Brain state dynamics and working memory in patients with major depressive disorder and bipolar disorder

Working memory (WM) deficits are common in major depressive disorder (MDD) and bipolar disorder (BD) and may reflect disrupted transitions between latent brain states during cognitive engagement. Using a robust algorithm called the Bayesian switching dynamical system (BSDS), we studied hidden brain states and their transitions related to WM deficits in MDD and BD.

Using the BSDS, we characterized hidden brain states and their transitions during an N-back task (fixation, 0-back, and 2-back) in 42 patients with MDD, 75 with BD, and 64 healthy controls (HCs), focusing on regions of interest within the default mode network (DMN) and frontoparietal network (FPN). We compared state occupancy, mean lifetime, and transition probabilities across groups while controlling for age, sex, education, and framewise displacement, applying false discovery rate (FDR) correction.

BSDS identified four states: an FPN-dominant task-related state (ST), a DMN dominant fixation-related state (SF), and two low-occupancy hybrid states, including a posterior DMN-dominant state (SP) and an anterior DMN-dominant state (SA). Relative to BD and HCs, MDD showed a longer mean lifetime of SA during 0-back, and higher SA occupancy and mean lifetime during 2-back. Within MDD, greater SA expression was associated with slower 2-back responses and lower WAIS-CR Information scores.

MDD and BD share preserved dominant FPN-DMN dynamics during WM, whereas MDD is characterized by heightened engagement of a low-occupancy anterior DMN-dominant state linked to inefficient performance.