Resting-state connectome-based support-vector-machine predictive modeling of internet gaming disorder

Area of Science:

• Neuroimaging
• Psychiatry
• Machine Learning

Background:

• Internet gaming disorder (IGD) is a global mental health concern.
• Previous studies show resting-state functional connectivity (rsFC) alterations in IGD.
• The link between abnormal connectivity and behavioral measures in IGD requires further investigation.

Purpose of the Study:

• To identify resting-state functional connections associated with Internet gaming disorder (IGD).
• To investigate the utility of connectome-based predictive modeling (CPM) for understanding IGD neural mechanisms.
• To explore the role of the default-mode network (DMN) in predicting IGD.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (fMRI) data from 72 individuals with IGD and 41 healthy controls.
• Adapted the CPM approach using a support vector machine for classification and regression analyses.
• Compared whole-brain and network-based analyses to pinpoint informative neural networks.

Main Results:

• The default-mode network (DMN) was the most informative network for predicting IGD.
• Achieved 78.76% accuracy in classifying individuals with IGD.
• Found a significant correlation (r=0.44, P<0.001) between predicted and actual psychometric scale scores for IGD severity.

Conclusions:

• Individual differences in DMN resting-state function are crucial for understanding IGD.
• The findings advance the characterization of aberrant DMN activity in IGD.
• This research provides a foundation for understanding IGD etiology and intervention outcomes.