Machine learning-based differentiation of major depressive disorder and bipolar disorder using entropy-derived EEG biomarkers in drug-naïve patients

Objective: Major depressive disorder (MDD) and bipolar disorder (BD) exhibit overlapping depressive presentations, making early differentiation difficult and misdiagnosis clinically consequential.Entropy-based resting-state electroencephalography (EEG) markers were evaluated for differentiating drug-nave patients with stable MDD, converted BD, and diagnosed BD.Methods: Resting-state EEG was acquired from 92 drug-nave patients: 50 with stable MDD, 16 who converted to bipolar II disorder over three years, and 26 with diagnosed BD.Bandscale entropy and cross-sample entropy (CSE) were analyzed.Group differences were tested using analysis of covariance adjusted for depressive severity, and classification was evaluated using nested cross-validation with permutation testing.Results: Relative to stable MDD, diagnosed BD showed higher theta-and alpha-band entropy in frontal and central regions and lower values in selected intra-and inter-regional CSE measures.In diagnosed BD, two of these CSE measures were positively correlated with nonplanning impulsiveness.The converted BD group showed intermediate entropy values, but pairwise differences were not significant.Prospective discrimination of stable MDD versus converted BD was modest (balanced accuracy, 65.4%; area under the receiver operating characteristic curve [ROC-AUC], 0.661).The strongest classification was obtained for stable MDD versus diagnosed BD, yielding 78.9% pooled accuracy, 76.6% balanced accuracy, and a ROC-AUC of 0.832 when entropy features were combined with depressive severity; the combined entropy set outperformed either single-family configuration.Conclusions: Entropy-derived EEG measures differentiated stable MDD from diagnosed BD and suggested a weaker, intermediate profile in converted BD.Entropy-derived markers may