2.1.

Participants

Sixty participants were recruited for this study, which comprised 30 children with ADHD (boys, 7 to 12 years) and 30 sex-, age-, and education-matched HCs. The children with ADHD were recruited from Peking University Sixth Hospital, Beijing, China, and HCs were enrolled from a primary school in the local community. For the children with ADHD, the inclusion criteria were as follows: (1) a full-scale intelligence quotient $(\mathrm{IQ})\ge 80$; (2) right-handed; and (3) drug-naïve and free of other medical intervention. In addition, children with a diagnosis or history of head trauma with loss of consciousness, a history of neurological illness or other severe disease such as epilepsy, schizophrenia, pervasive developmental disorders (including autism spectrum disorders) or mental retardation were excluded.

The ADHD and comorbidities were diagnosed according to the DSM-IV criteria based on a semistructured interview using the clinical diagnostic interview scale^50,51 by an experienced child psychiatrist. Accordingly, the children with ADHD can be categorized into inattentive type [ADHD-I, sample size $(n)=22$] and ADHD combined type (ADHD-C, $n=8$). Meanwhile, 25 of the children with ADHD exhibited comorbidities, such as disruptive behavior ($n=8$), anxiety disorder ($n=2$), mood disorder ($n=3$), tic disorder ($n=6$), and learning disorder ($n=17$). In addition, the DSM-IV symptoms of children with ADHD were scored using the ADHD rating scale-IV (ADHD RS-IV) to index the severity of ADHD.⁵² The items were rated on a four-point Likert-type scale (0 = never, 1 = sometimes, 2 = often, and 3 = always) by parents. Accordingly, the “inattentive,” “hyperactive/impulsive,” and “total” scores were computed for each child with ADHD.

This study was approved by the Ethics and Committee of Peking University Sixth Hospital/Institute of Mental Health. All subjects were treated according to the Declaration of Helsinki. Written informed consent was obtained from parents of the children. Meanwhile, the children aged above 10 years old also provided written informed consent by themselves. All the children provided oral consent before the experiment and were free to withdraw from the experiment at any time for any reason without prejudice to future care.

2.2.

fNIRS Data Acquisition and Preprocessing

fNIRS data were collected using a multichannel continuous wave near-infrared optical imaging system (Hui Chuang, China) with a sampling rate of 17 Hz. This system contains 24 light sources and 28 detectors. The optode arrays generated 80 different measurement channels with a fixed 3-cm interoptode distance, which covered primary regions of the whole head, e.g., frontal, temporal, parietal, and visual cortexes [Figs. 1(a) and 1(b)]. The optodes were placed according to the international 10–20 system, with the external auditory canals and vertex as the reference points. A resting-state fNIRS signal recording was lasted at least 12 min. During the recording, the participants were instructed to sit still and close their eyes without falling asleep. Such resting-state recording did not require overt perceptual input or behavioral output. Positions of the measurement channels were labeled by vitamin E capsules on an arbitrarily chosen participant, which were visible in the structural MRI imaging from a Siemens 3.0 Tesla scanner. According to the obtained spatial coordinates, these channels were displayed on Yeo et al.’s network template⁵³ [Fig. 1(c)], in which six functional networks (i.e., the default, frontoparietal control, ventral attention, somatomotor, dorsal attention, and visual networks) were presented and labeled by different colors.

Fig. 1

Schematic diagram of experimental data acquisition. (a) Photo obtained from a participant during the data collection. (b) Optodes and channels. The red circles represent the sources and the blue circles represent the detectors. Meanwhile, the green lines linking the sources and detectors represent the formed measurement channels. (c) The arrangement of the whole-head 80 measurement channels on a functional network brain template.⁵³

