Standardizing Number of EEG Sensors for AI-Driven Dementia Detection

Alzheimer's Disease (AD), the most prevalent dementia subtype, early detection is paramount for slowing disease progression because cures are not available. Artificial Intelligence (AI) has gained significant attention for AD detection, particularly through Electroencephalography (EEG)—a cost-effective, accessible technique that records brain activity from sensors placed on the scalp. However, the variability in EEG sensor configurations across different settings poses a significant challenge, as typical AI models require different models for specific setups. To address this limitation, this study explores using EEG microstates to standardise sensor configurations and propose a new AI model, E-FastKAN, to validate sensor generalisation, ensuring effectiveness regardless of the original number of sensors. Using data from 115 participants with 575 samples from high-density and low-density sensor setups, this study aims to open a new approach to reduce costs, increase accessibility, and broaden the usability of AI-driven dementia detection.