Decoding Natural Positive Emotional Behaviors from Human Fronto-Temporal Mesolimbic Structures.

Understanding the correlation between neural features and symptoms of mood disorders, such as depression, could provide objective measurements for diagnosis and facilitate clinical treatments. In this paper, we study the correlation of neural features with positive naturalistic emotional displays, e.g., smiling, in human subjects in a normal setup, without presenting any experimental stimuli to the subjects. We employed a data driven approach and utilized Random Forest classifiers to decode positive emotional displays from brain activity. Our results on all of our eight subjects show that neural features from mesolimbic circuits including cingulate, hippocampus, insula, amygdala and orbitofrontal cortex (OFC) can be used for decoding emotions (mean area under the ROC curve = 0.86 +− 0.04). The most important features based on the Random Forest models were mainly clustered in the gamma frequency band (30–100Hz) and low frequencies, with majority of them in theta band (4–8 Hz). These features were distributed across the limbic network, specific to each individual. Remarkably, the gamma cluster was selective to the positive emotions while the low frequency cluster showed selectivity to the neutral state. These results demonstrate that non-task-based emotions can be decoded from brain neuronal activity, and, may inform biomarker identification for objective symptom assessment in the treatment of severe mood disorders.