Radiomics-based feature exploration and auxiliary diagnostic model construction for gaming disorder.

Radiomics-based feature exploration and auxiliary diagnostic model construction for gaming disorder - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Sci Rep . 2026 Apr 10;16:16874. doi: 10.1038/s41598-026-47980-y Radiomics-based feature exploration and auxiliary diagnostic model construction for gaming disorder Yifan Li 1, # , Shuhong Lin 1, # , Ying Tang 1 , Qiuping Huang 2 , Zhenjiang Liao 1 , Xinxin Chen 1 , Chenhan Wang 3 , Jingyue Hao 1 , Xuhao Wang 1 , Ru Yang 4, ✉ , Hongxian Shen 1, ✉ Author information Article notes Copyright and License information 1 National Clinical Research Center for Mental Disorders, and Department of Psychiatry, The Second Xiangya Hospital of Central South University, 139 Renmin Road, Changsha, 410011 Hunan China 2 Department of Psychology, School of Humanities and Management, Hunan University of Chinese Medicine, Changsha, China 3 The Affiliated Children’s Hospital of Xiangya School of Medicine, Central South University, Changsha, China 4 Department of Radiology, The Second Xiangya Hospital of Central South University, 139 Renmin Road, Changsha, 410011 Hunan China ✉ Corresponding author. # Contributed equally. Received 2025 Dec 2; Accepted 2026 Apr 6; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13230922  PMID: 41963491 Abstract Subjective biases in current clinical assessments create an urgent need for objective biomarkers in Gaming Disorder (GD). Radiomics could help in diagnosis based on whole brain analysis and multidimensional indicators. This study aimed to find classification features of GD based on radiomics and to develop an auxiliary diagnostic model for GD. A total of 141 individuals with GD and 73 healthy controls underwent a clinical assessment and resting-state functional magnetic resonance imaging scans. Radiomics was used for feature extraction and selection, support vector machine was employed to construct a classification model, and permutation test was applied to verify model performance. The model incorporated 67 brain functional features (accuracy: 86%, sensitivity: 93%, specificity: 75%, AUC: 0.92), primarily concentrated in the Default Mode Network, Executive Control Network, and Salience Network. Among these, right precuneus-left anterior cingulate cortex connectivity not only contributed most to the model’s classification but also was significantly correlated with GD symptom severity (total score, withdrawal symptoms, tolerance, persistent engagement despite harm, and escape from problems/negative emotions). Radiomics provides a promising framework to identify GD features, unravel neural mechanisms, and assist objective diagnosis. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-026-47980-y. Keywords: gaming disorder, machine learning, radiomics, resting-state functional magnetic resonance imaging, brain network, diagnostic model Subject terms: Biomarkers, Computational biology and bioinformatics, Diseases, Health care, Medical research, Neurology, Neuroscience Introduction Gaming disorder (GD) is an uncontrolled gaming behavior that primarily affects adolescents, severely impairing their social functioning. GD directly contributes to vision impairment, obesity and biorhythmic disorders 1 , 2 , and has a significant negative impact on academic and work performance 3 , 4 . Currently, GD is common worldwide, with an overall prevalence of approximately 3.3% and a prevalence of up to 12% in East Asia 5 , 6 . Under the impact of the COVID-19 pandemic, the time adolescents spent on electronic games increased significantly during the pandemic period. A subset of them continued to engage in high-intensity gaming behavior even after the pandemic, which in turn led to a rise in the risk of GD and scores on its core addiction symptom (e.g., withdrawal symptoms, tolerance symptoms scores) 7 , 8 . However, due to a lack of awareness of the disease or a sense of stigma, people with GD tend to conceal or deny their behaviors 9 , 10 , and the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5) also states that people with GD “may lie to family members or doctors about their amount of gaming" 11 . Many studies have now diagnosed GD using self-report scales, which may lead to GD being missed or overshadowed by mood disorders 12 . Therefore, choosing the neurobiological indicators is valuable in the diagnosis of GD. Resting-state functional magnetic resonance imaging (rs-fMRI), as a tool that objectively reflects brain functional status, has been widely used in GD research 13 – 17 . Existing studies have reported that individuals with GD exhibit increased activity in the ventral striatum and decreased activity in the dorsolateral prefrontal cortex 18 , as well as abnormalities in functional connectivity (FC) between brain regions. Additionally, some studies have proposed that subtype classification of GD can be performed based on FC patterns 19 . For example, Wang et al. categorized GD patients into two subgroups (craving-related and mixed psychological) based on FC 19 . From the network perspective, the Default Mode Network (DMN), Executive Control Network (ECN), Salience Network (SN), and limbic system, among others, have all been reported to exhibit functional abnormalities in GD 20 – 22 . However, there is heterogeneity among these rs-fMRI findings: different studies often report distinct brain regions with abnormal activity, and even contradictory results have been documented for the same brain region. For example, some studies have found increased activity in the anterior cingulate cortex (ACC) in GD patients 23 , 24 , whereas others have reported decreased activity in this region 25 , 26 ; with respect to the functional connectivity of the DMN, some studies have demonstrated reduced connectivity 27 , 28 , while others have observed enhanced connectivity 29 . This heterogeneity may be related to the subjective selection of regions of interest (ROIs) or the adoption of different analysis strategies. Whole-brain-based multi-dimensional indicator analysis can reduce selection bias on a data-driven basis, but extensive computations and the integration of multiple types of indicators increase the difficulty of research. Radiomics, by converting imaging data into high-dimensional quantifiable features and integrating machine learning for screening and modeling, can efficiently process whole-brain and multimodal data, reduce subjective bias, and explore potential biomarkers 30 . This technique has been increasingly applied in the field of mental disorders, including schizophrenia and major depressive disorder, facilitating the identification of imaging biomarkers that assist in early diagnosis and mechanism exploration 31 – 34 . In the specific context of GD, preliminary efforts have also been made to leverage radiomics. For example, Han et al. constructed a GD classification model with accuracy of 73%, specificity of 68% and sensitivity of 77% by using T1-weighted image and diffusion tensor imaging (DTI) data with a random forest algorithm 35 ; Wang et al. also developed a GD radiomic model based on subcortical T1-weighted MRI features, achieving an AUC of 0.895 in the test cohort 