Abnormal Functional Connectivity in Cognitive Control Network, Default Mode Network, and Visual Attention Network in Internet Addiction: A Resting-State fMRI Study.

Internet addiction (IA) has become a global mental and social problem, which may lead to a series of psychiatric symptoms including uncontrolled use of internet, and lack of concentration. However, the exact pathophysiology of IA remains unclear. Most of functional connectivity studies were based on pre-selected regions of interest (ROI), which could not provide a comprehensive picture of the communication abnormalities in IA, and might lead to limited or bias observations. Using local functional connectivity density (lFCD), this study aimed to explore the whole-brain abnormalities of functional connectivity in IA. We evaluated the whole-brain lFCD resulting from resting-state fMRI data in 28 IA individuals and 30 demographically matched healthy control subjects (HCs). The correlations between clinical characteristics and aberrant lFCD were also assessed. Compared with HCs, subjects with IA exhibited heightened lFCD values in the right dorsolateral prefrontal cortex (DLPFC), left parahippocampal gyrus (PHG), and cerebellum, and the bilateral middle cingulate cortex (MCC) and superior temporal pole (STP), as well as decreased lFCD values in the right inferior parietal lobe (IPL), and bilateral calcarine and lingual gyrus. Voxel-based correlation analysis revealed the significant correlations between the Young's Internet Addiction Test (IAT) score and altered lFCD values in the left PHG and bilateral STP. These findings revealed the hyper-connectivity in cognitive control network and default mode network as well as the hypo-connectivity in visual attention network, verifying the common mechanism in IA and substance addiction, and the underlying association between IA, and attention deficit/hyperactivity disorder in terms of neurobiology.

Introduction

Internet addiction (IA), as a behavioral addiction, was conceptualized as the poorly controlled use of internet with impairment of interpersonal relationships, and psychological and social functioning (1). With the popularity of the internet and the improvement of information technology in the last 2 decades, IA has become a serious public health problem due to its high prevalence around the world (2, 3). In Hong Kong, the prevalence rate of IA has dramatic increased from 3.0 to 26.8% in the last 10 years, with the proportion of teenagers who spend more than 20 h online per week increasing from 25.4 to 42.0% (4). The potential risk of IA has attracted growing concerns from worldwide researchers, and mounting studies, and papers have been published. Besides uncontrolled use of internet, multiple comorbid psychiatric symptoms of IA have been revealed by clinical studies, including anxiety, depression, and lack of concentration (5, 6). Epidemiological research further demonstrated that attention deficit/hyperactivity disorder (ADHD), unsociable personality, and bad relationship with parents were high risk factors of IA (6–8). Additionally, ADHD was thought to be the main complication of IA (9). However, the exact pathophysiology of IA has not been fully understood yet, and the underlying mechanism of the coexistence of IA, and comorbid symptoms remains unclear.

Although the pathogenesis has not been well-elucidated, resting-state fMRI provides a useful platform for understanding the neuropathological changes in IA. Resting state functional connectivity (FC), which reflects the temporal coherence of inter-regional spontaneous blood oxygen level dependent (BOLD) fluctuations in cerebral activity (10), was extensively used to investigate the functional interactions between brain areas in psychiatric research, including IA. In previous fMRI studies, IA was associated with the altered FC of cognitive control network (CCN) (11), which was considered as the main neurobiological characteristic of substance addiction. For instance, enhanced FC between the dorsolateral prefrontal cortex (DLPFC), and the temporoparietal junction (TPJ) was observed in IA individuals (12), which was associated with many other cognitive disorders (13, 14). Furthermore, altered FC within the CCN was also demonstrated. Compared with healthy subjects, adolescents with IA exhibited heightened FC between the DLPFC, and anterior cingulate cortex (ACC), correlating with the symptom severity of IA (15). According to Volkow's addiction model, abnormal CCN was proposed as the key determinant in development, and maintenance of substance addiction (16). The altered FC of CCN found in IA individuals, giving a pathophysiological explanation for the uncontrolled use of internet, suggested that IA might share similar neurobiological mechanism with substance addiction (17). However, most previous FC studies in IA were based on a priori selection of region of interest (ROI). This seed-based analysis cannot provide a whole-brain picture of FC alterations in IA, and may lead to limited or biased observations due to the restriction on ROI.

