Depression is a prevalent mental health disorder with significant impacts on individuals’ quality of life and societal participation. Current diagnostic methods, primarily relying on self-reported scales and clinical interviews, are subjective, time-consuming, and susceptible to diagnostic inaccuracies. This study proposes FusionNet, a deep learning-based framework that integrates Traditional Chinese Medicine (TCM) consultation data with electroencephalography (EEG) signals obtained from an Oddball paradigm for depression state classification and recognition. The FusionNet model employs a feature-level fusion approach, utilizing EEGNet for spatial-temporal feature extraction and Long Short-Term Memory (LSTM) networks for temporal pattern recognition. Experimental results demonstrate that FusionNet achieves an 85% classification accuracy, highlighting its potential in improving diagnostic precision and efficiency for depression detection. The proposed method offers a robust, objective diagnostic tool that enhances early diagnosis and clinical decision-making, and provides a new approach for personalized treatment strategies.