Microstate analysis from resting state electroencephalography in patients with bipolar disorder converted from major depressive disorder

Limited comparative studies on the neuropathologies of major depressive disorder (MDD) and bipolar disorder (BD) necessitate deeper investigation. We compared the electroencephalography (EEG) microstate characteristics of patients with BD converted from MDD, using machine-learning to evaluate if specific features can distinguish between the two disorders. We enrolled 53 drug-naïve MDD and 17 transitioned to BD patients. Clinical parameters assessed included bipolarity, depression, anxiety, impulsivity, and cognitive functions. Resting-state EEG microstate analysis was conducted to compare features between groups and correlate them with clinical characteristics. In the whole frequency band (1–30 Hz), microstate B showed longer duration, increased occurrence, and wider coverage in the converted BD group compared to the MDD group. Transition probabilities to B from microstates A, C, and D were notably higher in converted BD patients. Transitions from D to A in the delta band correlated with impulsivity measures. Classification of groups using microstate features achieved 75.71% accuracy (95% CI 65.71%–85.71%; p < 0.001), 70.59% sensitivity (95% CI 47.06%–88.24%; p = 0.027), 77.36% specificity (95% CI 66.04%–88.68%; p < 0.001), and ROC-AUC of 0.766 (95% CI 0.644–0.875; p < 0.001). The study’s limitations include a small BD group size, age difference between groups, cross-sectional design limiting causal interpretations, and relatively low microstate global explained variance (GEV) value. These preliminary and exploratory findings suggest that microstate features may provide potentially informative candidate features for future studies aimed at distinguishing patients with MDD from those who later converted to BD.