Local functional connectivity density (lFCD) has been developed to analyze the whole-brain FC abnormalities by computing the temporal correlations of every pair of neighboring voxels in the entire brain (18, 19). Voxels with a higher number of functional connections have greater lFCD values, and are considered to be important functional hubs in the information processing. Unlike seed-based FC, lFCD provides an unbiased approach to the study of whole-brain functional connectivity (18). As a result, extra abnormalities have been revealed by lFCD in many disorders, which would not have been detected using traditional seed-based FC analysis methods (20–22). Investigating the alterations of lFCD in IA may provide the full picture of the communication abnormalities of whole-brain functional network, and offer a more reliable, and comprehensive result beyond the FC. This may help to improve our understanding of the neuropathological mechanism of IA and the common neurobiological dysfunction underlying the coexistence of IA, and comorbid symptoms.

In the current study, we aim to investigate the lFCD dysfunction in IA individuals and to detect the potential relationships between the altered lFCD values and the related symptom scores. Local FCD from 28 IA individuals were compared to those of 30 demographically matched healthy controls (HCs). Furthermore, a voxel-based correlation analysis was performed to examine the possible associations between the clinical variables and the lFCD values of the regions with significant differences.

Materials and Methods

Participants

A total of 58 subjects (28 IAs and 30 HCs) were recruited from university students. All participants were right-handed and native Chinese speakers. The inclusion criteria for IA group were as follows: (1) aged between 18 and 30 years; (2) in line with the diagnostic criteria of IA via Young's Diagnostic Questionnaire (YDQ) (1); (3) a score of 60 or higher on Young's Internet Addiction Test (IAT); (4) not under any form of therapeutic interventions; and (5) without any other organic or mental diseases. Thirty HCs, well-matched with IA group in age, gender, and education, were recruited with the following inclusion criteria: (1) aged between 18 and 30 years; (2) inconsistent with the IA diagnostic criteria by Young (1); (3) IAT score <50; and (4) without any other organic, or mental diseases. Participants with a history of substance addiction were excluded in our study. Additionally, pregnant, or lactating women were also excluded.

The protocol of this study was evaluated and approved by the Sichuan Regional Ethics Review Committee on traditional Chinese medicine (ethical approval number 2016KL-005), in accordance with the Declaration of Helsinki (2000). All participants signed informed consent forms before inclusion and were compensated for their participation.

Questionnaire

Besides IAT, the Yale-Brown Obsessive Compulsive Scale (YBOCS) (23) and the Barratt Impulsiveness Scale-11 (BIS-11) (24) were finished by all subjects, and were used to assess obsessive-compulsive and impulsive behaviors, respectively. All scales were translated into Chinese. We also collected information on sex, age, educational level, and years of internet use via self-designed questionnaire.

Data Acquisition

All fMRI data were collected on a 3.0T MRI scanner (GE Discovery MR 750, USA) with a standard 8-channel head coil. During scanning, participants were instructed to keep their eyes closed and head still, and remain awake without thinking anything in particular. Ear plugs and foam pads were used to restrict noise and displacement of head, respectively. Functional images were acquired using a standard Echo Planar Imaging sequence with the parameters as follows: repetition time = 2000 ms, echo time = 30 ms, flip angle = 90°, field of view = 24 × 24 cm2, image matrix = 64 × 64, no gap, and voxel size = 3.75 × 3.75 × 4.4 mm3. Two hundred and 55 volumes of image were acquired, and each volume included 35 slices.