36 . These results demonstrate the promising application of radiomics in the diagnosis of GD. However, most studies have demonstrated significant brain dysfunction in GD patients, and classification models constructed from structural MRI data alone are difficult to fully reflect their brain dysfunction. Notably, compared with structural MRI, which only captures morphological characteristics, fMRI has unique advantages in detecting dynamic functional activities of the brain. This is critical for revealing the core pathophysiological changes of GD, a disorder dominated by functional abnormalities. Using fMRI features can better capture these functional abnormalities, thereby enabling the development of more accurate auxiliary diagnostic models. Moreover, unlike traditional rs-fMRI analyses that mostly focus on single-dimensional indicators (e.g., amplitude of low-frequency fluctuations), the application of radiomics in multi-dimensional functional feature mining enables a more comprehensive characterization of brain functional alterations in GD. At present, there is a lack of studies that deeply mine the fMRI data of GD and uses multi-dimensional indicators to construct a diagnostic model for GD. Given the significant abnormalities in brain function observed in individuals with GD and the advantages of radiomics in processing whole-brain and multi-dimensional data, this study aims to utilize radiomics to comprehensively analyze rs-fMRI data from individuals with GD and normal subjects to obtain classification features and construct classification models. We hypothesize whether radiomics can yield brain imaging features conducive to identifying GD, and based on these features, construct an auxiliary diagnostic model for GD. This, in turn, will provide support for uncovering the neural mechanisms of GD and offering objective diagnostic evidence. Methods Participants This study recruited two groups of adolescents with GD and healthy controls from the psychiatric outpatient clinic of the Second Xiangya Hospital of Central South University and the public, respectively, from January 2024 to June 2025. Two professionally trained psychiatrists diagnosed GD after evaluating game use information and impaired social functioning described by the subjects themselves and their families according to the International Classification of Diseases 11th Revision (ICD-11). Meanwhile, psychiatric disorders such as mood disorders, schizophrenia, and substance use disorders were excluded via standardized assessment with the Mini International Neuropsychiatric Interview (MINI), according to ICD-11. Healthy controls were neighborhood residents recruited by the study advertisement. The specific inclusion and exclusion criteria were as follows: Inclusion criteria for the GD group: (1) meeting the diagnostic criteria of ICD-11 GD; (2) age 14–25 years, > 6 years of education, able to understand the study and complete it; and (3) right-handedness. Inclusion criteria for the healthy control group: (1) did not meet the diagnostic criteria for ICD-11 GD; (2) age, gender, dominant hand, and years of education matched those of the gaming disorder group, and (3) were able to understand the study and complete it. Common exclusion criteria for both groups: (1) suffering from other mental disorders; (2) having used psychotropic drugs or psychoactive substances within 6 months; (3) suffering from organic brain disorders or somatic disorders affecting hemodynamics; and (4) history of cranial and cerebral trauma, history of febrile coma, and contraindications to magnetic resonance scanning. Initially, this study recruited 147 patients with GD and 79 age- and sex-matched healthy controls. After MRI data preprocessing, 3 GD patients and 6 healthy controls were excluded due to excessive head motion, and another 3 GD patients were excluded due to obvious artifacts in the frontal lobe. Finally, 141 GD patients and 73 healthy controls were included for statistical analysis. Each subject read the contents of the study project carefully and signed the informed consent form. For subjects under the age of 18, both the subject and their legal guardian provided their signatures on the form. The study was reviewed and approved by the Ethics Committee of the Second Xiangya Hospital of Central South University (Ethical batch number: No. 2020004). All procedures in this study were performed in accordance with the Declaration of Helsinki and the relevant national and institutional guidelines and regulations. Data acquisition A self-administered questionnaire was used to collect socio-demographic information (age, sex, years of education) and gaming usage information (gaming expenditure[yuan/month], gaming time[hours/d]) from the participants in this study. Through interviews, participants’ gaming motivations were categorized into four types: seeking pleasure, escaping problems, relieving boredom, and maintaining relationships. The severity of GD was assessed using the DSM-5 structured interview. Imaging data were acquired using a 3.0T MRI scanner (uMR 790, United Imaging Healthcare, China) with 32 coil channels. Specific acquisition parameters were as follows: (1) Rs-fMRI: Synchronized multi-planar echo sequence was employed with TR = 700 ms, TE = 30 ms, flip angle = 52°, field of view (FOV) = 210 × 210 mm, matrix = 84 × 84, Voxel size = 2.50 × 2.50 × 2.50 mm 3 ; (2)T1-weighted images: TR/TE = 7.90/3.10 ms, flip angle = 10°, FOV = 230 × 256 mm, matrix = 460 × 512, voxel size = 1.00 × 1.00 × 1.00 mm 3 . The total scan duration of resting-state fMRI (rs-fMRI) was 7 min and 14 s, and the scan duration of T1-weighted structural MRI was 4 min. A total of 600 time points were acquired during the rs-fMRI scanning. During the scan, participants were in a supine position, with their eyes closed and kept awake; they wore earplugs to reduce noise interference, and head cushions were used to fix their heads to minimize motion artifacts. Rs-fMRI data preprocessing The rs-fMRI data were preprocessed using the GRETNA toolkit (v2.0) 37 : (1) Discarding the unstable data of the first 10 time points; (2) Conducting temporal correction and head motion correction (excluding participants with displacement > 2 mm or rotation > 2°, calculating the mean frame displacement (mean FD) for the remaining subjects); (3) Normalizing the T1-weighted images and rs-fMRI data to the Montreal Neurological Institute (MNI) space; (4) Using a 0.01–0.10 Hz band-pass filter to remove high-frequency physiological noise and low-frequency drift; (5) Regressing out covariates such as white matter signal and cerebrospinal fluid signals. Notably, global signal regression was not performed in this study. This decision was made to maximize the retention of spontaneous neural activity fluctuations associated with the core pathophysiological characteristics of GD, relying instead on the aforementioned combination of strict head motion control and covariate regression to ensure data reliability; (6) Eliminating linear drift. Calculation of Rs-fMRI metrics Three types of rs-fMRI metrics were extracted using DPABI software (v9.0) ( http://rfmri.org/dpabi ) 38 , as specified below: Regional Homogeneity (ReHo): Reflects the synchronization of neuronal