Altered Functional Brain Connectivity in Gaming Disorder: Insights from EEG and Machine Learning

Gaming disorder is increasingly recognized as a behavioral disorder involving impaired cognitive control and altered brain connectivity. This study investigates its neural mechanisms using electroencephalogram (EEG)-based classification and connectivity analysis. Participants identified as addicted or non-addicted were exposed to game-related audiovisual stimuli during EEG recording. A convolutional neural network-long short-term memory model integrating spatial and temporal features achieved the highest classification accuracy. Specifically, the classification accuracy between addicted and non-addicted groups exceeded 99%, remarkably distinguishing the two populations. Furthermore, the classification accuracy for the resting state was also exceptionally high in both groups, demonstrating that significant group differences were evident even at baseline. The model particularly leveraged high-frequency neural responses for classification. Functional connectivity analysis revealed that Pearson correlation coefficient features outperformed Phase-locked value features in classification tasks. Analysis revealed suppressed, localized brain activations in addicted individuals, mainly in the occipital region, while non-addicted individuals exhibited broader, long-range connectivity. Confusion matrix analysis further confirmed clear group separability and indicated different sensitivities to stimulus types. These findings suggest that gaming disorder is linked to disrupted brain network integration and altered visual attention, providing insight into its underlying neural mechanisms.