From read-out geometry to in-silico stimulation: a distributed functional-connectivity signature of Alzheimer's disease

Resting-state functional connectivity (FC) is altered in Alzheimer's disease (AD), widely regarded as a distributed network process; whether its FC signature reduces to a few focal sites has not been tested causally, a question central to targeted neuromodulation.

We fit subject-specific, cross-subject-identifiable reservoir-computing models that reconstruct each individual's lagged FC.

Two model read-outs classify AD from controls at modest accuracy, below that of structural atrophy; we nonetheless build on the functional read-out because it, not tissue loss, is what stimulation can act on.

The ideal correction that maps a patient's dynamics onto the control template is a \emph{distributed} change in the model's connectivity kernel, a coordinated multi-site pattern rather than a focal target: accordingly, a single-site drive at the node with the largest kernel change fails to revert the classification even at supra-physiological amplitudes, whereas selecting each patient's site by its \emph{effect on the disease discriminant} achieves complete, individualised reclassification from one site, and a real-time closed-loop controller reaches comparable efficacy at lower dose using only causally available information.

Optimal targets are cortical and heterogeneous: effective neuromodulation requires model-informed, personalised targeting, the site to stimulate is not where the read-out deviation is largest but where the network is most therapeutically responsive.