We used an in-house FC-NIRS package²⁵ to preprocess the fNIRS data. Similar to our previous studies,³³ a bandpass filter (0.01 to 0.1 Hz) was first conducted to eliminate the effects of low-frequency drift and high-frequency neurophysiological noise. Subsequently, we calculated the relative changes in the concentrations of oxygen-hemoglobin and deoxygen-hemoglobin using the modified Beer–Lambert law.³² Then we extracted 8 min stable hemoglobin time series for each participant. Finally, similar to our previous studies,^33,36,45,46 we conducted a temporal independent component analysis to remove systematic physiological noise (e.g., superficial signal) and motion-induced artifacts.⁵⁴ Specifically, these noise components were identified according to the components’ temporal profiles, spatial maps, and power spectra. A component would be considered noise if it met one of the following criteria: (1) the corresponding temporal profile included sudden jumps, slowly varied U or inverted U-shaped spike, or numerous intercurrent quick spikes (e.g., motion artifacts); (2) the dominant frequency of power spectra of the component was outside the range of 0.01 to 0.1 Hz; and (3) the spatial map of the component presented a global and spatially dispersive pattern (e.g., physiological interference). Once the noise components were identified, the concentration signal was subsequently reconstructed with these particular components eliminated from the original hemoglobin time course. The filtered concentration signal was used for further analysis. In this study, we used oxy-hemoglobin signal to present the following results because the HbO signal generally has a better signal-to-noise ratio than the HbR signal.⁵⁵

2.3.

Functional Connectivity Calculation and Brain Network Construction

For each participant, functional connectivity was calculated by conducting Pearson correlation analyses between time series of every pair of nodes, where the nodes were the measurement channels. This procedure generated an $80\times 80$ correlation matrix for each participant. Of note, these correlation coefficients ($r$) were normalized to $z$-values with Fisher’s $r$-to-$z$ transformation. With a predetermined sparsity that denotes the number of actual connections divided by the maximum possible number of connections in the network, the correlation matrix was then thresholded into a binary matrix that described the topological organization of the functional networks. As in our previous studies,^36,37 we chose the sparsity of 0.2 to construct the brain network.

2.4.

Brain Network Analysis

A graph theory method was used to characterize the topological organization of the brain functional networks in the ADHD and HC groups. Network measures were calculated using our FC-NIRS package.²⁵ In fNIRS-derived brain network studies, topological network efficiency has been frequently used to characterize the capacity of parallel information processing within a brain network. We, therefore, focused on efficiency-related parameters, i.e., nodal efficiency, global efficiency, and local efficiency, to examine the differences in these efficiency measures between the ADHD and HC groups. The definitions for these parameters are described below.

The global efficiency $E_{\mathrm{glob}}$ represents the information transfer efficiency across the network, which is defined as the inverse of the harmonic mean of the shortest path length between any two nodes⁵⁶

In addition, for a given node $i$, its efficiency in information transfer is measured by $E_{\mathrm{nod}}$, which is defined as the harmonic mean of the shortest path length between this node and its neighbors

2.5.

Statistical Analysis

Two-sample $t$-tests were adopted to compare the differences in demographics or core symptoms between the ADHD and HC groups. For functional connectivity, a network-based statistic approach^46,57 was adopted to compare the functional connectivity differences between the ADHD and HC groups. Specifically, two-sample $t$-tests with a threshold of $p<0.001$ were performed to identify the suprathreshold connections. These connections formed one or more subgraphs (components). Subsequently, 1000 permutations were performed to determine the significance of each component. Finally, the most significant component was selected to represent the altered functional connectivity. Furthermore, previous studies have found that the analogous regions in the resting-state network are strongly connected,⁵⁸ and altered homotopic connectivity has been associated with many psychiatric conditions.^59–61 In order to characterize the spatial attributes of the altered functional connectivity, we categorized the altered functional connectivity into three spatially different groups: (1) homotopic connectivity, denoting the interhemispheric connectivity between homologous regions; (2) intrahemispheric connectivity, denoting the connectivity between regions in the same hemisphere; and (3) heterotopic connectivity, denoting the interhemispheric connectivity that was not homotopic connectivity.⁶² For network efficiency, two-sample $t$-tests were also adopted to compare the differences between groups.

2.6.

Relationship Between Altered Brain Functional Connectivity/Network Features and ADHD Core Symptoms

To test the associations between altered brain functional connectivity/network features and core symptoms (e.g., inattentive, hyperactive/impulsive, and total scores) in ADHD, Pearson correlation analyses were performed in the ADHD group with significance threshold of $p<0.05$. Before the correlation analyses, the effects of age, sex, and years of education were removed by multiple linear regression.

3.1.

Demographic and Core Symptoms

The $t$-test results for the demographic and core symptoms between the two groups are listed in Table 1. The ADHD and HC groups showed no significant differences in age or IQ. However, the children with ADHD exhibited significantly higher scores in core symptoms including inattentive, hyperactive/impulsive, and total scores ($p<0.01$) compared to the HCs.

