Estudio de los correlatos neurales de la percepción emocional por análisis de patrones en multitud de voxeles
Emotion and its perception are fundamental psychological faculties for the survival of animals and social interaction. This is recognized by the emergence of whole areas of neuroscience devoted to understanding its neural basis. Although the basic components of such emotional system have been identified, the segregation of the milieu of affective experiences into different patterns of brain signals remains poorly understood. Recent functional imaging studies have implicated simultaneous distributed activity as a better correlate of emotional state than its univariate counterpart; however, those attempts have still restricted themselves to regions of interest and severely-filtered data. In this work we tested whether the visual perception of three basic emotions can be decoded from full brain activity using multivariate pattern classification, while keeping localizationist and encoding assumptions at a minimum. Beyond stimuli prediction, we also provide proof of-concept anatomical mapping and discovery of relevant structures.To this end, we ran a face perception experiment on a sample of 16 neurotypical participants while recording their brain activity using fMRI. Per-subject SVM classifiers were trained on the fMRI data, so that they could recognize the emotion class brains were presented with. Results were cross-validated and compared against performance by chance using resampling techniques; and the whole of our reproducible pipeline was further validated using more trivial contrasts embedded within the main emotional task. Thorough assessment of behavioral data points towards the validity of our task.Results show a robust and distributed representation of (perceived) happiness in humans, but not of negative-valence anger and sadness; contrary to the more optimistic (though less diligent) existing studies. Overall, our approach proved more sensitive and anatomically specific than the classical mass-univariate analysis, amidst high-dimensionality concerns. Group inference of SVM parameters suggests the defining information-bearing pattern emanates from known structures in the ventral visual pathway and emotion-related areas. Namely: the primary visual cortex (V1) and surroundings, the middle collateral sulcus and parahippocampal gyrus (mCS, mPHG), the amygdala, the medial prefrontal cortex (mPFC) and the anterior cerebellum around the vermis; all of them in bilateral fashion. Our work paves the way for further multivariate studies to provide a complementary picture of emotions (and other brain functions), according to its macroscale dynamics.