has been demonstrated in epilepsy, with longer disease duration correlating with greater functional changes (Wang et al., 2011). In left TLE, resting-state connectivity is altered relative to controls, whereby dominant hemisphere language modules demonstrate diminished connectivity (Waites et al., 2006).

Forthcoming research should also include comparison of normal controls against patients with neurological conditions known to have atypical connectivity patterns, such as those with epilepsy. Further development of our technique and findings might have a very practical usage allowing language mapping in patients on natural sleep or under light sedation not only in very young children but also for those otherwise unable to stay still during an awake scanning process. While knowing which areas are connected is important, only through an understanding of how they interact physiologically will this science take on true clinical relevance and ability to inform medical decision making.

In conclusion, we have shown connectivity asymmetries that appear to be related to language lateralization. The main asymmetries were related to Brodmann pairs BA42-BA47, BA39-BA44, BA21-BA40, and BA21-BA42 with right-side dominance, and BA44-BA38 and BA44-BA7 with left-side dominance. While further technical refinement is needed, task-free rs-fMRI procedures could become an alternative method of probing language lateralization in noncooperative patients requiring resective brain surgery.

Footnotes

Author Disclosure Statement

Dr. Byron Bernal is president and owner of fMRI Consulting. All other authors have no competing financial interests.

References

Ardila

, Bernal

, Rosselli

. 2015. Language and visual perception associations: meta-analytic connectivity modeling of Brodmann area 37. Behav Neurol, 2015:565871.

Ardila

, Bernal

, Rosselli

. 2016a. Connectivity of BA46 involvement in the executive control of language. Psicothema, 28:26–31.

Ardila

, Bernal

, Rosselli

. 2016b. How extended is Wernicke's area? Meta-analytic connectivity study of BA20 and integrative proposal. Neurosci J, 2016:4962562.

Ardila

, Bernal

, Rosselli

. 2016c. How localized are language brain areas? A review of Brodmann areas involvement in oral language. Arch Clin Neuropsychol, 31:112–122.

Bedny

, Thompson-Schill

. 2006. Neuroanatomically separable effects of imageability and grammatical class during single-word comprehension. Brain Lang, 98:127–139.

Belfi

, Tranel

. 2014. Impaired naming of famous musical melodies is associated with left temporal polar damage. Neuropsychology, 28:429–435.

Benson

, Ardilla

. 1996. Aphasia: A Clinical Perspective. New York, NY: Oxford University Press.

Bernal

, Ardila

, Rosselli

. 2015. Broca's area network in language function: a pooling-data connectivity study. Front Psychol, 6:687.

Bottini

, Corcoran

, Sterzi

, Paulesu

, Schenone

, Scarpa

, et al. 1994. The role of the right hemisphere in the interpretation of figurative aspects of language. A positron emission tomography activation study. Brain, 117(Pt 6):1241–1253.

10.

Brown

, Martinez

, Parsons

. 2006. Music and language side by side in the brain: a PET study of the generation of melodies and sentences. Eur J Neurosci, 23:2791–2803.

11.

Chou

, Booth

, Bitan

, Burman

, Bigio

, Cone

, et al. 2006. Developmental and skill effects on the neural correlates of semantic processing to visually presented words. Hum Brain Mapp, 27:915–924.

12.

DeSalvo

, Tanaka

, Douw

, Leveroni

, Buchbinder

, Greve

, Stufflebeam

. 2016. Resting-state functional MR imaging for determining language laterality in intractable epilepsy. Radiology, 281:264–269.

13.

Doucet

, Pustina

, Skidmore

, Sharan

, Sperling

, Tracy

. 2015. Resting-state functional connectivity predicts the strength of hemispheric lateralization for language processing in temporal lobe epilepsy and normals. Hum Brain Mapp, 36:288–303.

14.

Duzel

, Picton

, Cabeza

, Yonelinas

, Scheich

, Heinze

, Tulving

. 2001. Comparative electrophysiological and hemodynamic measures of neural activation during memory-retrieval. Hum Brain Mapp, 13:104–123.