activity in adjacent brain regions 39 . ReHo values was calculated using a 3 × 3 × 3 cubic voxel neighborhood centered on each individual voxel (a total of 27 voxels) in the whole brain; after segmentation and averaging using the Dosenbach160 (DOS160) brain template 40 (excluding the cerebellum), ReHo values for 142 ROIs were obtained. Amplitude of Low-Frequency Fluctuations (ALFF): Reflects the fluctuation intensity of spontaneous neuronal activity in the brain 41 . ALFF values were calculated for each voxel in the whole brain based on unfiltered data; after segmentation and averaging using the aforementioned template, ALFF values for 142 ROIs were obtained. FC: Reflects the correlation of neuronal activity between different brain regions 42 . After 142 ROIs were identified via segmentation using the aforementioned template, Pearson correlation analysis was performed on their time series, resulting in 10,011 FC values (142 × 141/2). To clarify the association between imaging features and GD, and to focus on the Limbic Network (LN) related to emotional regulation and reward processing in GD, this study did not use the DOS160 template for network partitioning. Instead, the selected features were matched to the Yeo seven-network system based on the MNI coordinates of brain regions for visualization 43 . Radiomics feature extraction and screening The original feature space constructed above includes 142 ReHo, 142 ALFF, and 10,011 FC features (total initial dimensionality: 10,295). Each feature in this high-dimensional space corresponds to one dimension, which not only leads to enormous computational load in subsequent modeling but also obscures key classification features and increases the risk of overfitting. Therefore, feature screening was performed to address these issues and identify key discriminative features, using Python’s scipy ( https://scipy.org/ ) and scikit-learn packages ( https://scikit-learn.org/ ) in three sequential stages: Independent samples t-test (Filter method): Select features with statistically significant inter-group differences ( P < 0.01), removing noise features without discriminative ability. Correlation-based redundancy elimination (Filter method): Calculate the Pearson correlation coefficients between all pairs of features. If the correlation coefficient of any two features is greater than 0.7, separately compute their respective mean correlation coefficients with all other features, and finally remove the feature with the higher mean correlation coefficient. Lasso regression (Embedded method): Perform further screening with group labels as the dependent variable. First, generate a candidate set of alpha values consisting of 50 logarithmically spaced points ranging from 10 − 4 to 10 2 to determine the optimal regularization parameter (alpha) for the model. Then, conduct 5-fold cross-validation to select the optimal alpha that minimizes the mean squared error (MSE) of the regression model. Construct the final Lasso model with this optimal alpha and retain features with non-zero regression coefficients. Model construction A Support Vector Machine (SVM) was used as the classifier, and an optimal classification model was constructed through 100 independent repeated experiments. To avoid sampling bias from a single random split and to comprehensively evaluate the model’s robustness under different data distributions, a sequentially varying random seed strategy (random_state = i, where i ranged from 0 to 99) was applied during these 100 experiments, ensuring both robustness and reproducibility. The specific workflow is as follows: In each experiment, the screened labeled classification features were split into a training set and a test set at a ratio of 70%:30% based on the specific random seed; on the training set, grid search combined with 5-fold cross-validation was adopted to optimize hyperparameters (the training set was divided into 5 mutually exclusive subsets—each time 4 subsets served as the training set and 1 subset as the validation set, and the optimal hyperparameters were selected after 5 cycles). For the grid search, the preliminary exploration was performed with C ∈ [10 − 5 ,10] (100 uniform candidates, np.linspace) and γ∈[10 − 5 ,10] (100 uniform candidates, np.linspace) with both radial basis function (RBF) and linear kernels included; subsequently, the RBF kernel was fixed for refined optimization, where C at was set to [0.5, 1.0] (30 logarithmic candidates, np.logspace, actual range: 3.16–10) and γ to [− 3.0,− 2.2] (30 logarithmic candidates, np.logspace, actual range: 0.001–0.006). The SVM model was fitted on the complete training set based on the optimal hyperparameters, and the test set was used to evaluate the model’s generalization ability. After 100 repeated experiments, the model with the highest test set accuracy was selected as the optimal model, and the mean and standard deviation of the model performance metrics (sensitivity, specificity, area under the receiver operating characteristic curve [AUC]) across all experiments were calculated. SHapley Additive exPlanations (SHAP) was used to analyze feature importance: A Kernel Explainer was constructed using the training set data, and the SHAP value of each feature in the test set was calculated to quantify the contribution of different features to the model’s classification results. Statistical analysis Post-hoc power analysis was performed using GPower 3.1 software to verify the statistical rationality of the sample size. SPSS 27.0 ( https://www.ibm.com/cn-zh/spss ) was used to analyze inter-group differences in clinical data (including sociodemographic information, gaming-related information, DSM-5 assessment results, etc.): The Kolmogorov-Smirnov test was applied to determine the normality of numerical data; independent samples t-tests were used for data conforming to a normal distribution, and chi-square tests were used for categorical data. The scipy and scikit-learn packages in Python 3.10 were used for correlation analysis and permutation tests: (1) For the feature with the highest SHAP value, Spearman correlation analysis was conducted with gaming motivations and scores of each DSM-5 item, while Pearson correlation analysis was performed with gaming expenditure, gaming time, and total DSM-5 scores; (2) Permutation test: Sample category labels were randomly shuffled, and the model was retrained. This process was repeated 10,000 times to construct the accuracy distribution of random models. The proportion of cases where the accuracy of the original model was higher than that of random models was calculated to determine the P -value. All statistical tests in this study were two-tailed, with a P -value < 0.05 considered statistically significant. Results Sociodemographic information A total of 214 subjects (including 141 GD and 73 healthy controls) were included in the current analyses. Post-hoc power analysis showed that for the intergroup comparison of core radiomic features between GD and HC groups, with a two-tailed α of 0.05, a moderate effect size (d = 0.5) and the actual sample size (141 GD patients, 73 HCs), the achieved statistical power was 0.932 (> 0.8). This indicated that the current sample size was sufficient to detect