High-density EEG network analysis in MCI: an exploratory study of electrode density and cognitive performance.

Introduction: Electroencephalography-derived functional connectivity and graph-theoretical measures are increasingly investigated as markers of cognitive dysfunction in mild cognitive impairment, but the extent to which their clinical sensitivity depends on electrode density remains unclear. This study examined whether source-level network metrics are associated with cognitive performance and whether electrode density influences their ability to capture and classify cognitive differences within a mild cognitive impairment cohort. Methods: Resting-state high-density electroencephalography was acquired from 17 individuals with mild cognitive impairment. From the original recordings, three montages were derived (173, 64, and 18 channels). Source-level lagged linear connectivity was reconstructed with exact low-resolution electromagnetic tomography across 84 cortical regions and used to compute graph-theoretical metrics of network integration and segregation. Associations with Montreal Cognitive Assessment scores were examined using Spearman correlations, between-group differences were assessed after stratification by cognitive performance, and single-feature binary classification was evaluated with leave-one-out cross-validation. Results: Correlation patterns appeared progressively clearer with increasing electrode density, especially in the alpha and beta bands, although none survived correction for multiple comparisons. In contrast, the 173-channel montage showed the strongest and most consistent between-group differences, with several metrics remaining significant after false discovery rate correction. The same configuration also achieved the best classification performance, with balanced accuracy up to 0.82 and area under the curve up to 0.94. Discussion: These findings suggest that electrode density may affect the sensitivity of source-level electroencephalography network analysis to cognition-related alterations. In this exploratory sample, high-density recordings showed the clearest between-group differences and the best classification performance, consistent with the possibility that denser montages provide richer network information within the mild cognitive impairment spectrum.