Nonlinear Feature Modeling Using Kolmogorov-Arnold Networks in Electroencephalography for Depression Classification: Deep Learning Study.

BACKGROUND: Depression is a common mental disorder that seriously affects mental health and quality of life. Traditional depression diagnosis mainly relies on clinical interviews and subjective assessments, which may lead to variability in diagnostic outcomes. Electroencephalogram (EEG)-based computer-aided diagnosis has emerged as a promising, objective approach for depression recognition. However, effectively modeling the complex nonlinear characteristics of EEG signals remains a major challenge for existing deep learning methods. OBJECTIVE: This study aimed to investigate the applicability of Kolmogorov-Arnold Networks (KANs) for EEG-based depression recognition and to evaluate the effectiveness of KAN-based nonlinear modeling frameworks on frequency-domain EEG features. METHODS: KAN was introduced into EEG-based depression recognition to enhance nonlinear feature representation of brain electrical signals. Two representative KAN-based convolutional architectures, ConvKAN and its efficiency-optimized variant ConvKAN Efficient, were systematically investigated for EEG classification tasks. These architectures incorporated KANLinear layers together with dynamic grid updating, enhanced weight initialization, and regularization strategies to improve nonlinear modeling capability while maintaining computational efficiency. Extensive experiments were conducted on the EDRA dataset, and the proposed framework was compared with conventional convolutional neural network-based methods and traditional machine learning approaches. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. RESULTS: Experimental results demonstrated that the KAN-based frameworks achieved superior EEG classification performance compared with conventional convolutional neural network-based methods and traditional machine learning approaches. Among the evaluated models, ConvKAN Efficient achieved the best overall performance, with accuracy, precision, recall, and F1-score all exceeding 94% on the EDRA dataset. The results indicate that KAN-based architectures can effectively capture the complex nonlinear characteristics of EEG signals while preserving high computational efficiency. CONCLUSIONS: This study systematically demonstrates the applicability of KAN-based nonlinear modeling frameworks to EEG-based depression recognition. The proposed framework provides a scalable and effective approach for EEG signal analysis and offers valuable technical support for both theoretical research and clinical applications in computer-aided depression recognition.