the true intergroup differences in radiomic features, meeting the statistical requirements for the construction of the classification model and ensuring the reliability of the model’s statistical foundation. Results of inter-group baseline data comparison showed that there were no statistically significant differences between the GD group and the healthy control group in terms of gender (χ 2 = 0.01, P = 0.95) and age (t = 1.67, P = 0.10). Statistically significant differences were observed in years of education (t = 4.22, P < 0.001), gaming time (t = − 5.52, P < 0.001), gaming expenditure (t = − 2.08, P = 0.04), and certain gaming motivations (seeking pleasure: χ 2 = 23.00, P < 0.001; escaping problems: χ 2 = 20.53, P < 0.001; relieving boredom: χ 2 = 21.04, P < 0.001). Specifically, the GD group had fewer years of education, longer gaming time, higher gaming expenditure, stronger motivations for seeking pleasure and escaping problems, and weaker motivations for relieving boredom. There was no statistically significant difference between the two groups in the “maintaining relationships” motivation (χ 2 = 1.31, P = 0.25). Detailed data are shown in Table 1 . Table 1. Demographic information, gaming usage information, and DSM-5 scores of the GD and healthy control group. Variables Healthy control group n = 73 Mean ± SD/n(%) Gaming disorder group n = 141 Mean ± SD/n(%) t/chique P Sex Male 68 (93.2%) 131 (92.9%) 0.004 0.95 Female 5 (6.8%) 10 (7.1%) Age 19.93 ± 2.42 19.20 ± 3.29 1.67 0.10 Years of education 14.30 ± 2.16 12.63 ± 3.01 4.22 < 0.001 Gaming time (hours/d) 4.07 ± 3.28 7.17 ± 4.17 − 5.52 < 0.001 Gaming expenditure (yuan/month) 71.96 ± 97.94 314.15 ± 991.14 − 2.08 0.04 Gaming motivations Seeking pleasure 30 (41.1%) 105 (74.5%) 23.00 < 0.001 Escaping problems 8 (11.0%) 58 (41.1%) 20.53 < 0.001 Relieving boredom 62 (84.9%) 75 (53.2%) 23.04 < 0.001 Maintaining relationships 16 (21.9%) 22 (15.6%) 1.31 0.25 DSM-5 total score 1.27 ± 1.25 6.72 ± 1.41 − 27.84 < 0.001 Open in a new tab Classification features Through a three-stage radiomics features screening, the initial 10,295 imaging features were screened, ultimately yielding 67 classification features, including 66 FC features (98.5%) and 1 ALFF feature (1.5%). These features are highly concentrated in typical brain functional networks, including the DMN, ECN, Ventral Attention Network, and SN. The corresponding brain regions and networks for each feature are detailed in Fig. 1 and Supplementary Table 1 (see Additional file 1 ). Fig. 1. Open in a new tab The classification features screened via radiomics and their corresponding brain networks. ( a ) Midline sagittal plane; ( b ) coronal plane; ( c ) axial plane; ( d ) brain regions and network connectivity. Performance of the classification model After optimizing hyperparameters using 5-fold cross-validation (embedded in grid search, with average validation accuracy as the evaluation index for parameter selection), the optimal parameters of the SVM classification model were as follows: the RBF was used as the kernel, with a penalty coefficient C = 8.53 and a kernel parameter γ = 2.43 ×  10 − 3 . The classification performance metrics of this model on the test set were as follows: accuracy of 0.86 ± 0.04 (range 0.78–0.95), sensitivity of 0.93 ± 0.04 (range 0.81–1.00), specificity of 0.75 ± 0.08 (range 0.53–0.91), and area under the receiver operating characteristic curve (AUC) of 0.92 ± 0.03 (range 0.86–0.99). Table 2 presents the performance of the classification model on the training set, validation set and test set. Table 2. Performance of the classification model on the training set, validation set and test set. Accuracy (mean ± SD [range]) Sensitivity (mean ± SD [range]) Specificity (mean ± SD [range]) AUC a (mean ± SD [range]) Training set 0.96 ± 0.03 (0.91-1.00) 0.98 ± 0.02 (0.89-1.00) 0.89 ± 0.06 (0.74-1.00) 0.99 ± 0.01 (0.97-1.00) Validation set 0.88 ± 0.02 (0.83–0.93) 0.98 ± 0.02 (0.89-1.00) 0.89 ± 0.06 (0.74-1.00) 0.99 ± 0.01 (0.97-1.00) Test set 0.86 ± 0.04 (0.78–0.95) 0.93 ± 0.04 (0.81-1.00) 0.75 ± 0.08 (0.53–0.91) 0.92 ± 0.03 (0.86–0.99) Open in a new tab a AUC: area under the receiver operating characteristic curve. Results of the permutation test showed that the average accuracy of the original model was significantly higher than the accuracy distribution of models obtained from 10,000 permutations ( P < 0.001), which indicates that the classification results were not obtained randomly. Detailed data are shown in Fig. 2 . Fig. 2. Open in a new tab The receiver operating characteristic curve (ROC curve) ( a ) and permutation test results of the gaming disorder classification model ( b ). Correlation between the top-SHAP-value feature and clinical indicators In this study, the feature with the highest SHAP value was identified as the FC between the right precuneus and the left ACC. Detailed data are shown in Fig. 3 . Correlation analysis was conducted between this FC value and gaming motivations, gaming expenditure, gaming time, individual DSM-5 items, and total DSM-5 score. The results showed that this FC value was positively correlated with DSM-5 Item 2 (withdrawal symptoms) ( r = 0.20, P = 0.004), DSM-5 Item 3 (tolerance symptoms) ( r = 0.14, P = 0.04), DSM-5 Item 6 (persistent engagement despite knowledge of harm) ( r = 0.16, P = 0.02), and DSM-5 Item 8 (use to escape problems or relieve negative emotions) ( r = 0.21, P = 0.002). It was also positively correlated with the total DSM-5 diagnostic score ( r = 0.18, P = 0.01). Fig. 3. Open in a new tab Results of SHAP interpretation ( a ) and the correlation between the feature with the largest SHAP value and DSM-5 scores ( b ). Discussion This study employed an rs-fMRI radiomics approach to develop a machine learning model that can effectively distinguish between adolescents with GD and healthy controls. Based on the 67 screened brain functional features (primarily FC), the model exhibited strong classification performance on the test set (accuracy of 86%, AUC of 0.92) and passed the permutation test ( P < 0.001), confirming that the classification results were not obtained randomly. SHAP interpretation showed that the FC between the right precuneus and left ACC made the greatest contribution to classification. More importantly, the strength of this connectivity was significantly positively correlated with the severity of GD symptoms as assessed by the DSM-5. This suggests that this FC not only has diagnostic discriminative potential but may also reflect the clinical severity of this disorder. In this study, a radiomics-based auxiliary diagnostic model for GD was constructed using rs-fMRI features, achieving an accuracy of 86% (AUC = 0.92) in GD identification. Previous studies have made advances in developing discriminative models between GD and healthy people by integrating structural or single-dimensional fMRI features, but their performance is limited. Structural features-based models (e.g., T1-weighted imaging gray matter and diffusion tensor imaging white matter features) only reached 73% accuracy 35 , and subcortical structural feature-based models required additional clinical variables (e.g., gender, internet addiction scores) to achieve satisfactory efficacy (C-index = 0.987) 36 . Single-dimensional fMRI feature-based models (e.g., FC or ReHo alone) also showed limited performance (78.8% and 82.7% accuracy, respectively) 44 , 45 , due to the subjective selection of single indicators that fail to comprehensively reflect brain activity and network connectivity abnormalities in GD, compromising model classification performance and generalization ability. In contrast, this study directly utilized multi-dimensional rs-fMRI features and adopted radiomics methods to systematically and unbiasedly screen the most discriminative quantitative features from massive imaging features. In addition, permutation test this research used is well-suited for the heterogeneous rs-fMRI functional connectivity data and limited adolescent sample size, and when combined with 5-fold cross-validation for hyperparameter optimization, 10,000 permutations robustly verify that the model’s performance derives from genuine feature-label correlations rather than random chance. The finally constructed model exhibits superior stability and generalization ability, offering a more reliable technical support for the auxiliary diagnosis of GD. Further network attribution analysis revealed that the above-mentioned 67 features do not exist in isolation, but are highly concentrated in typical brain functional networks, including the DMN, ECN, SN. These features also involve core nodes of the SN, such as the ACC and insula 43 , 46 . These findings point to functional impairments within the classic “triple-network model” in addiction neuroscience 21 , 47 . Specifically, the core function of the DMN is to support self-referential processing, internal thinking, and memory retrieval 48 , 49 . In this study, DMN-related features (e.g., abnormal connectivity between the precuneus and ventromedial prefrontal cortex) were highly consistent with the clinical manifestations of GD: excessive immersion in gaming-related internal activities (e.g., repeatedly reliving gaming experiences) 50 . This suggests that DMN dysfunction may be the neural basis for patients’ difficulty in escaping “gaming-related mental representations”. In contrast, the ECN is primarily responsible for cognitive control, impulse inhibition, and goal-directed decision-making 51 . The abnormal connectivity of nodes such as the dorsolateral prefrontal cortex and dorsomedial prefrontal cortex in this study reveals that GD patients have reduced ability to regulate their gaming behaviors, thus failing to effectively inhibit the urge to engage in gaming excessively 52 , 53 . The core nodes of the SN—the ACC and insula—are also key features identified in this screening. The main role of this network is to identify and prioritize the processing of internal and external “salient information" 54 . However, abnormal SN-related FC in GD may lead to excessive attentional bias and emotional responses to gaming cues, while reducing sensitivity to non-gaming information in real life 55 , ultimately giving gaming-related behaviors much higher priority than other social functions. In summary, the classification features identified in this study collectively depict a picture of extensive dysfunction in brain functional networks among GD patients. In particular, the balance among the DMN, ECN, and SN is disrupted, which supports the hypothesis of abnormal network function in GD. In the model constructed in this study, the FC between the precuneus and the ACC played a crucial role in the model’s classification performance, indicating that precuneus-ACC FC is a key factor in distinguishing between GD patients and healthy controls. The precuneus is a core component of the DMN and is closely associated with cognitive functions such as self-awareness, episodic memory retrieval, and mental imagery 48 , 56 . In the context of GD, precuneus activity may be closely linked to individuals’ intense gaming cravings, repeated mental rehearsal of gaming scenarios, and adoption of the “player-avatar” as part of their identity 57 – 60 . Separately, the ACC—acting as the intersection of the ECN and the SN—plays a core role in conflict monitoring, emotional regulation, and impulse control 61 , 62 . Impairments in ACC function are a repeatedly reported phenomenon in many addictive behaviors, and are particularly closely associated with difficulties in inhibiting addictive impulses 63 . Therefore, abnormalities in the precuneus-ACC FC may reveal the core contradiction of GD patients at the neural level: “an intense gaming desire and intrinsic motivation originating from the precuneus” vs. “a gradual impairment of cognitive control and emotional regulation mediated by the ACC”. This FC dysfunction further embodies the deep-seated internal conflict in patients: the imbalance and conflict between the urge to “play” and the rational control to “stop" 64 , 65 . Previous studies have also reported abnormalities in precuneus-ACC FC in patients with GD, and this alteration has been found to be more pronounced in GD patients with comorbid depression 66 . In addition, studies on other substance addictions have also demonstrated the presence of abnormal precuneus-ACC FC in heroin, areca nut, and nicotine addiction 67 – 69 . Among these, one study further indicated that this FC abnormality is associated with nicotine dependence and craving levels 69 . This suggests that the precuneus-anterior cingulate cortex circuit may possess addiction-related commonality, thereby providing important clues for exploring the universality of addiction mechanisms and identifying cross-addiction intervention targets. This study found that precuneus-ACC FC was positively correlated with the total DSM-5 score and scores of specific items (e.g., withdrawal, tolerance, loss of control, and escape from negative emotions through gaming). This result indicates this metric may not only help distinguish between pathological and non-pathological gaming behaviors but also show potential in reflecting disease severity. First, the association between precuneus-ACC circuit connectivity and “withdrawal” or “escape from negative emotions” confirms the key role of this neural circuit in emotional regulation disorders 66 , 70 . Studies have shown that functional deficits of the ACC in emotional monitoring and regulation may lead individuals to escape unmanageable negative emotions via gaming 71 , 72 . Meanwhile, its positive correlation with “tolerance” and “loss of control” further highlights the connectivity’s critical role in abnormal reward mechanisms and executive function deficits. A core feature of GD is abnormal sensitivity of the reward system, along with impaired impulse control and self-regulation 73 , 74 . This finding aligns with neurobiological models of addiction, particularly supporting the “reward-driven and control-deficit” dual-pathway mechanism 64 . Thus, precuneus-ACC FC may not only serve as a biomarker for distinguishing affected from unaffected individuals but also help uncover functional deficits in addictive behaviors, thereby enhancing understanding of the neural mechanisms underlying GD. Based on these findings, future studies could explore whether this connectivity can act as a biomarker for different mental disorders, offering a new