Photic stimulation is a routine clinical probe of posterior alpha circuits but its signature in frontotemporal dementia (FTD) and direct comparison with Alzheimer's disease (AD) at the subject level were previously uncharacterized. This study tested whether photic responses, particularly in the alpha band, differ between AD, FTD, and healthy controls, and whether photic measures add diagnostic value beyond resting EEG.
The analysis used two openly available OpenNeuro datasets comprising the same cohort of 88 participants. After quality control 86 participants were retained: 35 with Alzheimer's disease (AD), 22 with frontotemporal dementia (FTD), and 29 healthy controls (CN). The datasets are ds004504 (resting-state eyes-closed scalp EEG) and ds006036 (intermittent photic stimulation scalp EEG). Data were originally released by the 2nd Department of Neurology, AHEPA University Hospital, Aristotle University of Thessaloniki, Greece, under CC0.
Photic stimulation comprised intermittent flashes at 5, 10, 15, and 20 Hz with alternating eye states; resting EEG was recorded with eyes closed. Both datasets correspond to routine clinical EEG recordings described in their respective dataset metadata and publications cited in the source. Participant matching across datasets was performed using BIDS identifiers.
Data underwent independent component analysis (ICA) preprocessing. The primary photic response measure was a computed photic driving index (DI) at stimulation frequencies. Resting-state spectral features included individual alpha frequency (IAF) and posterior theta/alpha ratios. Functional connectivity was estimated using the weighted phase-lag index (wPLI). Derived subject-level feature matrices and analysis code are publicly available.
Group effects were assessed with linear mixed-effects models. Sensitivity analyses included log-transformed, winsorized, and rank-based approaches. Equivalence was evaluated with two one-sided tests (TOST) for connectivity. Diagnostic utility was quantified with leave-one-subject-out cross-validation using five classifiers; balanced accuracy and McNemar tests compared classification performance between feature sets.
The most prominent photic finding was a selective reduction of DI at 10 Hz in both patient groups. DI at 10 Hz was reduced by approximately 50% in AD (β = −6.96, p = 0.004) and in FTD (β = −6.74, p = 0.012). These reductions remained robust across log, winsorized, and rank-based sensitivity analyses, indicating a consistent amplitude decrement in the alpha-band photic response for both disorders.
Resting spectral analyses revealed disease-specific alterations. AD showed a 1.5 Hz slowing of individual alpha frequency (IAF) relative to controls. Posterior theta/alpha ratios were elevated markedly: up to sixfold in AD and 3.3-fold in FTD. These resting spectral markers distinguished groups in expected directions and highlight alterations in intrinsic posterior rhythms in neurodegenerative disease.
Weighted phase-lag index (wPLI) connectivity analysis did not show group differences. However, equivalence between groups for wPLI was not established (TOST p ≥ 0.16), so the data do not support a definitive conclusion that synchrony is preserved across groups. The authors note that longer recordings and designs powered for equivalence testing are required to resolve whether amplitude reductions occur with preserved phase-synchrony.
The split-half reliability of the 10 Hz DI was low (intraclass correlation coefficient ICC = 0.06), indicating poor internal consistency for the photic driving amplitude measure over recording halves. In classification experiments, balanced accuracy for resting features alone was 0.577; when combining resting and photic features balanced accuracy was 0.543. The difference was not significant by McNemar test (p = 0.85), indicating that adding photic DI did not improve diagnostic performance in this sample.
The study demonstrates a selective alpha-band amplitude deficit, specifically reduced 10-Hz photic driving, that is shared between Alzheimer's disease and frontotemporal dementia. Despite this shared deficit, photic DI provided no added diagnostic value over resting EEG in this dataset. The absence of observed wPLI differences does not confirm preserved synchrony because equivalence testing was not achieved. The authors highlight the low reliability of the 10-Hz DI and indicate that longer recordings designed and powered for equivalence testing are necessary to dissociate reduced cortical gain from preserved synchrony.
All raw EEG data and analysis code, including derived feature matrices and statistical outputs, are openly available via OpenNeuro (ds004504, ds006036) and a GitHub repository linked in the source. The datasets were released with appropriate ethical approvals and licenses as reported by the original data contributors.