P300-based support vector machine model for identifying adolescent with first-episode, drug-naïve major depressive disorder.

BACKGROUND: Major depressive disorder (MDD) is diagnosed mainly through clinical interviews, highlighting a need for objective neurophysiological measures. Although P300 abnormalities have been well documented in adults with MDD, evidence in adolescents remains scarce. This study examined P300 alterations in adolescents with first-episode, drug-naïve MDD and evaluated its diagnostic potential. METHODS: A total of 182 adolescents with first-episode, drug-naïve MDD and 127 healthy controls (HCs) participated in the study. Electroencephalogram was recorded during a visual oddball task, and P300 amplitude and latency were extracted as features for support vector machine (SVM) classification between groups. RESULTS: Adolescents with MDD exhibited significantly reduced P300 amplitudes in response to both standard and target stimuli compared to HCs (all p < 0.05, Cohen's d = 0.263-1.139), with no group differences were observed in P300 latency (all p > 0.05). In the MDD group, P300 amplitude was negatively correlated with the Children's Depression Inventory scores (p < 0.001, r = -0.232 - -0.347). The SVM classifier based on P300 features achieved a maximum accuracy of 90.63%. LIMITATIONS: The proportion of females was significantly higher in the MDD group than in the HCs group. The performance of the SVM model has not yet been validated using an independent external dataset. CONCLUSION: Adolescents with MDD exhibit reduced P300 amplitudes, which were significantly associated with greater symptom severity. The SVM model demonstrated that P300 features effectively distinguish patients from HCs, underscoring their potential as objective neurophysiological measures for early identification and clinical assessment of adolescents MDD.