Date Preprocessing

Functional images were preprocessed using the neuroscience information toolbox (NIT, http://www.neuro.uestc.edu.cn/NIT.html), which is based on Statistical Parametric Mapping (SPM8, http://www.fil.ion.ucl.ac.uk/spm/). Prior to preprocessing, the initial 5 volumes of each subject were removed to minimize the interference of instability in the initial signals. Subsequently, slice timing was conducted to correct the time delay between slices. Spatial realignment was performed to correct the head motion. Participants with more than 2 mm displacement or more than 2° rotation were excluded. The fMRI images were then normalized to standard Montreal Neurological Institute (MNI) template with a resolution of 3 × 3 × 3 mm3. After that, we regressed out 24 head motion parameters and signals from cerebral spinal fluid, and white matter. Finally, band-pass filtering (0.01–0.08 Hz) was performed to reduce the interference of low-frequency drift and high-frequency noise (25, 26).

Local FCD Calculation

Local FCD, which is defined as a sum of the number of efficient neighboring functional connections of a given voxel, reflects the voxel-wise whole-brain functional connectivity strength. In our study, lFCD calculation was conducted using the NIT toolbox, based on the proposal by Tomasi and Volkow (18, 19). First, Pearson correlation coefficients between the time course of a given voxel and those of its neighboring voxels were computed. According to the seminal paper (18), functional connections between two voxels with a correlation coefficient R > 0.6 were considered significant, as thresholds of R < 0.4 might lead to too many false positives, while thresholds of R > 0.7 might reduce the sensitivity of lFCD maps (27). All voxels were calculated to generate the whole brain lFCD map. Then we smoothed the lFCD maps with a Gaussian kernel of 6 mm. A more detailed procedure of lFCD calculation can be found in our previous study (28).

Statistical Analysis

Using SPSS 18.0, we compared demographic data and clinical variables between IA individuals and HCs. Two-sample t-test was performed to compare the intergroup differences in continuous variables (e.g., age, years of internet use, scores on clinical scales) while chi-square test was used for categorical variable (e.g., gender). A P < 0.05 was considered to be statistically significant. Then a two-sample t-test was performed to compare lFCD values between IA individuals and HCs, and a P < 0.01 with cluster size > 23 adjacent voxels was deemed significant according to the random field theory (29), which set the threshold for the spatial extent of clusters is extent−threshold = 638 mm3. In addition, a voxel-based correlation analysis was conducted between the lFCD values and the clinical scores and then the potential association between aberrant lFCD and severity of symptoms was obtained by intersection of lFCD and correlation results. Statistical significance was set at P < 0.05 (FDR corrected).

Results

Demographic Characteristics and Clinical Measures

Subject demographics and clinical scores of 28 IA individuals and 30 HCs are shown in Table 1. No significant intergroup difference was observed in age, sex, or years of education between two groups (P > 0.05). Consistent with the inclusion, IA individuals exhibited higher scores of IAT, YBOCS, BIS-11 significantly (P < 0.001), although the years of internet use showed no difference (P > 0.05).

Table 1

Demographics and clinical characteristics of the individuals with internet addiction and the healthy controls.

	Internet addiction group (n = 28)	Healthy control group (n = 30)	P-value#
	m ± sd	m ± sd
Age (years)	21.32 ± 1.96	21.73 ± 2.08	0.45
Gender (male/female)	21/7	22/8	0.89
Education (years)	15.21 ± 1.84	15.77 ± 1.82	0.26
Internet use (years)	8.14 ± 2.84	8.23 ± 2.33	0.90
Framewise displacement	0.030 ± 0.011	0.031 ± 0.011	0.79
Young's internet addiction test (IAT)	73.89 ± 6.76	29.90 ± 7.18	<0.001
Yale-Brown Obsessive Compulsive Scale (YBOCS)	15.64 ± 7.40	4.30 ± 5.07	<0.001
Barratt Impulsiveness Scale-11 (BIS-11)	79.57 ± 9.37	62.53 ± 8.57	<0.001

Two-sample t-test and chi-square test were performed to evaluate the intergroup difference in continuous and categorical variables, respectively.

