Virtual brain and electroencephalography explain the variance of memory alterations in mild cognitive impairment

Abstract Background Mild Cognitive Impairment (MCI) is a heterogeneous clinical condition characterized by a wide spectrum of cognitive and behavioural manifestations. Despite numerous studies, the link between neuropsychological performance and pathophysiological signatures of the disease—including Aβ and tau accumulation along with altered excitation/inhibition (E/I) balance and brain rhythms—remains elusive. Methods Here Aβ/tau biomarkers were used to distinguish positive (MCI + - prodromal Alzheimer’s disease) and negative (MCI − ) subjects in a cohort of 30 MCI patients (18 MCI + and 12 MCI − ). Virtual brain models based on high-field magnetic resonance imaging data were then developed to determine the inter-node coupling and E/I profile in resting-state networks, while node spectral information was obtained from source analysis of high-density electroencephalography (HD-EEG). Finally, virtual brains and HD-EEG parameters, creating brain digital twins of individual subjects, were correlated with cognitive performance. Results While virtual brain simulations did not reveal E/I differences between MCI + and MCI − , a positive correlation emerged between synaptic parameters of the limbic network and verbal episodic memory for both groups. EEG power spectral density revealed a lower high-frequency/low-frequency ratio in MCI + largely due to a reduced alpha band in the default mode, limbic, attention, frontoparietal, visual and somatomotor networks. A strong correlation emerged between multimodal parameters and memory functions, supporting that brain digital twin simulations can effectively explain the variability of neuropsychological performance in MCI patients beyond the sensitivity of individual techniques alone. In particular, the combination of HD-EEG and virtual brain parameters explained more than 90% of variance for episodic memory patients’ scores, confirming the compound origin of memory performance involving network specific E/I levels and electroencephalographic activity acting in concert. Conclusions This multimodal and multiparametric analysis combining virtual brain modelling with HD-EEG and molecular data enhances the stratification of MCI patients and could be used to develop digital biomarkers of progression to dementia, opening new perspectives for personalized prognosis and treatment.