TdCCA with Dual-Modal Signal Fusion: Degenerated Occipital and Frontal Connectivity of Adult Moyamoya Disease for Early Identification.

Abstract

Cognitive impairment in patients with moyamoya disease (MMD) manifests earlier than clinical symptoms. Early identification of brain connectivity changes is essential for uncovering the pathogenesis of cognitive impairment in MMD. We proposed a temporally driven canonical correlation analysis (TdCCA) method to achieve dual-modal synchronous information fusion from electroencephalogram (EEG) and functional near-infrared spectroscopy (fNIRS) for exploring the differences in brain connectivity between MMD and normal control groups. The dual-modal fusion features were extracted based on the imaginary part of coherence of the EEG signal (EEG iCOH) and the Pearson correlation coefficients of the fNIRS signal (fNIRS COR) in the resting and working memory state. The machine learning model showed that the accuracy of TdCCA method reached 97%, far higher than single-modal features and feature-level fusion CCA method. Brain connectivity analysis revealed a significant reduction in the strength of the connections between the right occipital lobe and frontal lobes (EEG iOCH: p = 0.022, fNIRS COR p = 0.011) in MMD. These differences reflected the impaired transient memory and executive function in MMD patients. This study contributes to the understanding of the neurophysiological nature of cognitive impairment in MMD and provides a potential adjuvant early identification method for individuals with chronic cerebral ischemia.