Local FCD Results

As shown in Table 2 and Figure 1, in comparison to HCs IA individuals exhibited lower lFCD values in the visual attention regions (P < 0.01, with the cluster-level corrected), including the right inferior parietal lobe (IPL, comprising supramarginal gyrus and inferior parietal cortex), and bilateral calcarine and lingual gyrus. Higher lFCD values (P < 0.01, with the cluster-level corrected) were observed in brain regions belonging to the DMN, including the left parahippocampal gyrus (PHG), and bilateral superior temporal pole (STP). In addition, enhanced lFCD values in the CCN were also demonstrated (P < 0.01, with the cluster-level corrected), including the right DLPFC (comprising superior frontal gyrus), left cerebellum, and bilateral middle cingulum cortex (MCC). We further evaluated the influence of framewise displacement (FD), which might partially index physiological noise (30, 31). No significant intergroup difference in FD and no correlation between FD and lFCD values were found (P > 0.05).

Table 2

Differences in lFCD between IA individuals and HCs.

AAL Regions	Cluster voxels	MNI coordinates			Peak T-value	P uncorrected
		X	Y	Z
CER_L	29	−39	−39	−33	4.2465	<0.01
STP_R	32	9	9	−15	3.8903	<0.01
STP_L	57	−21	10	−30	3.8258	<0.01
PHG_L	40	−21	0	−30	4.1046	<0.01
CAL_R	173	15	−81	9	−5.5509	<0.01
CAL_L	87	3	−85	7	−4.3274	<0.01
LING_R	156	10	−66	4	−4.3057	<0.01
LING_L	53	−6	−82	0	−3.2804	<0.01
SMG_R	42	63	−48	42	−3.6324	<0.01
IPC_R	34	59	−49	42	−3.4958	<0.01
MCC_R	74	9	−15	36	3.7217	<0.01
MCC_L	24	−7	−17	39	3.5067	<0.01
SFG_R	40	21	−3	57	3.6759	<0.01

lFCD, local functional connectivity density; IA, internet addiction; HCs, healthy controls; MNI, Montreal Neurological Institute; L, left; R, right; CER, cerebellum; STP, superior temporal pole; PHG, parahippocampal gyrus; CAL, calcarine; LING, lingual; SMG, supramarginal gyrus; IPC, inferior parietal cortex; MCC, middle cingulum cortex; SFG, superior frontal gyrus.

Figure 1

Significant intergroup differences in lFCD between the internet addiction individuals and healthy control subjects. Compared with healthy control subjects, individuals with IA exhibited increased lFCD values in the right superior frontal gyrus, left cerebellum, and parahippocampal gyrus, and bilateral middle cingulum cortex, and superior temporal pole. Decreased lFCD values were exhibited in the right inferior parietal lobe and bilateral calcarine and lingual gyrus.

Relationship Between lFCD Values and Clinical Scores

Voxel-based correlation analysis revealed positive correlations between the IAT score and the heightened lFCD values of left PHG (r = 0.522, P = 0.004) and bilateral STP (Left: r = 0.500, P = 0.007; Right: r = 0.634, P < 0.001) after FDR correction (Figure 2). However, there was no remarkable association between the neural activity changes and BIS-11 or YBOCS after multiple comparison correction, although the IA subjects exhibited higher BIS-11 and YBOCS scores.

Figure 2

Correlations between lFCD values and IAT score in IA group. The IAT score was positively associated with lFCD values in the (A) STP_L, (B) STP_R, and (C) PHG_L. lFCD, local functional connectivity density; IAT, Young's internet addiction test; L, left; R, right; PHG, parahippocampal gyrus; STP, superior temporal pole.

Discussion

Using a voxel-wise data-driven method, the current study investigated the changes of whole-brain functional connectivity in IA individuals. Similar to substance addiction, IA was manifested with enhanced lFCD in the CCN. In addition, increased lFCD in the DMN and decreased lFCD in the visual attention network were both related to impaired attention, and were also observed in our study. Furthermore, we demonstrated the association between the severity of IA and altered lFCD values in DMN. These findings might provide novel insights into the underlying pathophysiological mechanisms of IA and the underlying association between IA and complications.

