Network-targeted neurophysiological biomarkers of dysconnectivity and cognitive decline in Alzheimer’s disease

Abstract Background Alzheimer’s disease (AD) is marked by progressive network disconnection that begins decades before symptoms emerge, yet detecting early functional disruptions remains a key challenge. Transcranial magnetic stimulation combined with electroencephalography (TMS–EEG) provides a direct, non-invasive probe of cortical dynamics and may yield biomarkers that index cortical network dysfunction, track progression, and inform personalized neuromodulation strategies. Methods We enrolled 86 participants: 28 HC (mean age 69.7 ± 6.0, 15 female), 27 MCI (73.1 ± 5.2, 16 female), and 31 ADD (75.8 ± 4.9, 17 female). TMS–EEG targeted individualized default mode (DMN) and executive control (ECN) network nodes identified via resting-state fMRI. Using a data-driven strategy, we extracted early TMS-evoked potential (TEP) components within 50 ms. We assessed group differences, associations with hippocampal volume, white matter integrity, and cognition, and prediction of cognitive decline over ~ 1.5 years ( n = 55). Results After left DMN stimulation, the contralateral frontal N20F peak was increased in both MCI (P-adj. = 0.0001, Δ = 1.89 µV) and ADD (P-adj. = 0.0003, Δ = 1.76 µV) relative to HC, achieving an AUC of 0.82 for HC-versus-MCI classification. In the ECN, N20C amplitude showed a diagnosis-by-hemisphere interaction: reduced in MCI after left stimulation (P-adj. = 0.036, Δ = −13.64 µV) but increased in MCI and ADD after right stimulation (P-adj. = 0.028, Δ = 14.59 µV and P-adj. = 0.048, Δ = 10.30 µV). Left DMN N20C amplitude correlated with superior longitudinal fasciculus integrity ( P = 0.009, β = 37.15). TEPs were selectively linked to executive function: in ADD, larger left ECN N20F predicted worse Trail Making Test scores (P-adj. = 0.0003, β = −24.35). Longitudinally, lower left DMN N20C predicted steeper global cognitive decline (β = 0.74, P = 0.014), and right ECN N20C predicted executive decline in MCI (β = −35.15, P = 0.0004). Conclusions Early network-targeted TEPs capture functional alterations linked to AD progression. Limitations include the moderate sample size, clinical rather than biomarker-confirmed diagnosis, and single-site design. These network-specific, hemisphere-sensitive measures may serve as biomarkers for diagnosis, tracking, and prognosis of cognitive decline, and could inform personalized network-targeted neuromodulation in AD. Trial registration Not applicable (observational study).