15.

Ethofer

, Anders

, Erb

, Herbert

, Wiethoff

, Kissler

, et al. 2006. Cerebral pathways in processing of affective prosody: a dynamic causal modeling study. Neuroimage, 30:580–587.

16.

Friedman

, Kenny

, Wise

, Wu

, Stuve

, Miller

, et al. 1998. Brain activation during silent word generation evaluated with functional MRI. Brain Lang, 64:231–256.

17.

Giraud

, Kell

, Thierfelder

, Sterzer

, Russ

, Preibisch

, Kleinschmidt

. 2004. Contributions of sensory input, auditory search and verbal comprehension to cortical activity during speech processing. Cereb Cortex, 14:247–255.

18.

Greicius

, Kiviniemi

, Tervonen

, Vainionpaa

, Alahuhta

, Reiss

, Menon

. 2008. Persistent default-mode network connectivity during light sedation. Hum Brain Mapp, 29:839–847.

19.

Guillen

, Adjouadi

, Ayala

, You

, Barreto

, Bernal

, Gaillard

. Toward fMRI Group Identification Based on Brain Lateralization. In On Advanced Information Networking and Applications (AINA-09). Presented at the IEEE 23rd International Conference, Bradford, UK, 2009, p. 1025.

20.

Hampson

, Tokoglu

, Sun

, Schafer

, Skudlarski

, Gore

, Constable

. 2006. Connectivity-behavior analysis reveals that functional connectivity between left BA39 and Broca's area varies with reading ability. Neuroimage, 31:513–519.

21.

Hesling

, Clement

, Bordessoules

, Allard

. 2005. Cerebral mechanisms of prosodic integration: evidence from connected speech. Neuroimage, 24:937–947.

22.

Joliot

, Jobard

, Naveau

, Delcroix

, Petit

, Zago

, et al. 2015. AICHA: an atlas of intrinsic connectivity of homotopic areas. J Neurosci Methods, 254:46–59.

23.

Joliot

, Tzourio-Mazoyer

, Mazoyer

. 2016. Intra-hemispheric intrinsic connectivity asymmetry and its relationships with handedness and language Lateralization. Neuropsychologia, 93:437–447.

24.

Knecht

, Drager

, Deppe

, Bobe

, Lohmann

, Floel

, et al. 2000. Handedness and hemispheric language dominance in healthy humans. Brain, 123(Pt 12):2512–2518.

25.

Lindell

. 2006. In your right mind: right hemisphere contributions to language processing and production. Neuropsychol Rev, 16:131–148.

26.

Lou

, Peck

, Brennan

, Mallela

, Holodny

. 2017. Left-lateralization of resting state functional connectivity between the presupplementary motor area and primary language areas. Neuroreport, 28:545–550.

27.

Luria

. 1972. The Working Brain: Introduction to Neuropsychology. Harmondsworth, UK: Penguin Books Ltd.

28.

Manning

, Courchesne

, Fox

. 2013. Intrinsic connectivity network mapping in young children during natural sleep. Neuroimage, 83:288–293.

29.

Matsumoto

, Okada

, Mikuni

, Mitsueda-Ono

, Taki

, Sawamoto

, et al. 2008. Hemispheric asymmetry of the arcuate fasciculus: a preliminary diffusion tensor tractography study in patients with unilateral language dominance defined by Wada test. J Neurol, 255:1703–1711.

30.

McDermott

, Petersen

, Watson

, Ojemann

. 2003. A procedure for identifying regions preferentially activated by attention to semantic and phonological relations using functional magnetic resonance imaging. Neuropsychologia, 41:293–303.

31.

Mehta

, Inoue

, Rudrauf

, Damasio

, Tranel

, Grabowski

. 2016. Segregation of anterior temporal regions critical for retrieving names of unique and non-unique entities reflects underlying long-range connectivity. Cortex, 75:1–19.

32.

