Analysis of brain activity based on EEG and fNIRS data using explainable artificial intelligence

The paper examines the possibility of combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) for brain activity analysis. Application of explainable AI methods (XAI SHAP analysis) confirmed the biological interpretability of the results: the dominance of EEG features is consistent with known neurophysiological markers, while the contribution of fNIRS remains limited due to low temporal resolution. A key limitation is the lack of consideration of the time lag of neurovascular coupling, which reduces the usefulness of fNIRS data. A promising direction for further research is the development of asymmetric models that explicitly take into account time delays between modalities (e.g., through cross-modal attention or temporal alignment).