Optimizing Depression Classification Using Combined Datasets and Hyperparameter Tuning with Optuna

Abstract

Highlights: What are the main findings? The study proposes an EEGNet model optimized with Optuna for depression state classification, using data from two public databases and an independent one, achieving 93.27% accuracy on the independent Proprietary dataset. The main contribution of this study lies in its approach to unifying EEG datasets collected from different sources, providing a method for real-world applications where data are more diverse and not confined to a single location. What is the implication of the main finding? The results demonstrate the potential of hyperparameter optimization, offering a framework for handling real-world EEG data variability for depression detection. Unlike recent research, which primarily focuses on individual publicly available datasets, this work emphasizes dataset integration to enhance scalability. The objective was not solely to achieve high classification accuracy but to develop a framework that facilitates the unification of multiple datasets. The study proposes a standardized pipeline for preprocessing and classification, ensuring that the method remains applicable across different EEG datasets. By utilizing a lightweight neural network, the approach is designed to be easily integrated into portable systems, enhancing accessibility in real-world clinical settings. This research focuses on the depression states classification of EEG signals using the EEGNet model optimized with Optuna. The purpose was to increase model performance by combining data from healthy and depressed subjects, which ensured model robustness across datasets. The methodology comprised the construction of a preprocessing pipeline, which included noise filtering, artifact removal, and signal segmentation. Additive extraction from time and frequency domains further captured important features of EEG signals. The model was developed on a merged dataset (DepressionRest and MDD vs. Control) and evaluated on an independent dataset, 93.27% (±0.0610) accuracy with a 34.16 KB int8 model, ideal for portable EEG diagnostics. These results are promising in terms of model performance and depression state-of-the-art classification accuracy. The results suggest that the hyperparameter-optimized Optuna model performs adequately to cope with the variability of real-world data. Furthermore, the model will need improvement before generalization to other datasets, such as the DepressionRest dataset, can be realized. The research identifies the advantages of EEGNet models and optimization using Optuna for clinical diagnostics, with remarkable performance for deployed real-world models. Future work includes the incorporation of the model into portable clinical systems while ensuring compatibility with current EEG devices, as well as the continuous improvement of model performance.