Table 1

The demographic and clinical characteristics of children with ADHD and HCs.

3.2.

Decreased Functional Connectivity in Children with ADHD

Figure 2(a) shows the group-averaged connectivity strength in children with ADHD and HC groups. It was found that the averaged connectivity strength in ADHD was much lower (e.g., in somatomotor and dorsal attention networks) than that in HC although the spatial patterns of the functional connectivity maps between two groups exhibited obvious similarity. Quantitatively, the mean values of connectivity strength and its standard deviations were $0.37\pm 0.13$ for ADHD group and $0.42\pm 0.14$ for HC group [Fig. 2(b)]. Furthermore, the number of functional connectivity strength lower than 0.4 was much larger in ADHD group compared to that in HC group [Fig. 2(c)].

Fig. 2

Spatial patterns of the functional connectivity in ADHD and HC groups. (a) Functional connectivity maps for these two groups. (b) Histograms of the functional connectivity distribution. The functional connectivity displayed approximately normal configuration in both ADHD and HC groups. (c) The stacked bar chart of functional connectivity across different thresholds.

Figure 3 shows the statistical differences in functional connectivity between ADHD and HC groups, in which significantly decreased functional connectivity was consistently found in ADHD group ($p<0.05$). Specifically, the changes in homotopic functional connectivity were mainly located in the default mode network, visual network, and between frontoparietal and dorsal attention networks. For intrahemispheric functional connectivity, the significantly altered connectivity was primarily centered in the right hemisphere involving the regions of default mode network, dorsal attention network and visual networks. For heterotopic functional connectivity, the changes were mainly located in the regions of frontoparietal and somatomotor networks in the right hemisphere and the regions of parietal and visual cortex in the left hemisphere.

Fig. 3

Significantly decreased functional connectivity in children with ADHD. The decreased functional connectivity was categorized into three groups: homotopic, intrahemispheric, and heterotopic connections. The dots represent measurement channels, and the colors label the cortical location of these channels.

3.3.

Disrupted Brain Network Topology in Children with ADHD

For global network properties, the global efficiency in children with ADHD significantly decreased as compared to that in the HC group [Fig. 4(a)]. However, no significant difference was found in the local efficiency between two groups [Fig. 4(b)].

Fig. 4

Group differences in (a) global, (b) local, and (c) nodal efficiencies. In (c), the red circles with white and black points indicate that the nodal efficiency significantly decreased and increased, respectively, in children with ADHD as compared to HC group.

For regional nodal characteristic, the ADHD group exhibited both decreased and increased nodal efficiency in some primary brain regions [Fig. 4(c)] ($p<0.05$). Specifically, the decreased nodal efficiency was mainly located in the right hemisphere involving the somatomotor, default mode, and frontoparietal networks, and the increased nodal efficiency was mainly located in the left hemisphere involving the visual and dorsal attention networks.

3.4.

Relationship Between Brain Network Features and Core Symptoms

Figure 5 shows the correlation relationships between functional connectivity and core symptoms. Four functional connections were found to be associated with the core symptoms. Specifically, functional connectivity that linked the left frontoparietal and right somatomotor networks [green line in Fig. 5(a)] showed a significantly negative correlation with the hyperactive/impulsive score in the children with ADHD [Fig. 5(b)]. The functional connectivity that linked the right frontoparietal and visual networks [brown lines in Fig. 5(a)] showed a significantly negative correlation with both the hyperactive/impulsive score and the total score in the children with ADHD [Fig. 5(c)].

Fig. 5

The relationship between functional connectivity and core symptoms in the ADHD group. (a) The connections showed significant correlation with core symptoms. (b) The scatter plotted between core symptoms and functional connectivity. (c) The scatter plotted between core symptoms and mean functional connectivity between right frontoparietal and visual networks.

For nodal efficiency, we identified one node in the right somatomotor network, as indicated by a black arrow in Fig. 6(a), showed a significantly negative correlation with both the hyperactive/impulsive and the total scores [Figs. 6(b) and 6(c)]. For the global and local network efficiency measures, no significant correlations were found between these features and core symptoms.

Fig. 6

The relationship between nodal efficiency in the right somatomotor network and core symptoms.