This paper proposes a high-precision theoretical and computational neurorehabilitation framework for Mild Cognitive Impairment (MCI), connecting computational neuroscience and clinical practice through qEEG-guided neurofeedback training (NFT). By employing sLORETA to identify putative pathological nodes within the Default Mode Network (DMN)—specifically the Precuneus and the Posterior Cingulate—the model utilizes spectral decomposition to isolate the aperiodic 1/f component, reducing background noise bias and allowing the calculation of a pure individual alpha frequency (IAF) to inform recalibration of Weber’s Cognitive Threshold. The core architecture uses Bayesian algorithms and stochastic modeling to drive a Dynamic Weight Change mechanism. To support Long-Term Potentiation (LTP) and Hebbian learning, reward thresholds are modulated in real time to target a 70% success rate, as a strategic rationale to anticipate neural fatigue while maintaining the Reward Prediction Error required for synaptic strengthening. As a prospective validation pathway, future studies may assess clinical value through changes in MoCA and RAVLT scores, as well as by examining normalization of cortical coherence in the Default Mode Network (DMN). By merging computational neuroscience with biological models of synaptic plasticity, this work outlines how individual biology can be mapped into an explicit mathematical model. The proposed framework may inform an individualized protocol that provides an objective model-based measure of cognitive recovery, suggesting a replicable and robust strategy for neurorehabilitation during the prodromal phase of dementia, and providing a new approach to neuroscience-based cognitive rehabilitation. This work is intended as a theoretical and computational framework; no complete empirical dataset is reported in the present manuscript.
Dasilva et al. (Fri,) studied this question.