40 Hz biparietal transcranial alternating current stimulation for Alzheimer’s disease: a prospective study Authors: Xiaowei Ma, Mengwei Yuan, Yan Gao, Ziyu Guo, Jiayi Liang, Fangyuan Yan, Shuaixiang Wang, Huiqing Qiu, Sha He, Yang Li, Yuping Wang Published: 2026 ABSTRACT AbstractBackgroundAlzheimer’s disease (AD) is characterized by dysfunction in multiple cognitive domains. Patients with AD demonstrate a relative attenuation and dysregulation of theta and gamma oscillations in the temporal and parietal lobes at an early stage. Transcranial alternating current stimulation (tACS) has been shown to modulate neural oscillations in different regions of the brain, affecting higher cognitive functions such as motor function, memory, and learning.ObjectiveTo observe the clinical efficacy and after-effects of biparietal 40 Hz tACS in treating AD patients, to analyze its safety and feasibility, and to deeply explore the effects of tACS on time-varying brain networks and dynamic functional connectivity.MethodsIn the study, patients with mild AD who were treated in the First Hospital of Hebei Medical University from October 2022 to August 2023 were recruited, randomly divided into tACS group or Sham group, and then respectively received either biparietal 40 Hz tACS or sham stimulation for 15 consecutive days, each lasting 30 min. Neuropsychological assessment scales were collected to assess the efficacy at three time points: pre-stimulation, post-stimulation, and 10-week follow-up. Additionally, transcranial magnetic stimulation with electroencephalography (TMS-EEG) and functional magnetic resonance imaging (fMRI) were collected to explore time-varying brain networks and functional connectivity.ResultsPatients with AD exhibited improvements in global cognition, memory, language, attention, and executive functions following tACS treatment compared with those before treatment. Additionally, the neuropsychiatric inventory scores were decreased significantly after tACS treatment. The clinical efficacies were not obvious in Sham group. These results indicated that tACS treatment had significant short-term efficacy and good stability. The brain time-varying EEG network patterns were analyzed using the adaptive directed transfer function. The analysis showed enhanced information flows from the temporal and prefrontal regions and from the posterior to anterior regions after tACS treatment. In contrast, decreased information flow was observed from the left to right frontal regions. The fMRI analysis revealed enhanced connectivity within default mode network and between default mode network and Frontoparietal network after active treatment.ConclusionsThe 40 Hz biparietal tACS is a potentially safe and effective treatment for AD with certain long-term efficacy by modulating dynamic functional connectivity.Trial registrationThe trial was retrospectively registered at the Chinese Clinical Trial Registry (registration number: ChiCTR2500115019).Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12916-026-04983-w.Keywords: Transcranial alternating current stimulation, Biparietal, Gamma-frequency, Alzheimer’s disease, Time-varying brain network connection, Dynamic functional connectivity Background Alzheimer’s disease (AD), as a progressive and irreversible neurodegenerative disease, is clinically characterized by cognitive dysfunction, mental behavioral abnormalities, and reduced functioning in social life, which progresses gradually to dementia [1, 2]. With the aging of the population, the incidence and prevalence of AD are increasing annually. It is estimated that approximately 110 million individuals worldwide will be affected by this disease by 2050, which results in a heavy social and economic burden [3–6]. At present, the treatment of AD is dominated by drugs, including traditional cholinesterase inhibitors and partial antagonists of N-methyl-D-aspartate receptors as well as the emerging three anti-amyloid-beta (Aβ) antibodies, aducanumab, lecanemab and donanemab [7–14]. However, pharmacologic treatments are often constrained by concomitant adverse effects, limitations in efficacy, as well as the stringent conditions of use and accompanying high costs set by monotherapy [15–17]. Non-pharmacological interventions have attracted considerable attention for their efficacy, fewer side effects, maneuverability, and high acceptance by patients and their families [1, 18–21]. AD has a diverse pathology including Aβ and tau protein accumulation, microglia