Functional MRI in Depression—Multivariate Analysis of Emotional Task

Abstract

The field of translational neuroscience suffers from an extremely low replication levels compared to other life science fields. The objective of the present study was to test the hypothesis that multivariate analysis of a classical emotional pictures paradigm would produce meaningful brain signatures with some power to discriminate depressed patients from healthy subjects. Participants in the study were eighteen medicated depressed patients and eighteen sex and age matched healthy controls. Functional MRI paradigm with a visual presentation of emotional pictures (positive, negative and neutral) from the International Affective Pictures System was used. The multivariate linear method (MLM) was used to derive the specific brain signatures on an individual and on a group level. The predictive power of the brain signatures is tested by use of linear discriminant analysis. Following the individual and group MLM, the three brain patterns that summarized all the individual variabilities of the individual brain patterns were produced. The discriminant analysis yielded accuracy levels for the three brain signatures ranging from 67 to 98%. The present study demonstrated that the multivariate linear method resulted in meaningful brain signatures with significant potential for distinction between healthy and depressed subjects. Such findings will fuel the emerging paradigm shift from more conventional statistical analysis to the probably more appropriate for the field of functional neuroimaging machine learning techniques.