Relations of gray matter volume to dimensional measures of cognition and affect in mood disorders

Understanding the relationship between brain measurements and behavioral performance is an important step in developing approaches for early identification of any psychiatric difficulties and interventions to modify these challenges. Conventional methods to identify associations between regional brain volume and behavioral measures are not optimized, either in scale, scope, or specificity. To find meaningful associations between brain and behavior with greater sensitivity and precision, we applied data-driven factor analytic models to identify and extract individual differences in latent cognitive functions embedded across several computerized cognitive tasks. Furthermore, we simultaneously utilized a keyword-based neuroimaging meta-analytic tool (i.e., NeuroSynth), restricted atlas-parcel matching, and factor-analytic models to narrow down the scope of search and to further aggregate gray matter volume (GMV) data into empirical clusters. We recruited an early adult community cross-sectional sample (Total n = 177, age 18-30) that consisted of individuals with no history of any mood disorder (healthy controls, n = 44), those with remitted major depressive disorder (rMDD, n = 104), and those with a diagnosis of bipolar disorder currently in euthymic state (eBP, n = 29). Study participants underwent structural magnetic resonance imaging (MRI) scans and separately completed behavioral testing using computerized measures. Factor-analyzing five computerized tasks used to assess aspects of cognitive and affective processing resulted in seven latent dimensions: (a) Emotional Memory, (b) Interference Resolution, (c) Reward Sensitivity, (d) Complex Inhibitory Control, (e) Facial Emotion Sensitivity, (f) Sustained attention, and (g)Simple Impulsivity/Response Style. These seven dimensions were then labeled with specific keywords which were used to create neuroanatomical maps using NeuroSynth. These masks were further subdivided into GMV clusters. Using regression, we identified GMV clusters that were predictive of individual differences across each of the aforementioned seven cognitive dimensions. We demonstrate that a dimensional approach consistent with core principles of RDoC can be utilized to identify structural variability predictive of critical dimensions of human behavior.