dysfunction, neurodegeneration, and altered activity of oscillatory network activity [22, 23]. Among these, non-invasive brain stimulation techniques are expected to delay the onset and progression of AD by modulating neurological function and ameliorating pathology [24]. Transcranial alternating current stimulation (tACS) manipulates neuronal oscillations by applying sinusoidal currents externally to the scalp at specific frequencies, amplitudes, and phases of oscillations [25, 26]. tACS improves cognitive abilities, such as episodic memory, working memory, long-term memory, attention and executive control [27]. The amyloid cascade has long been recognized as the main explanation for the pathogenesis of AD, which is the most common clinical therapeutic target for AD [28, 29]. The findings from animal models and patients with AD indicate that long-term tACS ameliorated memory deficits by increasing microglial activity, regulating inflammation, and inducing clearance of Aβ and phosphorylated tau deposits [30–33]. It can improve various cognitive functions by influencing intrinsic neural oscillations, improving brain networks connectivity, increasing cerebral blood flow, and reducing neurotoxic substances in the brain [26, 27, 34]. In addition, tACS intervention can produce long-term sustained therapeutic effects and is often used as one of the clinical interventions [35]. Different brain regions and cognitive processes are associated with different frequency bands in electroencephalogram (EEG). tACS has the advantage of being frequency-selective and site-selective in improving cortical excitability and brain function. Rhythmic neuromodulation targeting gamma in the left dorsolateral prefrontal lobe (DLPFC) preferentially improves long-term memory, whereas rhythmic neuromodulation targeting theta in the left inferior parietal lobe selectively improves working memory in older adults [36]. The early stage of AD is characterized by changes in neuronal hyperexcitability, spontaneous epileptiform activity, and cerebral oscillations, particularly decreases in the alpha (7–13 Hz), beta (14–30 Hz)- and/or gamma (30–140 Hz) frequency bands, and increases in the theta (4–7 Hz) frequency bands in EEG [37, 38]. Of these, oscillations in the low gamma frequency (30–80 Hz) have shown the strongest association with higher order cognitive functions, particularly working memory, learning, and attention [39]. Notably, significant abnormalities or disturbances in gamma frequency are positively correlated with the progression of AD [38, 40]. Zhou et al. found that applying 40 Hz tACS to the bilateral temporal lobes could significantly improve the cognitive function of patients with AD, accompanied by changes in peripheral blood with the ratio of Aβ40 to Aβ42, although the efficiency could not last until 12 weeks [31]. On the contrary, Sprugnoli et al. showed no change in global cognitive function in patients with mild to moderate AD after tACS treatment. However, it is exciting that the perfusion of bilateral temporal lobes increased significantly, and the perfusion changes were positively correlated with changes in episodic memory and changes in gamma-band spectral power [30]. Thus, neural oscillations at gamma frequencies play an important role in cognitive processing, and the effect of the 40 Hz on cognitive function is particularly pronounced in animals and humans. The tACS stimulation sites most commonly used to improve cognitive function in AD are typically the DLPFC and parietal lobes, with the parietal lobe potentially showing greater improvement in cognitive function [41, 42]. The parietal cortex, as a component of the cortico-hippocampal network, has been associated with memory functions such as episodic and working memory [43, 44]. The posterior parietal cortex (PPC) is considered a key node in this cognitive dysfunction [45]. Dual-target TMS showed that PPC modulatory effects in AD patients were positively correlated with cognitive performance [46]. Therefore, gamma tACS targeting the parietal cortex would be a promising intervention to improve cognitive function, but whether tACS improves cognitive function and the duration of efficacy maintenance in AD patients remains to be validated by further studies. In addition, tACS may be particularly effective in promoting interregional connectivity by improving phase synchronization [47]. Since functional connectivity within and between brain networks is reduced in AD patients [48], whether tACS targeted