Identify a brain-stimulation network target for memory dysfunction.

Therapeutic brain stimulation holds promise in treating memory impairment, but results have been inconsistent in part due to uncertainty in the neuroanatomical target. Here we combine data from patients with brain lesions, DBS, and TMS to map out a convergent memory network and evaluate if its potential as a therapeutic stimulation target.

We analyzed 12 independent datasets (n=1,247): focal lesions (n=985), DBS (n=207), and TMS (n=72). For each lesion or stimulation site, we computed normative functional connectivity and mapped connections covarying with episodic memory outcomes within each dataset. (1) We tested whether lesion-, DBS-, and TMS-derived maps converged on a common brain network. (2) We evaluated whether connectivity to this network explained memory variance in held-out lesion, DBS, and TMS cohorts. (3) We assessed if connectivity to the convergent memory network could explain prior outcomes from Alzheimer’s disease (AD) brain stimulation trials. (4) In retrospective AD DBS patients, we derived theoretically optimal DBS parameters that maximally engaged the convergent memory network and assessed if being closer to these optimal parameters was associated with better memory outcomes.

Lesion, DBS, and TMS maps converged on a common memory network (spatial R=0.75, p=0.0092) with significant nodes in hippocampus, precuneus, cerebellum, lateral parietal, retrosplenial, and prefrontal cortex. Connectivity to this network explained memory variance in held-out datasets and explained significantly more variance than connectivity to prior localizations like the hippocampus, Papez circuit, or default mode network (average ΔR²=23%, p=0.0361). Across 19 AD stimulation trials, sites more positively connected to the convergent network were associated with better memory outcomes (t=4.76, p=0.0005). In AD DBS cohorts, outcomes improved as clinician-programmed settings approached theoretically optimal parameters (ρ=0.55, p=0.0021).