Increased lFCD Values in CCN

The CCN, which underlies the ability to allow for appropriate actions through the suppression of inappropriate motor, is thought to be the key determinant in substance dependence model (16). In our study, the IA individuals exhibited heightened lFCD values in the right DLPFC, left cerebellum and bilateral MCC, all of which are main components of the CCN.

The DLPFC is one of the most well-investigated areas in the brain, in particular in the context of cognitive function. As the core node of CCN, the DLPFC is involved in a variety of advanced cognitive functions essential for decision-making and goal-directed behavior, such as self-control (32), action planning (33), and intertemporal choice (34). Dysfunction in DLPFC has been found in many cognitive control disorders (35–37), in particular in substance dependence (38, 39). Evidence from previous neuroimaging studies has demonstrated the pivotal role of DLPFC dysfunction in craving for marijuana (40), cigarette (41), and alcohol (42), and, thus, the altered DLPFC function was presumed as a neuropathological explanation of substance addiction (43). The cingulate cortex, involved in cognition and decision making, is also implicated in various addictions (44, 45). As a vital component of cingulate cortex, the MCC is engaged in decision making for reward/approach and fear/avoidance selection by assessment of potential outcomes (46, 47), and therefore closely related to substance dependence (48).

As to IA, multiple studies attempting to elucidate the pathological mechanism of IA have converged on the abnormalities of the DLPFC and cingulate cortex (49, 50). Higher Regional homogeneity (ReHo) values in the DLPFC and cingulate cortex have been revealed, reflecting the enhanced synchronization between the DLPFC, and cingulate cortex in individuals with IA (51). Additionally, hyperactivity in the DLPFC and cingulate cortex was also found when the IA subjects were exposed to an internet-related cue (52). Furthermore, altered gray matter in the DLPFC and MCC were revealed in IA group by structural MRI (sMRI), suggesting that the functional alterations in the DLPFC, and MCC were based on the structural abnormalities (53, 54). In our study, the heightened lFCD values in the right DLPFC and bilateral MCC observed in IA individuals further demonstrated the dysfunctional CCN in IA, providing new evidence supporting the hypothesis that IA shares similar neurobiological mechanism with substance addiction.

The cerebellum was traditionally considered as a motor center, well-known for its involvement in postural, and motor control. However, in recent studies, cerebellum abnormalities were found in cognitive disorders (55, 56). Conversely, cerebellum resections or lesions caused various types of abnormal social behavior (57, 58). Providing new insights into the cerebellum function, recent studies revealed the role of cerebellum in cognitive control function (59, 60). Furthermore, structural and functional changes in the cerebellum were found in substance-dependent individuals who exhibited uncontrolled craving for substance (61, 62). Similar to those in substance addiction, cerebellum abnormalities were also demonstrated in IA, associated with the severity of addiction (63). Consistent with the previous studies, our results proved the remarkably increased lFCD in cerebellum in IA group, supporting the speculation that there were homological neurobiological changes in the CCN in IA and substance addiction.

Increased lFCD in DMN and Decreased lFCD in Visual Attention Network

The DMN is generally regarded as the most significant component of the resting state network, due to its typical deactivation during attention-demanding cognitive tasks and activation during internal mentation (e.g., retrieval of episodic memory, self-referential, and imagination of future) (64–67). Dysfunctional DMN is implicated in numerous mental disorders, in particular in ADHD. In our study, enhanced lFCD values in the left PHG and bilateral STP, both are integral parts of the DMN, were revealed in IA group.

As an important part of limbic system, the PHG is responding to emotional, sensory, and environmental stimuli (68–70), and involved in contextual associations, which means processing schema (object-context) associations (71). Due to the function of linking specific sensations to contextual predictions, the PHG is thought to play an essential role in attention (72). Similarly, as an integral component of paralimbic region, the temporal pole is involved in spatial navigation and retrieval of memory (73, 74), which are also thought to be important in attentional allocation (72). In previous studies, structural, and functional alterations in the PHG and temporal pole were implicated in the neuropathology of ADHD (75–78). Consistent with those in ADHD, lFCD in the PHG and STP were revealed in our study, implicating the enhanced functional connectivity of DMN in IA individuals.

