91PORN

A collaborative research team led by Professor Jung-Seok Choi from the Department of Psychiatry at Samsung Medical Center, Sungkyunkwan University, and Professor Woo-Young Ahn from the Department of Psychology at Seoul National University has identified the shared and distinct psychological and neurophysiological characteristics of IGD and AUD using artificial intelligence techniques. This study was recently published in the Comprehensive Psychiatry, and the research was conducted by Ji-Yoon Lee (first author, Department of Healthcare and Convergence Science, Seoul National University) and Myeong Seop Song (co-first author, Department of Psychology, Seoul National University), among others.

Substance use disorder typically involves the repeated use of substances that directly affect the body—most notably alcohol and drugs. In recent years, however, excessive engagement in certain behaviors such as gambling, gaming, and shopping has gained attention as another form of addiction, referred to as behavioral addiction. While the triggers may differ—substances versus behaviors—numerous studies have reported that substance use disorder and behavioral addiction share similarities in terms of clinical symptoms, disease progression, genetic underpinnings, and neural abnormalities.

Based on these similarities, the 11th revision of the International Classification of Diseases (ICD-11) officially included "disorders due to addictive behaviors", and in 2018, both gambling disorder and gaming disorder were recognized as formal diagnoses by the World Health Organization. However, despite this recognition, the neurological basis of behavioral addictions remains insufficiently understood, and there is ongoing debate as to whether they should be considered brain disorders on par with substance addictions. This underscores the need to elucidate the neural mechanisms underlying behavioral addiction and determine how they overlap or diverge from those associated with substance addiction.

In South Korea, one of the most prevalent behavioral addictions is IGD characterized by excessive and persistent use of online games. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) includes IGD as a condition warranting further study. Previous research has indicated that individuals with IGD often exhibit psychological symptoms such as depression, anxiety, and impulsivity, similar to those with AUD. However, the specific neural mechanisms that distinguish or connect these two disorders have not been fully clarified. Therefore, the present study aimed to compare the neurophysiological and psychological characteristics of IGD and AUD using a multimodal machine learning framework, integrating both EEG data and neuropsychological features.

We analyzed both the neurophysiological and psychological characteristics of IGD and AUD using artificial intelligence models applied to multimodal data—resting-state EEG signals recorded with eyes closed, and standardized psychological assessments. A total of 191 participants were included in the study: 67 individuals with IGD, 58 with AUD, and 66 healthy controls. From the EEG data, the researchers extracted both sensor-level (channel-based) and source-level (brain-region-based) connectivity features. In parallel, they collected psychological data, including measures of depression, anxiety, impulsivity, and intelligence quotient (IQ). Three machine learning algorithms were used for classification: L1-norm logistic regression, support vector machines, and random forest. We compared models trained using EEG data alone, psychological data alone, and a multimodal model that integrated both.

Figure 1. Multimodal analysis framework

The multimodal L1-norm logistic regression model achieved the highest performance in distinguishing IGD from AUD, with a classification accuracy of 71.2%—surpassing the models with neuropsychogical or EEG data. Notably, the results revealed that connectivity differences in delta and beta frequency bands—particularly within the right orbitofrontal cortex, prefrontal cortex, temporal lobe, and anterior cingulate cortex—played a key role in distinguishing the two disorders. These regions are associated with reward processing and cognitive control, suggesting distinct patterns of neural dysfunction between IGD and AUD. In contrast, psychological features such as depression, anxiety, and impulsivity did not significantly differ between the two groups, highlighting that while IGD and AUD may share similar psychological profiles and exhibit distinct neurophysiological patterns.  

Figure 2. Feature importance according to beta coefficients: comparison between IGD and AUD

This study is the first to compare behavioral addiction and substance use disorder using both non-invasive and cost-effective EEG data and neuropsychological assessments. It offers a potential technical foundation for the early diagnosis, personalized treatment, and potential development of digital therapeutics for addiction-related disorders. Moreover, the multimodal machine learning approach achieved high classification performance and shows great potential for broader application in the diagnosis and prognosis of various psychiatric conditions.