to PPC improves cognitive function by modulating changes in functional connectivity within and between brain networks in AD patients is an urgent matter for further investigation. In this study, we conducted a pilot, single-blinded, sham-controlled clinical trial, recruiting patients with mild AD for 40 Hz high definition-tACS (HD-tACS) or sham stimulation for 15 days and assessing cognitive function using neuropsychological assessment scales. In addition, we used the combination of transcranial magnetic stimulation and electroencephalography (TMS-EEG) and resting-state functional magnetic resonance imaging (rs-fMRI) techniques to explore the probable mechanisms of tACS for AD. Methods Study design and participants This randomized single-blinded controlled trial has been registered at the Chinese Clinical Trial Registry (ChiCTR2500115019). Participants were recruited from the First Hospital of Hebei Medical University, Shijiazhuang, China, from October 2022 to August 2023. Forty patients with probable AD were recruited and randomly divided into tACS group and Sham group. The primary outcome was the change in the Alzheimer’s Disease Assessment Scale-Cognitive subscale (ADAS-Cog) score following intervention. The secondary outcomes included changes in global cognition, episodic memory, language, executive function, attention, neuropsychiatric symptoms, TMS-EEG network characteristics, and rs-fMRI connectivity measures. Outcomes were assessed at pre-stimulation, post-stimulation, and 10-week follow-up. A flow chart of the study is shown in Fig. 1A. Experimental protocol. A Study design and relevant assessments. B HD-tACS electrode layout. In the 2D and 3D electrode layout, the red dots represent the central electrodes and the blue dots represent the surrounding electrodes. On the left hemisphere, a central electrode was placed on PPC, corresponding to P3; four surrounding electrodes were placed square around the central electrode, corresponding to P1, P5, CP3 and PO3. On the right hemisphere, a central electrode was placed at P4; four surrounding electrodes were placed square around the center electrode, corresponding to P2, P6, CP4 and PO4. C The modeling of electric field for participants receiving bilateral parietal lobe HD-tACS stimulation. The intensity of the simulated field is indicated by the color bar. Blue indicates zero electric field (0 V/m) and red indicates peak magnitude (0.14 V/m) Inclusion criteria were as follows: (1) right-handed patients aged 55–85 years, (2) at least 6 years of education, (3) probable AD dementia diagnosed according to the core clinical criteria of the 2011 National Institute on Aging AD (NIA-AA) guidelines [49], (4) Clinical Dementia Rating (CDR) score of 1.0. Exclusion criteria were as follows: (1) complications with other neurological diseases, structural brain abnormalities, or severe diseases of the heart, liver, lung, or kidney; (2) metallic implants, such as pacemakers, or cardiac stents; (3) currently taking benzodiazepines or having a history of drug abuse; (4) a diagnosis of major depressive disorder or other severe mental illness. Participants who were on medication are eligible for inclusion in the study, provided that their treatment dosage remains stable from 6 weeks prior to the intervention until the end of the follow-up period. This study was approved by the local medical research ethics committee at the First Hospital of Hebei Medical University, Shijiazhuang, China. The consents were provided by all participants. Randomization and blinding After providing informed consent, patients were randomly assigned in a 1:1 ratio to either the tACS group or the Sham group using computer-generated random number sequences. Allocation concealment was achieved using sequentially numbered, opaque, sealed envelopes. Participants and their caregivers were blinded to group assignment. The operator independently set the transcranial alternating current stimulation parameters, and the display content of the real and false stimulation parameters is consistent on the screen. For the sham stimulation, the device was programmed to mimic the active stimulation by incorporating current ramp-up and ramp-down phases at the beginning and end of each session. Participants could not determine the authenticity of the tACS stimulation either by the appearance of the stimulator or by the sensation experienced. Sample size Based on previous studies, sample size estimation was performed using G*Power 3.1 (Heinrich