Abstract
Social anxiety is a symptom widely prevalent among young adults, and when present in excess, can lead to maladaptive patterns of social behavior. Recent approaches that incorporate brain functional radiomic features and machine learning have shown potential for predicting certain phenotypes or disorders from functional magnetic resonance images. In this study, we aimed to predict the level of social anxiety in young adult participants by training machine learning models with resting-state brain functional radiomic features including the regional homogeneity, fractional amplitude of low-frequency fluctuation, fractional resting-state physiological fluctuation amplitude, and degree centrality. Among the machine learning models, the XGBoost model achieved the best performance with balanced accuracy of 77.7% and F1 score of 0.815. Analysis of input feature importance demonstrated that the orbitofrontal cortex and the degree centrality were most relevant to predicting the level of social anxiety among the input brain regions and the input type of radiomic features, respectively. These results suggest potential validity for predicting social anxiety with machine learning of the resting-state brain functional radiomic features and provide further understanding of the neural basis of the symptom.
Introduction
Social anxiety is a highly prevalent condition among young adults, often resulting in maladaptive social behaviors. Despite its impact, many individuals do not seek clinical assistance due to the stigma surrounding mental health or a lack of objective diagnostic measures. While traditional neuroimaging has identified brain regions associated with anxiety at a group level, individual-level prediction remains a significant challenge. This study explores the intersection of neuroimaging and machine learning, specifically using functional radiomics, to provide an objective lens for screening social anxiety.
Methods
The researchers analyzed resting-state fMRI (rs-fMRI) data from 116 young adults (average age ~23). Participants were classified into "Low Social Anxiety" (LSA) and "High Social Anxiety" (HSA) groups based on the median score of the Liebowitz Social Anxiety Scale (LSAS).
Four functional radiomic features were extracted: Regional Homogeneity (ReHo), fractional Amplitude of Low-Frequency Fluctuation (fALFF), fractional Resting-State Physiological Fluctuation Amplitude (fRSFA), and Degree Centrality (DC). These features were focused on seven key brain regions involved in the limbic system and social processing, including the orbitofrontal cortex (OFC) and the amygdala. Five machine learning classifiers—Logistic Regression, Support Vector Machine (SVM), Random Forest, Multi-layer Perceptron (MLP), and XGBoost—were trained and compared.
Results
The XGBoost model demonstrated superior performance, achieving a balanced accuracy of 77.7% and an F1 score of 0.815, significantly outperforming the baseline SVM model.
Feature importance analysis using SHAP values revealed that the orbitofrontal cortex (OFC) was the most predictive brain region. Specifically, the fRSFA and Degree Centrality (DC) of the left OFC, along with the DC of the left amygdala, were the most relevant features. The findings suggest that social anxiety is better predicted by global brain connectivity (Degree Centrality) than by localized activity levels.
Discussion
This study highlights the potential of using machine learning and functional radiomics as an objective screening tool for social anxiety. The prominence of the OFC in the model's predictions aligns with its known role in evaluating social rewards and consequences. However, the study's limitations include a relatively small sample size and a focus on the general population rather than clinical patients. Future research should aim to validate these models in larger, more diverse cohorts and test their ability to distinguish between subclinical anxiety and diagnosed Social Anxiety Disorder (SAD). Ultimately, this approach serves as a potential screening aid to encourage individuals to seek professional help.
Summary updated