Identification of spatial patterns with maximum association between power of resting state neural oscillations and trait anxiety

Carmen Vidaurre,Vadim V. Nikulin,Maria Herrojo Ruiz
DOI: https://doi.org/10.1007/s00521-022-07847-5
2022-10-01
Neural Computing and Applications
Abstract:Abstract Anxiety affects approximately 5–10% of the adult population worldwide, placing a large burden on the health systems. Despite its omnipresence and impact on mental and physical health, most of the individuals affected by anxiety do not receive appropriate treatment. Current research in the field of psychiatry emphasizes the need to identify and validate biological markers relevant to this condition. Neurophysiological preclinical studies are a prominent approach to determine brain rhythms that can be reliable markers of key features of anxiety. However, while neuroimaging research consistently implicated prefrontal cortex and subcortical structures, such as amygdala and hippocampus, in anxiety, there is still a lack of consensus on the underlying neurophysiological processes contributing to this condition. Methods allowing non-invasive recording and assessment of cortical processing may provide an opportunity to help identify anxiety signatures that could be used as intervention targets. In this study, we apply Source-Power Comodulation (SPoC) to electroencephalography (EEG) recordings in a sample of participants with different levels of trait anxiety. SPoC was developed to find spatial filters and patterns whose power comodulates with an external variable in individual participants. The obtained patterns can be interpreted neurophysiologically. Here, we extend the use of SPoC to a multi-subject setting and test its validity using simulated data with a realistic head model. Next, we apply our SPoC framework to resting state EEG of 43 human participants for whom trait anxiety scores were available. SPoC inter-subject analysis of narrow frequency band data reveals neurophysiologically meaningful spatial patterns in the theta band (4–7 Hz) that are negatively correlated with anxiety. The outcome is specific to the theta band and not observed in the alpha (8–12 Hz) or beta (13–30 Hz) frequency range. The theta-band spatial pattern is primarily localised to the superior frontal gyrus. We discuss the relevance of our spatial pattern results for the search of biomarkers for anxiety and their application in neurofeedback studies.
computer science, artificial intelligence
What problem does this paper attempt to address?