Background: Discriminating between unipolar and bipolar dimensions of illness represents a clinical challenge in the treatment of mood disorders. Concurrently, the heterogeneity in brain function and symptom profiles makes it difficult to translate basic neuroscientific research to clinical practice. To address these challenges, we developed a brain-based graph-theoretical predictive model (GPM) to prospectively predict the trajectory of symptoms of anhedonia, mania, and impulsivity among patients with mood disorders.
Methods: 79 patients with unipolar depression or bipolar disorder underwent an fMRI scan at baseline, including (i) resting-state and (ii) a reinforcement-learning (RL) task. Reward-related symptoms of anhedonia, impulsivity, and mania were measured at baseline as well as at 3-month and 6-month follow-ups. For each fMRI task, a whole-brain functional connectome was computed, and our machine learning GPM was applied for symptom prediction. Graph-theoretical metrics of integration, segregation, and centrality were computed for the entire brain network and reward circuit, and the best graph metric was identified and selected for symptom prediction using cross-validation.
Results: Cross-sectionally, the GPM predicted anhedonia from global efficiency (integration metric) during the RL task (r=0.31, p=0.026) and impulsivity from the centrality of the left anterior cingulate cortex during resting-state (r=0.31, p=0.013). Global efficiency quantifies information transfer across the brain and centrality refers to the importance of a certain brain region for the spread of information within the network. At 6-month follow-up, the GPM predicted anhedonia from the centrality of the left caudate during resting-state (r=0.52, p=0.023), with the added variance explained by the brain predictor on top of that of baseline symptoms quantified as a 9.87% increase in R square.
Conclusions: Across DSM categories, global efficiency and centrality of the reward circuit predicted symptoms of anhedonia and impulsivity, cross-sectionally and prospectively. The GPM is an innovative modeling approach that can be further utilized for clinical prediction at the individual level.
Live Zoom Session – March 9th
Rotem Dan, PhD, Alexis E. Whitton, PhD, Ashleigh V. Rutherford, BA, Poornima Kumar, PhD, Manon L. Ironside, BS, Diego A. Pizzagalli, PhD
Diego A. Pizzagalli, PhD