Nucifora

, Verma

, Melhem

, Gur

. 2005. Leftward asymmetry in relative fiber density of the arcuate fasciculus. Neuroreport, 16:791–794.

33.

Oldfield

. 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia, 9:97–113.

34.

Rosselli

, Ardila

, Bernal

. 2015. Angular gyrus connectivity model for language: a functional neuroimaging meta-analysis [in Spanish]. Rev Neurol, 60:495–503.

35.

Ruff

, Petrovich Brennan

, Peck

, Hou

, Tabar

, Brennan

, Holodny

. 2008. Assessment of the language laterality index in patients with brain tumor using functional MR imaging: effects of thresholding, task selection, and prior surgery. Am J Neuroradiol, 29:528–535.

36.

Seghier

, Lazeyras

, Pegna

, Annoni

, Zimine

, Mayer

, et al. 2004. Variability of fMRI activation during a phonological and semantic language task in healthy subjects. Hum Brain Mapp, 23:140–155.

37.

Shibata

, Abe

, Terao

, Miyamoto

. 2007. Neural mechanisms involved in the comprehension of metaphoric and literal sentences: an fMRI study. Brain Res, 1166:92–102.

38.

Shimony

, Zhang

, Johnston

, Fox

, Roy

, Leuthardt

. 2009. Resting-state spontaneous fluctuations in brain activity: a new paradigm for presurgical planning using fMRI. Acad Radiol, 16:578–583.

39.

Shivde

, Thompson-Schill

. 2004. Dissociating semantic and phonological maintenance using fMRI. Cogn Affect Behav Neurosci, 4:10–19.

40.

Shmuel

, Yacoub

, Pfeuffer

, Van de Moortele

, Adriany

, Hu

, Ugurbil

. 2002. Sustained negative BOLD, blood flow and oxygen consumption response and its coupling to the positive response in the human brain. Neuron, 36:1195–1210.

41.

Tranel

. 2009. The left temporal pole is important for retrieving words for unique concrete entities. Aphasiology, 23:867.

42.

Tsukiura

, Fujii

, Takahashi

, Xiao

, Inase

, Iijima

, et al. 2001. Neuroanatomical discrimination between manipulating and maintaining processes involved in verbal working memory; a functional MRI study. Brain Res Cogn Brain Res, 11:13–21.

43.

Uddin

, Supekar

, Ryali

, Menon

. 2011. Dynamic reconfiguration of structural and functional connectivity across core neurocognitive brain networks with development. J Neurosci, 31:18578–18589.

44.

Vandenberghe

, Nobre

, Price

. 2002. The response of left temporal cortex to sentences. J Cogn Neurosci, 14:550–560.

45.

Waites

, Briellmann

, Saling

, Abbott

, Jackson

. 2006. Functional connectivity networks are disrupted in left temporal lobe epilepsy. Ann Neurol, 59:335–343.

46.

Wang

, Lu

, Zhang

, Zhong

, Jiao

, Zhang

, et al. 2011. Altered resting state networks in epileptic patients with generalized tonic-clonic seizures. Brain Res, 1374:134–141.

47.

Wechsler

. 2014. Wechsler Adult Intelligence Scale (WAIS–IV), 4th ed. New York, NY.

48.

Wendelken

, Ferrer

, Whitaker

, Bunge

. 2016. Fronto-parietal network reconfiguration supports the development of reasoning ability. Cereb Cortex, 26:2178–2190.

49.

Whitaker

. 2010. Concise Encyclopedia of Brain and Language. Oxford, UK: Elsevier.

50.

Whitfield-Gabrieli

, Nieto-Castanon

. 2012. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connect, 2:125–141.

51.

Xiao

, Friederici

, Margulies

, Brauer

. 2016. Development of a selective left-hemispheric fronto-temporal network for processing syntactic complexity in language comprehension. Neuropsychologia, 83:274–282.

52.

, Kemeny

, Park

, Frattali

, Braun

. 2005. Language in context: emergent features of word, sentence, and narrative comprehension. Neuroimage, 25:1002–1015.