perspective for multidimensional symptom assessment of GD. While this study yielded valuable findings, it has several limitations. First, the cross-sectional design precludes inference of a causal relationship between abnormal brain function and GD; future longitudinal studies could track high-risk populations to clarify whether these brain function features are risk factors for GD or consequences of the disorder. Second, the sample was sourced from a single center, and the proposed auxiliary diagnostic model was evaluated solely using internal cross-validation. The lack of an independent external validation cohort may limit the generalizability and robustness of the model when applied to data acquired from different scanners or diverse populations. Validating the model across multiple centers remains a critical next step. Third, the study sample has an unbalanced gender ratio (males account for approximately 93%), which may limit the model’s generalization performance in female populations; future studies should expand the female sample size and construct a gender-balanced sample cohort to improve the model’s universality. Fourth, although common psychiatric comorbidities were excluded, this may also mean the sample does not fully represent the complex comorbid profiles frequently seen in clinical practice. Finally, this study focused primarily on resting-state brain function; future research could integrate task-state fMRI (e.g., gaming cue exposure tasks) or multimodal imaging data (e.g., DTI) to more comprehensively unravel the neural mechanisms of GD. Addressing these limitations will be a key direction for future research, facilitating the development of more reliable and clinically useful neurobiological biomarkers for GD. Conclusion In this study, the classification model constructed using rs-fMRI radiomics and machine learning methods can effectively distinguish adolescent GD patients from healthy controls (accuracy: 86%, AUC: 0.92, p < 0.001). The FC between the precuneus and ACC plays a dominant role in the model and is positively correlated with the severity of GD symptoms, suggesting its potential value in diagnostic discrimination and assessment of disease severity. The results at the network level support the key role of functional imbalance among the DMN, ECN, and SN in the pathogenesis of GD. This study innovatively adopted whole-brain multi-dimensional rs-fMRI features combined with radiomics methods to screen the optimal discriminative features, and the constructed pure imaging diagnostic model has both high accuracy and stability, potentially assist in GD identification. Moreover, it further suggests that precuneus-ACC FC might serve as a promising candidate biomarker, providing a potential neural indicator for simultaneous GD diagnosis and disease severity assessment. Although further validation is required through longitudinal, multi-center, and multi-modal studies, this research lays a preliminary foundation for the development of objective biomarkers based on brain functional imaging and for an in-depth understanding of the neural mechanisms underlying GD. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 (30.9KB, docx) Acknowledgements The authors would like to thank all the participants for their support for our study. Author contributions YF L, SH L, HX S, R Y, and QP H designed the study. YF L, SH L, Y T, ZJ L, XX C, JY H, CH W, and XH W collected data. YF L, SH L, Y T, and CH W performed statistical analysis. YF L, SH L, HX S, and R Y interpreted data. YF L and SH L prepared the manuscript. QP H, R Y, and HX S revised the manuscript. HX S, R Y, and QP H supervised the study. HX S and QP H acquired funding. All authors had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Funding This work was supported by Chinese National Programs for Brain Science and Brain-like Intelligence Technology [No.2021ZD0202102], National Natural Science Foundation of China [No.82501793]. All funders had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript, or the decision to submit the paper for publication. Data availability The datasets generated and analysed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yifan Li and Shuhong Lin contributed equally to this work. Contributor Information Ru Yang, Email: yr_smu@126.com. Hongxian Shen, Email: shenhx2018@csu.edu.cn. References 1. Chan, G., Huo, Y., Kelly, S., Leung, J. & Tisdale, C. The impact of eSports and online video gaming on lifestyle behaviours in youth: A systematic review. Comput. Hum. Behav. 126 , 106974 (2022). [ Google Scholar ] 2. Turel, O., Romashkin, A. & Morrison, K. M. Health outcomes of information system use lifestyles among adolescents: Videogame addiction, sleep curtailment and cardio-metabolic deficiencies. PloS One . 11 , e0154764 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. King, D. L. & Delfabbro, P. H. The concept of ‘harm’ in Internet gaming disorder. J. Behav. Addict. 7 , 562–564 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Delfabbro, P., King, D. L. & Carey, P. Harm severity in internet gaming disorder and problem gambling: A comparative study. Comput. Hum. Behav. 124 , 106898 (2021). [ Google Scholar ] 5. Kim, H. S. et al. Prevalence of gaming disorder: A meta-analysis. Addict. Behav. 126 , 107183 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 6. Liao, Z., Chen, X., Huang, Q. & Shen, H. Prevalence of gaming disorder in East Asia: A comprehensive meta-analysis. J. Behav. Addict 11 , 727–738 (2022). [ DOI ] [ PMC free article ] [ PubMed ] 7. Han, T. S., Cho, H., Sung, D. & Park, M. H. A systematic review of the impact of COVID-19 on the game addiction of children and adolescents. Front. Psychiatry . 13 , 976601 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Putra, P. Y., Fithriyah, I. & Zahra, Z. Internet addiction and online gaming disorder in children and adolescents during COVID-19 pandemic: A systematic review. Psychiatry Investig . 20 , 196–204 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Galanis, C. R., Weber, N., Delfabbro, P. H., Billieux, J. & King, D. L. Gaming disorder and stigma-related judgements of gaming individuals: An online randomized controlled trial. Addiction 118 , 1687–1698 (2023). [ DOI ] [ PubMed ] 10. Casale S. et al. Stigmatizing attitudes toward Internet gaming disorder, Problematic smartphone use and Problematic social networking site use: An experimental vignette study. Addict Behav 141 , 107665 (2023). [ DOI ] [ PubMed ] 11. American Psychiatric Association DSM- Task Force Arlington VA US. Diagnostic and statistical manual of mental disorders: DSM-5 ™ (5th ed.). Codas 25 , 191 (2013). [ DOI ] [ PubMed ] [ Google Scholar ] 12. Shao, T. et al. The recognition of gaming disorder in China: A case series of 223 patients. PeerJ 9 , e10827 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. León Méndez, M., Padrón, I., Fumero, A. & Marrero, R. J. Effects of internet and smartphone addiction on cognitive control in adolescents and young adults: A systematic review of fMRI studies. Neurosci. Biobehav. Rev. 159 , 105572 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 14. Zhang, Z. et al. Altered dynamic reconfiguration of brain functional networks during gaming and deprivation in individuals with internet gaming disorder. J. Behav. Addict. 