Twenty-eight college students with IGD were recruited in our study based on the criteria of the modified Young Diagnostic Questionnaire for Internet addiction (YDQ) by Beard and Wolf (Young, 1998; Beard and Wolf, 2001). Young suggested that respondents who answered five or more “yes” for the eight questions were considered to be an internet dependent user (Young, 1998). Beard and Wolf modified the YDQ criteria (Beard and Wolf, 2001), proposed that respondents who answered “yes” to questions 1 to 5 and at least to any one of the remaining three questions were classified as suffering from IA, which was used for screening subjects for the present study. We asked the subjects to recall their life-style when they were initially addicted to the internet, which was a retrospective measure for the addiction is a gradual process and we planned to explore linear changes of the brain structure. We retested them with the YDQ criteria modified by Beard and Wolf (Brand et al., 2014) to verify that they qualified for IA diagnosis. By communicating with their parents via telephone we confirmed the reliability of the self-reports from the IGD subjects. We also confirmed this information from their roommates and classmates that if they often played internet game till late night so that disturbing others’ life. Twenty-eight age and gender matched (p > 0.05) healthy controls without personal or family history of psychiatric disorders were also recruited in our study. In order to ensure that the healthy controls were not suffering from IGD, they were administered by modified YDQ for Internet addiction of Beard and Wolf. All recruited participants were native Chinese speakers, right-handed. Urine test was performed for all subjects to exclude substance abuse before magnetic resonance imaging (MRI) scanning. Exclusion criteria for both groups were (1) neurological disorders or physical illness, including brain tumor, hepatitis, or epilepsy assessed by clinical evaluations and medical records; (2) alcohol, nicotine or drug abuse; and (3) pregnancy or menstrual period in women; Written consent forms were obtained by all the patients and controls. More detailed demographic information was given in Table 1.

Brain imaging scan was performed on a 3T GE scanner at imaging center of Xi’an Jiontong university first affiliated hospital. A standard birdcage head coil and restraining foam pads were used to minimize head motion and protect hearing. The axial 3D T1-weighted images were obtained with a spoiled gradient recall sequence and the following parameters: repetition time (TR) = 8.5 ms; echo time (TE) = 3.4 ms; flip angle (FA) = 12°; field of view (FOV) = 240 × 240 mm²; data matrix = 240 × 240; slices = 140; voxel size = 1 × 1 × 1 mm.

MRI structure data was analyzed with FSL-VBM (Douaud et al., 2007),¹ an optimized VBM protocol (Good et al., 2001) of FSL (Smith et al., 2004). First, structural images were brain-extracted and segmented gray matter was registered to the MNI 152 standard space using non-linear registration (Andersson et al., 2007). The resulting images were averaged and flipped along x-axis to create a left-right symmetric, study-specific gray matter template. Second, all native gray matter images were non-linearly registered to this study-specific template and “modulated” to correct for local expansion (or contraction) due to the non-linear component of the spatial transformation. The modulated gray matter images were then smoothed with an isotropic Gaussian kernel with a sigma of 3 mm. Finally, voxel wise GLM was applied by correcting for multiple comparisons across space. Regional structure in gray matter were assessed by permutation-based non-parametric testing (5000 times) (Nichols and Holmes, 2002).

Color-word Stroop task was implemented by E-prime 2.0 software. This task included a block design with three conditions, i.e., congruent, incongruent and rest. Red, Blue, and Green, three words were displayed in three colors (red, blue and green) as the congruent and incongruent stimuli. During rest, the subjects just focused their eyes on the cross displayed at the center of the screen. We designed two runs with different sequences of congruent and incongruent blocks (Xing et al., 2014). We tested participants individually in a quiet room and the participants kept a calm state of mind. Each of them was instructed to respond to the displayed color as fast as possible by pressing a button on a Serial Response Box TM with the right hand. The index, middle, and ring finger of right hand corresponding to red, blue, and green were used to press button respectively. The behavior data was collected two or three days before MRI scanning after practice.

Analysis of covariance (ANCOVA) was employed with age, gender effects and total intracranial volume as covariates. We used a post hoc correlation analysis to investigate relationship between GMV and behavioral performances in IGD group, and response errors and response time for incongruent condition of color-word Stroop task were employed to be the factors of correlation respectively of IGD group.

A significant Stroop effect were observed in both group, where the RT was longer for the incongruent relative to the congruent condition (IGD group: 628.24 ± 59.20 vs. 549.38 ± 44.17 and control group: 707.52 ± 66.43 vs. 581.97 ± 39.35; p < 0.005). The IGD group committed more errors than the control group during the incongruent condition (IGD group: 8.67 ± 5.41 vs. control group: 6.64 ± 3.65; p < 0.05), and the response delay (RD) measured by RT during the incongruent condition minus congruent conditions was significantly different between these two groups (IGD group: 78.87 ± 45.38 vs. control group :125.56 ± 49.20; p < 0.05) (Table 2).

