Hybrid EEG-fNIRS Detection of MCI Subtypes Based on Transformer Network

Preventing the progression of mild cognitive impairment (MCI) to dementia is of paramount importance in the field of healthcare. Furthermore, detecting the subtle indications of MCI in its early stages is difficult but necessary. MCI can appear as either amnestic (aMCI) or non-amnestic (naMCI), confounding the diagnosis even more. However, with the development of improved neuroimaging techniques, computational algorithms that analyze data acquired during clinical tests might provide a solution. Among them, functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG) offer invaluable insights into an individual’s cognitive state. In this paper, we present a novel approach for distinguishing individuals diagnosed with MCI, specifically aMCI and naMCI, from healthy controls (HC). We employ a hybrid fusion of EEG and fNIRS signals, acquired during a cognitive task. To address this complex classification problem, we propose a new method based on the transformer network for MCI detection. Our experimental results demonstrate a remarkable classification accuracy of 99.78%, highlighting the significant improvement in MCI detection capabilities achieved through the integration of multimodal neuroimaging data.