Anxiety is a diffuse condition that can range from mild to more severe manifestations, including proper anxiety disorders. Specific sensitive periods such as adolescence and young adulthood are particularly vulnerable to anxious states and may favour the onset of future anxiety disorders. Until now, neuroanatomical research on anxiety has focused mainly on adults, employed univariate inference-based approaches, and considered one single neuroimaging modality, thus leading to an incomplete picture. The aim of the present study is to characterize the joint GM-WM contribution in high versus low trait anxiety, in a large sample of young individuals, exploiting a data fusion machine learning technique known as Parallel ICA, and to build a predictive model of trait anxiety based on a Random Forest classifier. Additionally, we aimed to characterize high anxiety individuals for their usage of maladaptive coping strategies, and other affective dimensions such as anger, impulsivity, and stress, and to test their relevance in predicting new cases of high trait anxiety.
At the neural level we found a fronto-parieto-cerebellar network to have a decrease gray matter concentration in high anxious individuals, and a parieto-temporal network to have an increase white matter concentration in high anxious individuals.
Additionally, at the psychological level, individuals with high anxiety are characterized by higher stress, cognitive and motor impulsivity, and avoidance/emotional coping.
Lastly, the Random Forest classifier robustly confirmed the goodness of the morphometric and psychological factors in predicting new cases of trait anxiety. As such, these findings may pave the road for the creation of an early biomarker of trait anxiety in young individuals, contributing to an early intervention to prevent the future development of anxiety disorders.