VBM comparison indicated decreased GMV in several brain areas, i.e., the bilateral ACC, precuneus, supplementary motor area (SMA), superior parietal cortex, left DLPFC, left insula and bilateral cerebellum in the IGD group compared with the control group (Figure 1).

Correlation analysis showed that the GMV of the ACC negatively correlated with Stroop task response errors for incongruent condition in the IGD group (Figure 1), but there was no statistical correlation between the GMV and RT for incongruent condition in the IGD group.

In order to verify impaired cognitive control ability in adolescents with IGD, a color-word Stroop task was used in the current study. Consistent with previous findings (Dong et al., 2011, 2013a; Yuan et al., 2013a,b), the IGD group committed more errors than the control group during the incongruent condition, which demonstrated that adolescents with IGD showed impaired cognitive control ability, as measured by the color-word Stroop test. The result that the RT during incongruent condition and RD of IGD group were shorter than control group may be imply that IGD subjects showed a different reaction pattern relative to controls, and they responded fast but taking the risk of making more errors, which was clearly a change in the response strategy. The study also found that the GMV of the ACC, DLPFC, precuneus, SMA, superior parietal cortex, insula, and cerebellum in the IGD group changed, which is in line with published IGD studies. Zhou and Weng et al. reported GMV reduction or abnormal activation in some brain areas in IGD subjects (Yuan et al., 2011; Zhou et al., 2011; Sun et al., 2012; Ko et al., 2013b; Weng et al., 2013). Although no study reported GMV of the precuneus decreased, fMRI study reported that the precuneus showed abnormal activation during cue-induced task in IGD subject (Ko et al., 2013a,b). The superior parietal cortex was found to be related with cognitive control (Durston et al., 2002, 2003; Ko et al., 2013a).

The correlation between the GMV of ACC and response errors showed that less GMV of ACC in IGD group was associated with more response errors during the incongruent condition in color-word Stroop task, which is a promising finding for the present study. The role of the ACC in cognitive control was well established and has been reported in a number of fMRI studies on Stroop interference paradigm in normal participants. Botvinick et al. reported the ACC was involved with conflict-monitoring function, for the ACC was more active under conditions of high conflict (Botvinick et al., 1999). Another research of Angus W. MacDonald III discovered that the activity of ACC was dissociable from top–down control, and it played a consistent role in monitoring conflict during response period (MacDonald et al., 2000). The study of Kerns revealed that the conflict-related activity of ACC predicted both greater prefrontal cortex activity and adjustments in behavior, supporting a role of ACC in conflict monitoring and cognitive control (Kerns et al., 2004). Furthermore, Matsumoto demonstrated that the cognitive control recruited by the ACC may be “consequential” based on conflicts between evoked plans and concrete actions (Matsumoto and Tanaka, 2004). A large body of experimental evidence on numerous diseases has accrued to support the important function of ACC in cognitive control. Akio Soeda et al. studied traumatic brain injury (TBI) patients and found out that the decreased activation in the ACC may be associated with alteration in functional cerebral activity, which may reflect either cortical disinhibition attributable to disconnection or compensation for an inefficient cognitive process (Soeda et al., 2005). Abnormal activity of ACC has been found in many mental problems, including obsessive–compulsive disorder (OCD), attention deficit-hyperactivity disorder (ADHD), and major depressive disorder (MDD; Ursu et al., 2003; Liotti et al., 2005; Murali and George, 2007). Recent neuroimaging studies also found altered activation of the ACC in heroin- and opioid-dependent individuals in GO/NOGO paradigm (Forman et al., 2004), suggesting the ACC is a key area in response inhibition (Fu et al., 2008). The research on cocaine users confirmed activity of ACC in inhibitory control (Kaufman et al., 2003; Goldstein et al., 2007, 2009). A Magnetic Resonance Spectroscopy (MRS) study on nicotine dependence showed that glutamate + glutamine (Glx) levels reduced in ACC, indicating the ACC was involved in cognitive control by modulating behavior (Wheelock et al., 2014). In a word, the ACC is important for cognitive control ability. The ACC’s structural abnormalities and dysfunction in IGD have been reported in previous studies. The VBM results of Zhou et al. showed that the GMV of ACC decreased in IGD compared with controls (Yuan et al., 2011; Zhou et al., 2011). Many researches on IGD indicated that the ACC participated in the cognitive control, such as inhibitory control, error monitoring, and decision making (Dong et al., 2012, 2013a,b).