14 , 776–792 (2025). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Wang, X. et al. Resting-state functional connectivity alterations in intermet gaming disorder: A fMRI study combining voxel-based morphometry meta-analysis. Brain Res. Bull. 232 , 111588 (2025). [ DOI ] [ PubMed ] [ Google Scholar ] 16. Zhang, Z. et al. Brain responses to positive and negative events in individuals with internet gaming disorder during real gaming. J. Behav. Addict. 12 , 758–774 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Dong, H., Zheng, H., Wang, M., Ye, S. & Dong, G. H. The unbalanced behavioral activation and inhibition system sensitivity in internet gaming disorder: Evidence from resting-state Granger causal connectivity analysis. Prog Neuropsychopharmacol. Biol. Psychiatry . 119 , 110582 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 18. Zhang, J. T. et al. Altered resting-state functional connectivity of the insula in young adults with Internet gaming disorder. Addict. Biol. 21 , 743–751 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Wang, Z. L. et al. Subgroups of internet gaming disorder based on addiction-related resting-state functional connectivity. Addiction 118 , 327–339 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 20. Yan, H., Li, Q., Yu, K. & Zhao, G. Large-scale network dysfunction in youths with Internet gaming disorder: A meta-analysis of resting-state functional connectivity studies. Prog Neuropsychopharmacol. Biol. Psychiatry . 109 , 110242 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 21. Zhang, J. T. et al. Altered coupling of default-mode, executive-control and salience networks in Internet gaming disorder. Eur. Psychiatry J. Assoc. Eur. Psychiatry 45 , 114–120 (2017). [ DOI ] [ PubMed ] [ Google Scholar ] 22. Wang, L. et al. Deficient dynamics of prefrontal-striatal and striatal-default mode network (DMN) neural circuits in internet gaming disorder. J. Affect. Disord . 323 , 336–344 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 23. Niu, X. et al. Meta-analysis of structural and functional brain alterations in internet gaming disorder. Front. Psychiatry . 13 , 1029344 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Zhang, J. T. et al. Effects of craving behavioral intervention on neural substrates of cue-induced craving in Internet gaming disorder. NeuroImage Clin. 12 , 591–599 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Zhou, W. et al. Sex difference in neural responses to gaming cues in Internet gaming disorder: Implications for why males are more vulnerable to cue-induced cravings than females. Neurosci. Lett. 760 , 136001 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 26. Zhang, J. et al. Altered brain activities associated with cue reactivity during forced break in subjects with internet gaming disorder. Addict. Behav. 102 , 106203 (2020). [ DOI ] [ PubMed ] [ Google Scholar ] 27. Bae, S., Han, D. H., Jung, J., Nam, K. C. & Renshaw, P. F. Comparison of brain connectivity between internet gambling disorder and Internet gaming disorder: A preliminary study. J. Behav. Addict. 6 , 505–515 (2017). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Yan, H., Li, Q., Yu, K. & Zhao, G. Large-scale network dysfunction in youths with Internet gaming disorder: A meta-analysis of resting-state functional connectivity studies. Prog Neuropsychopharmacol. Biol. Psychiatry . 109 , 110242 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 29. Lee, D., Lee, J., Lee, J. E. & Jung, Y. C. Altered functional connectivity in default mode network in Internet gaming disorder: Influence of childhood ADHD. Prog. Neuropsychopharmacol. Biol. Psychiatry . 75 , 135–141 (2017). [ DOI ] [ PubMed ] [ Google Scholar ] 30. Mayerhoefer, M. E. et al. Introduction to radiomics. J. Nucl. Med. 61 , 488–495 (2020). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Cui, L. B. et al. Thalamus radiomics-based disease identification and prediction of early treatment response for schizophrenia. Front. Neurosci. 15 , 682777 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Sun, K. et al. A novel MSN-II feature extracted from T1-weighted MRI for discriminating between BD patients and MDD patients. J. Affect. Disord . 371 , 36–44 (2025). [ DOI ] [ PubMed ] [ Google Scholar ] 33. Bang, M. et al. Identification of schizophrenia by applying interpretable radiomics modeling with structural magnetic resonance imaging of the cerebellum. Psychiatry Clin. Neurosci. 78 , 527–535 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 34. Schiulaz, A. et al. Radiomic analysis of striatal [18F]FDOPA PET imaging in patients with psychosis for the identification of antipsychotic response. Mol. Imaging Biol. 27 , 365–378 (2025). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Han, X. et al. MRI-based radiomic machine-learning model may accurately distinguish between subjects with internet gaming disorder and healthy controls. Brain Sci. 12 , 44 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Wang, L. et al. Identification of patients with internet gaming disorder via a radiomics-based machine learning model of subcortical structures in high-resolution T1-weighted MRI. Prog Neuropsychopharmacol. Biol. Psychiatry . 133 , 111026 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 37. Wang, J. et al. GRETNA: a graph theoretical network analysis toolbox for imaging connectomics. Front. Hum. Neurosci. 9 , 386 (2015). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Gao, X. et al. Structural and functional brain abnormalities in internet gaming disorder and attention-deficit/hyperactivity disorder: A comparative meta-analysis. Front. Psychiatry . 12 , 679437 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Han, X. et al. Alterations of resting-state static and dynamic functional connectivity of the dorsolateral prefrontal cortex in subjects with internet gaming disorder. Front. Hum. Neurosci. 12 , 41 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Dosenbach, N. U. F. et al. Prediction of individual brain maturity using fMRI. Science 329 , 1358–1361 (2010). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Du, Y., Jiang, W. & Vance, A. Longer term effect of randomized, controlled group cognitive behavioural therapy for Internet addiction in adolescent students in Shanghai. Aust N Z. J. Psychiatry . 44 , 129–134 (2010). [ DOI ] [ PubMed ] [ Google Scholar ] 42. Kuczenski, R. & Segal, D. S. Caudate-putamen and nucleus accumbens extracellular acetylcholine responses to methamphetamine binges. Brain Res. 923 , 32–38 (2001). [ DOI ] [ PubMed ] [ Google Scholar ] 43. Yeo, B. T. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J. Neurophysiol. 106 , 1125–1165 (2011). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Song, K. R. et al. Resting-state connectome-based support-vector-machine predictive modeling of internet gaming disorder. Addict. Biol. 26 , e12969 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 45. Wang, Z., Dong, H., Du, X., Zhang, J. T. & Dong, G. H. Decreased effective connection from the parahippocampal gyrus to the prefrontal cortex in Internet gaming disorder: A MVPA and spDCM study. J Behav. Addict 9 , 105–115 (2020). [ DOI ] [ PMC free article ] [ PubMed ] 46. Menon, V. Large-scale brain networks and psychopathology: A unifying triple network model. Trends Cogn. Sci. 15 , 483–506 (2011). [ DOI ] [ PubMed ] [ Google Scholar ] 47. Sutherland, M. T., McHugh, M. J., Pariyadath, V. & Stein, E. A. Resting state functional connectivity in addiction: Lessons learned and a road ahead. NeuroImage 62 , 2281–2295 (2012). