In this paper, we present a novel approach for identifying salient brain regions and interpreting the ability of nonlinear EEG features to discriminate between anxiety disorders and healthy controls. The proposed method involves the integration of advanced EEG preprocessing and artefact correction, nonlinear feature extraction using conditional permutation entropy, and interpretable machine learning to identify relevant electrodes. The extracted nonlinear features show statistically significant differences between classes, demonstrating high discriminative ability. The discriminative ability was confirmed with T-tests (p = 1.05e-10) and Mann-Whitney U tests (p = 2.65e-11), demonstrating robust statistical significance. Classification results support these findings and guide the identification of relevant electrodes, enhancing the interpretability of the discriminative features. This approach highlights potential brain regions critical for anxiety disorder diagnosis, paving the way for more targeted interventions and improved clinical outcomes.
Biloborodova, Tetiana Skarga-Bandurova, Inna Derkach, Maryna Matiuk, Danylo Zagorodna, Nataliya
School of Engineering, Computing and Mathematics
Year of publication: 2024Date of RADAR deposit: 2025-07-04