Contrary to the enhanced DMN, the weaker lFCD values in the right IPL, and bilateral calcarine, and lingual gyrus were observed in our study. All of these regions are components of the visual attention network (79, 80), involved in the bottom-up attentional pathway which is proposed as a circuit-breaker to reorient attention to new and external information (81). Abnormalities in visual attention network were associated with impaired concentration in previous studies (72, 82).

As the primary visual cortex, the calcarine plays a major role in visual information integration and attention processing (83). Structural and functional alterations in the calcarine were revealed in ADHD (84, 85). As the secondary visual cortex, the lingual gyrus has a critical function in spatial memory (86), and visual attention (87). Likewise, abnormalities in the lingual gyrus in ADHD children were identified in the existing literatures (88, 89). Due to its core role in attentional reorienting and visuomotor integration (80), the IPL is regarded as a supramodal core in the ventral attention network. Hypoactivity in ventral attention network, implicating insensitivity with external environment, is proposed as a key determinant in impairment of attention (90). Both structural and functional abnormalities in IPL were associated with ADHD in previous studies (91, 92). As to IA, abnormalities in the IPL and lingual gyrus were also revealed in IA adolescents, implicating the role of visual attention network in the impaired concentration in IA (93, 94).

The association between IA and ADHD has been demonstrated in previous epidemiology studies. ADHD is considered as the most remarkably associated mental disorder of IA (9). Adolescents with IA have obviously higher ADHD symptoms, among which lack of concentration is the most associated one with IA (95). Prior neural evidence has implicated the importance of hyper-connectivity within DMN (96, 97) and hypo-connectivity within the visual attention network (98) in the inattention in patients with ADHD. In our study, the increased lFCD values in the left PHG and bilateral STP and the decreased lFCD values in the right IPL and bilateral calcarine and lingual gyrus were observed, indicating the heightened functional connectivity of DMN and reduced functional connectivity of visual attention network in IA, which were similar with those in ADHD. To our knowledge, this was the first time the STP and calcarine were associated to IA. These findings, providing direct neural explanation for the impaired concentration in IA, demonstrated the underlying association between IA and ADHD in terms of neurobiology. Additionally, we further revealed the positive correlations between the IAT score and the lFCD in left PHG and bilateral STP, implicating the pivotal role of DMN dysfunction in development of IA.

Limitations

Our study has several limitations. First, FCD was restricted to the local functional connectivity in our study. We did not calculate the global FCD due to the misgiving of its high variability (99). And furthermore, according to the seminal paper, the local FCD, and global FCD are proportional to one another (18). Second, due to the small subject size and the sex difference in morbidity, our study fails to analyze the intergender differences. Finally, subjects are restricted to college students, which limits the generalization of this study.

Conclusion

In the present study, a voxel-wise method was performed to analyze the whole-brain FC abnormalities in IA subjects. Our results revealed that individuals with IA exhibited heightened lFCD values in the CCN and DMN as well as decreased lFCD values in the visual attention network. Our observations, providing neural explanation for the lack of concentration and uncontrolled use of internet in IA individuals, verified the similar neurobiological mechanism in IA with that in substance addiction, and implicated the underlying association between IA and ADHD in terms of neurobiology.

Data Availability

The datasets generated for this study are available on request to the corresponding author.

Ethics Statement

This study was carried out in accordance with the recommendations of Ethical Principles for Medical Research Involving Human of Chengdu University of TCM with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Sichuan Regional Ethics Review Committee on traditional Chinese medicine.

Author Contributions

TZ, CL, and DY conceptualized the study, designed the protocol, and administered the project. HL supervised the study. YW, YD, CW, LZ, and CZ preformed the experiments. YQ, BS, ZL, and XL conducted the statistical analysis. YW wrote the first draft of the manuscript. All authors read and approved the manuscript.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.