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Buckner, R. L., Andrews-Hanna, J. R. & Schacter, D. L. The brain’s default network: Anatomy, function, and relevance to disease. Ann. N Y Acad. Sci. 1124 , 1–38 (2008). [ DOI ] [ PubMed ] [ Google Scholar ] 49. Raichle, M. E. The brain’s default mode network. Annu. Rev. Neurosci. 38 , 433–447 (2015). [ DOI ] [ PubMed ] [ Google Scholar ] 50. Guglielmucci, F. et al. Dissociation in problematic gaming: A systematic review. Curr. Addict. Rep. 6 , 1–14 (2019). [ Google Scholar ] 51. Seeley, W. W. et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J. Neurosci. 27 , 2349–2356 (2007). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Kim, S. J. et al. Differences in resting-state functional connectivity according to the level of impulsiveness in patients with internet gaming disorder. J. Behav. Addict. 10 , 88–98 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 53. Zhang, Z., Zheng, H., Zhou, W. & Dong, G. H. Brain responses to decision-making in easy and hard choices in internet gaming disorder: Implications for irrepressible gaming behaviours. J. Psychiatr Res. 165 , 233–240 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 54. Menon, V. & Uddin, L. Q. Saliency, switching, attention and control: A network model of insula function. Brain Struct. Funct. 214 , 655–667 (2010). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 55. Wang, Z. et al. Altered brain functional networks in Internet gaming disorder: Independent component and graph theoretical analysis under a probability discounting task. CNS Spectr. 24 , 544–556 (2019). [ DOI ] [ PubMed ] [ Google Scholar ] 56. Cavanna, A. E. & Trimble, M. R. The precuneus: A review of its functional anatomy and behavioural correlates. Brain J. Neurol. 129 , 564–583 (2006). [ DOI ] [ PubMed ] [ Google Scholar ] 57. Dong, G. H. et al. Addiction severity modulates the precuneus involvement in internet gaming disorder: Functionality, morphology and effective connectivity. Prog Neuropsychopharmacol. Biol. Psychiatry . 98 , 109829 (2020). [ DOI ] [ PubMed ] [ Google Scholar ] 58. Ko, C. H. et al. Brain correlates of craving for online gaming under cue exposure in subjects with Internet gaming addiction and in remitted subjects. Addict. Biol. 18 , 559–569 (2013). [ DOI ] [ PubMed ] [ Google Scholar ] 59. Leménager, T. et al. Exploring the neural basis of avatar identification in pathological internet gamers and of self-reflection in pathological social network users. J. Behav. Addict. 5 , 485–499 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 60. Green, R., Delfabbro, P. H. & King, D. L. Avatar- and self-related processes and problematic gaming: A systematic review. Addict. Behav. 108 , 106461 (2020). [ DOI ] [ PubMed ] [ Google Scholar ] 61. Bush, G., Luu, P. & Posner, M. I. Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn. Sci. 4 , 215–222 (2000). [ DOI ] [ PubMed ] [ Google Scholar ] 62. Shackman, A. J. et al. The integration of negative affect, pain and cognitive control in the cingulate cortex. Nat. Rev. Neurosci. 12 , 154–167 (2011). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 63. Zhao, Y. et al. Anterior cingulate cortex in addiction: New insights for neuromodulation. Neuromodulation J. Int. Neuromodulation Soc. 10.1111/ner.13291 (2020). [ DOI ] [ PubMed ] [ Google Scholar ] 64. Wang, M. et al. Disrupted dynamic network reconfiguration of the executive and reward networks in internet gaming disorder. Psychol. Med. 53 , 5478–5487 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 65. Zhang, J. et al. Disturbed craving regulation to gaming cues in internet gaming disorder: Implications for uncontrolled gaming behaviors. J. Psychiatr Res. 140 , 250–259 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 66. Lee, D., Lee, J., Namkoong, K. & Jung, Y. C. Subregions of the anterior cingulate cortex form distinct functional connectivity patterns in young males with internet gaming disorder with comorbid depression. Front. Psychiatry . 9 , 380 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 67. Liu, T. et al. Betel quid dependence is associated with functional connectivity changes of the anterior cingulate cortex: a resting-state fMRI study. J. Transl Med. 14 , 33 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 68. Wang, W. et al. Changes in functional connectivity of ventral anterior cingulate cortex in heroin abusers. Chin. Med. J. (Engl) . 123 , 1582–1588 (2010). [ PubMed ] [ Google Scholar ] 69. Huang, W. et al. The development and expression of physical nicotine dependence corresponds to structural and functional alterations in the anterior cingulate-precuneus pathway. Brain Behav. 4 , 408–417 (2014). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 70. Lai, C. H. The regional homogeneity of cingulate-precuneus regions: The putative biomarker for depression and anxiety. J. Affect. Disord . 229 , 171–176 (2018). [ DOI ] [ PubMed ] [ Google Scholar ] 71. Lee, D., Park, J., Namkoong, K., Kim, I. Y. & Jung, Y. C. Gray matter differences in the anterior cingulate and orbitofrontal cortex of young adults with Internet gaming disorder: Surface-based morphometry. J. Behav. Addict. 7 , 21–30 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 72. Lee, J. et al. Compromised prefrontal cognitive control over emotional interference in adolescents with internet gaming disorder. Cyberpsychol. Behav. Soc. Netw. 18 , 661–668 (2015). [ DOI ] [ PubMed ] [ Google Scholar ] 73. Zhou, W. R. et al. Imbalanced sensitivities to primary and secondary rewards in internet gaming disorder. J. Behav. Addict. 10 , 990–1004 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 74. Dong, G. H. et al. Disrupted prefrontal regulation of striatum-related craving in Internet gaming disorder revealed by dynamic causal modeling: Results from a cue-reactivity task. Psychol. Med. 51 , 1549–1561 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Material 1 (30.9KB, docx) Data Availability Statement The datasets generated and analysed during the current study are available from the corresponding author on reasonable request. Articles from Scientific Reports are provided here courtesy of Nature Publishing Group ACTIONS View on publisher site PDF (2.1 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• A transdiagnostic computational account linking mental fatigue and cognitive effort#1180033
• A transdiagnostic computational account linking mental fatigue and cognitive effort#1179167
• Evaluation of Decision-Making Fatigue After Traumatic Brain Injury#1170472
• Radiomics for risk stratification in brain metastases: potential in melanoma brain metastases.#440778
• Causal relationship between inflammatory proteins and attention deficit hyperactivity disorder: A serum-metabolites-mediated Mendelian randomization analysis.#250965