Anxious Brains: A Combined Data Fusion Machine Learning Approach to Predict Trait Anxiety from Morphometric Features

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Trait anxiety is a stable tendency to experience negative emotions and perceive threats.
• Previous neuroanatomical studies on anxiety yielded inconsistent results due to univariate analyses.

Purpose of the Study:

• To develop a predictive model for individual differences in trait anxiety using brain morphometric features.
• To investigate the role of anxiety-related brain networks and their modulation of other networks.
• To examine the association between trait anxiety, cognitive emotion regulation strategies, and age.

Main Methods:

• A data fusion unsupervised machine learning approach (Parallel ICA) was used to decompose structural brain images.
• Supervised machine learning (decision tree) and backward regression were employed to build and validate a predictive model.
• Neuroimaging data from 158 participants were analyzed.

Main Results:

• Two distinct covarying gray and white matter networks successfully predicted trait anxiety.
• These networks involved parietal, temporal, and frontal regions, including the precuneus and anterior cingulate.
• Trait anxiety correlated positively with maladaptive strategies (catastrophizing, rumination) and negatively with adaptive ones (reappraisal).
• Trait anxiety showed a negative association with age.

Conclusions:

• Individual differences in trait anxiety can be predicted from specific brain morphometric networks.
• Anxiety is associated with particular cognitive emotion regulation strategies and decreases with age.
• This research offers a foundation for developing future diagnostic predictive models for anxiety.