Editors' ChoiceDepression

Decoding the mood network

See allHide authors and affiliations

Science Translational Medicine  28 Nov 2018:
Vol. 10, Issue 469, eaav9147
DOI: 10.1126/scitranslmed.aav9147


Machine learning and intracranial electroencephalography recordings reveal a circuit mechanism encoding mood state fluctuations over time.

Biomarkers are vital tools for informing diagnosis and treatment decisions in many fields of medicine, but they have proven relatively elusive in psychiatry, due in part to technical obstacles to probing neural circuits in the living human brain. Kirkby et al. report the discovery of a brain network that predicts real-time fluctuations in mood.

Previous neuroimaging studies have identified changes in brain network organization that correlated with depression and could be used to predict antidepressant response. However, these methods lack the temporal precision necessary for studying rapid oscillations in brain activity. Consequently, our understanding of how brain network oscillations contribute to changes in mood over time remains limited. Kirkby et al. used intracranial electroencephalography (iEEG) to record brain network activity in real time in 21 hospitalized epileptic patients with surgically implanted intracranial electrodes (which were needed for their epilepsy treatment). Using machine learning methods and longitudinal recordings from limbic brain structures that have been implicated in depression, they discovered an emotion-related subnetwork involving the amygdala and ventral hippocampus. This subnetwork was detectable in 62% (13 of 21) of their subjects, and its activity state encoded fluctuations in subjective well-being, as indexed by a self-reported mood scale. Up to 50% of the variance in an individual subject’s emotional state over time could be predicted based solely on this iEEG measure. Remarkably, the subnetwork was also closely related to an individual’s self-reported anxiety level several months before their hospitalization, suggesting that the presence of this network might also influence stable traits like resilience to anxiety or stress reactivity. If this specific iEEG feature can be validated in larger independent cohorts, in other clinical contexts—especially in depressed individuals who do not have epilepsy—it could also have important treatment implications, by providing a highly specific target for closed-loop deep brain stimulation, in which dysfunctional brain circuits could be rescued based on individualized recordings of their activity state in real time.

Highlighted Article

View Abstract

Navigate This Article