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

Area of Science:

• Neuroimaging
• Computational Psychiatry
• Medical Diagnostics

Background:

• Clinical assessments for Gaming Disorder (GD) suffer from subjective biases, necessitating objective biomarkers.
• Radiomics offers a potential solution for whole-brain analysis and identifying multidimensional indicators for GD diagnosis.

Purpose of the Study:

• To identify radiomic classification features for Gaming Disorder (GD).
• To develop an auxiliary diagnostic model for GD using radiomics.
• To explore the neural mechanisms underlying GD.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (fMRI) in 141 individuals with GD and 73 healthy controls.
• Applied radiomics for feature extraction and selection, constructing a classification model with a support vector machine.
• Employed permutation testing to validate model performance.

Main Results:

• Developed a classification model incorporating 67 brain functional features with 86% accuracy, 93% sensitivity, 75% specificity, and an AUC of 0.92.
• Identified key features concentrated in the Default Mode Network, Executive Control Network, and Salience Network.
• Found that right precuneus-left anterior cingulate cortex connectivity was a primary contributor and correlated significantly with GD symptom severity.

Conclusions:

• Radiomics presents a promising framework for identifying objective features of GD.
• The developed model can assist in the objective diagnosis of GD.
• Radiomics aids in unraveling the neural mechanisms associated with GD.