Heine University Düsseldorf, Düsseldorf, Germany). Assuming a moderate effect size, a statistical power of 80%, and a significance level of 5%, a minimum of 14 participants per group was required. Assuming a 20% dropout rate, at least 34 participants were planned for recruitment. tACS Stimulation was performed with an HD-tACS stimulator (VTS-801D, Nanjing Volcan Medical Technology Co., Ltd, Jiangsu, China). In real stimulation, an alternating sinusoidal current of 1.5 mA peak to peak (current density: 0.86 mA/cm2) at a frequency of 40 Hz was applied for 30 min. Stimulation was initiated on the right side of the brain first, followed by the left side. There was a current rise time of approximately 30 s before the onset of stimulation and a current fall time of approximately 30 s at the end of stimulation. The central electrode was placed on the PPC (according to the 10–20 international EEG coordinates, the left center was at P3, and the right center was at P4), and the remaining four surrounding electrodes were placed around the central electrode (left side: P1, P5, CP3 and PO3, right side: P2, P6, CP4, PO4) (Fig. 1B). Electric field simulation was performed (Fig. 1C). All ring electrodes (outer diameter × inner diameter × thickness dimensions = 12 mm × 5.7 mm × 2.5 mm) were fixed with fixation caps and ring electrode holders, and were injected with conductive paste to minimize contact impedance (< 5 kΩ). In the Sham group, the device did not deliver a sustained current. To confuse the subjective perception of the participants, the current was increased to 1.5 mA and then decreased during the first and last 60 s of stimulation. During stimulation, participants sat in a comfortable chair in a well-lit, quiet room, kept their eyes open, and were asked not to speak or move visibly. Safety assessment Adverse events were systematically monitored throughout the study. At each stimulation session, participants were specifically asked about predefined stimulation-related adverse events, including headache, dizziness, scalp discomfort, tingling sensation, fatigue, and sleep disturbance. All reported adverse events occurring during or after stimulation were documented. Neuropsychological assessment General cognitive function was assessed using the ADAS-Cog, the Mini Mental State Examination (MMSE), and the Montreal Cognitive Assessment (MoCA). Episodic memory was assessed using the Auditory Vocabulary Learning Test (AVLT). Language function was assessed using the Boston Naming Test (BNT). Executive function was assessed using parts A and B of the Trail Making Test (TMT). The Digit Span Test (DST) was used to assess attention. The Neuropsychiatric Inventory (NPI) was used to assess patients’ mental status. All patients completed assessments of these scales before and after treatment and at the 10-week follow-up. In addition, data on patients’ demographic and clinical characteristics were collected at baseline. All assessments were evaluated by experienced clinicians blind to the trial. TMS-EEG acquisition Single pulse was performed with a TMS stimulator (MagPro R30, Tonica Elektronik A/S, Farum, Denmark). The patient wore an electrode cap (GT Cap PRO, Greentek, Wuhan, China) with 32 electrodes positioned according to 10–20 Montage. 8 min of TMS-EEG data were acquired with a magnetic field compatible EEG amplifier (PN-NET, Yunshen, Beijing, China). The sampling rate was 256 Hz, and the electrode impedance was kept at 5 kΩ. The electrodes on the earlobes were used for reference, and AFz was used as the ground electrode. Single-pulse TMS at 90% resting motor threshold (RMT) was delivered through a figure-8-coil on the P3. A total of 120 pulses were delivered with a 4 s interval to avoid TMS effects. The examination was collected at baseline and within 1 h after intervention. During data acquisition, subjects were asked to remain still and close their eyes. TMS-EEG analysis MATLAB (R2019b, MathWorks, USA) was used for preprocessing and time-varying network analysis. First, the filter bandwidth of the EEG data was adjusted to 3 ~ 30 Hz, and the data sampling rate was reduced to 128 Hz. The EEG data from − 1000 to 2000 ms were intercepted as data segments. About 80–100 data segments (accounting for 66.7%-83.3% of total segments) were retained for individual EEG data after removing the bad segments. Second, the neural signals were baseline-corrected relative to the pre-TMS interval (-500 to -40 ms) prior to analyzing the post-TMS response window (60-2000 ms). Third, a time-varying multivariate adaptive autoregressive (tv-MVAAR) model was developed and the adaptive directed transfer function (ADTF) matrix was calculated. Finally, we mapped the brain network. Time-varying EEG networks were analyzed using paired-sample t test. The procedure can be found in the Supplementary Material, in accordance with the studies of Song and Liu et al. [50–54]. rs-fMRI data acquisition The rs-fMRI is an effective tool for detecting changes in brain activity and connectivity. The 3.0 T MRI scanner (MACNETOM Prisma, Siemens Healthineers, Erlangen, Germany) with a 20-channel head coil was used for MRI data acquisition. Structural images were scanned using a sagittal 3D T1 weighted image sequence with the following parameters: TR = 800 ms, TE = 2.15 ms, flip angle = 8°, layer thickness = 1 mm, voxel = 1 mm × 1 mm × 1 mm, FOV = 256 mm × 256 mm, matrix = 256 × 256, and number of layers = 144. The resting-state functional image was scanned using an axial plane echo imaging sequence: TR = 2000 ms, TE = 30 ms, flip angle = 90°, voxel = 3.8 mm × 3.8 mm × 3 mm, FOV = 240 mm × 240 mm, matrix = 64 × 64, layer thickness = 3 mm, number of layers = 36. The scan was repeated 240 times, and took 8 min 8 s. Image acquisition was completed at baseline and within 24 h after intervention. During the scan, subjects were asked to lie supine with their heads immobilized with foam padding and to wear earplugs to reduce noise, to breathe evenly, to close their eyes, to minimize their thought activity, and to remain still but conscious throughout the scan. rs-fMRI preprocessing The rs-fMRI data analysis was conducted using the SPM, DPABI, RESTPLUS software within MATLAB. Before processing, DICOM-formatted data acquired from clinical MRI scanner were converted into NIFTI format using the dcm2niigui feature within the MRIcron tool. This format can describe the pixel values and spatial coordinates of voxels in detail, and more accurately reflect the original metadata. The specific process of pretreatment is as follows: (1) Removal of the first 10 time points to reduce the interference of environmental factors on participants and the impact of initial magnetic field instability on imaging. (2) Perform an origin correction and adjust the orientation of the images to ensure that all images are compared and analyzed in the same anatomical direction. (3) The spatial normalization process was executed, aligning the images to the Montreal Neurological Institute standard space. During this alignment, individual T1 structural images were used as references for registration. The T1 structural images were registered to their corresponding mean functional images, facilitating subsequent comparisons of brain image data across different individuals. (4) Image smoothing was applied using a Gaussian kernel with a full-width at half-maximum of 6 mm to improve noise characteristics and enhance signal accuracy. (5) A linear trend removal step was implemented to exclude confounding factors. (6) Interferences were mitigated, including the removal of head motion effects, as well as the regression of signals from white matter and cerebrospinal fluid. (7) A band-pass filter (0.01–0.08 Hz) was applied to eliminate high-frequency noise and low-frequency signal drifts. Dynamic functional connectivity (dFC) Changes in dFC can serve as a non-invasive diagnostic and prognostic indicator for AD. The dFC analysis was performed using the DynamicBC software. The regions of interest were defined in the cortex using Schaefer 400 atlases. 400 cortical areas were divided into seven networks, including visual network (VN), somatomotor network (SMN), dorsal attention network (DAN), ventral attention network (VAN), limbic network (LN), frontoparietal network (FPN), and default mode network (DMN). The sliding time window method was used to construct the dFC network. The size of the sliding window was set to 30 TRs of a rectangular window convolved with a Gaussian (= 3 TRs) and sliding in 1 TR. A FC matrix is derived from each sliding window. It generates a total of 1360 (17 subjects × 80 windows) FC matrices. K-means clustering with Euclidean distance squared, 500 iterations, and 150 repetitions of the dFC window was used to evaluate the repetition pattern of FCs. The optimal number of cluster states is estimated by the elbow criterion. The number was determined to be 3, which means 3 FC states, showing different FC patterns. The paired-sample t test was performed on the mean dFC correlations before and after stimulation for each state. Finally, two temporal metrics were computed for each participant in each state: frequency and mean dwell time. Frequency is the percentage of time each state occurred during the scan. Mean dwell time is the average consecutive time that each state lasts. Statistical analysis Statistical analyses were performed with SPSS 22.0 (SPSS, Inc., Chicago, USA). Normality tests were conducted for all the data. Data are expressed as mean ± standard deviation, unless otherwise noted. Baseline demographic and clinical variables were compared between groups using student’s t test, Mann-Whitney test, or Chi-square test. Repeated-measures ANOVA was performed to evaluate the effect of the tACS intervention, with treatment as a between-group factor and time as a within-group factor. Post hoc tests were performed with Bonferroni correction for multiple comparisons. Sphericity of the variable set was assessed using the Mauchly test, and if it was violated, the Greenhousee Geisser correction was used. Generalized estimating equations were selected when biased abnormal data repeated measures were analyzed. Participants who did not complete all follow-up assessments were excluded from the final analyses, and missing data were not imputed. Results Demographic and clinical characteristics A total of 40 patients with AD were randomly assigned to either the Sham or the tACS group. 3 patients were excluded from the study because they were unable to adhere to the entire course of treatment. 2 patients chose alternative treatments, and 1 patient was excluded due to a diagnosis of esophageal cancer during the follow-up period (Fig. 2). The remaining 34 patients (17 in each group) participated in the entire trial. No statistically significant differences were observed in the demographic characteristics of the Sham and tACS groups with regard to gender, age, and education (p > 0.05) (Table 1). Furthermore, no statistically significant differences were observed between the Sham and tACS groups in terms of disease duration, hypertension and hyperlipidemia (Table 1). Study flow chart Table 1Demographic features at baselineVariablesSham(N = 17)tACS(N = 17)χ2/t p Gender (Female, n%) a10(58.82%)10(58.82%)<0.0011.000Age (years)b65.47 ± 7.4865.41 ± 7.760.8900.419Education (years)b11.88 ± 2.5811.94 ± 2.630.5430.585Duration (months) b29.76 ± 9.6330.76 ± 9.37-0.3070.761Hypertension (Yes, n%) a8(47.1%)6(35.3%)0.4860.486Hyperlipidemia (Yes, n%) a6(35.3%)7(41.2%)0.1250.724 Demographic features at baseline The data are express as mean ± standard deviation or counts (%). Results of the Chi-square test (a) or two independent-sample t test (b) comparing the three groups in the baseline are reported. Neuropsychological characteristic The results of the statistical analyses for scales are presented in Table 2; Fig. 3. In the study, the interaction effect of scores on all neuropsychological assessments was found to be statistically significant (Fig. 3). Bonferroni post hoc tests were employed to facilitate the comparison of differences in time intervals. In the tACS group, all neuropsychological scales demonstrated improvement at post-stimulation in comparison to pre-stimulation. At 10-week follow-up, a tendency was observed for the MMSE, MoCA, ADAS-cog, BNT, DST-forward, DST-backward, TMT-A, TMT-B, and NPI-Total to revert to pre-stimulation. However, improvements remained significantly superior to baseline levels. In the Sham group, a significant improvement was observed in the ADAS-cog, TMT-A, TMT-B, and NPI assessment at post-stimulation compared to pre- stimulation. At 10-week follow-up, a trend toward a return to baseline was observed, while improvements remained significantly better than baseline for TMT-A, TMT-B, and NPI-Total. Conversely, performance on the MMSE was poorer than baseline. At post-stimulation, statistically significant differences were observed in MMSE, MoCA, ADAS-cog, BNT, DST-forward, DST-backward, and AVLT-Total scores between the tACS and Sham groups. At 10-week follow-up, statistically significant differences were observed in MMSE, MoCA, and BNT scores between the two groups. However, no statistically significant differences were observed in the remaining scales comparing improvement between the two groups throughout the study period. Table 2Comparison of the neuropsychological performance between the Sham group and the tACS groupScalesmeasuring timetime*groupgroupF/Waldχ2 p F/Waldχ2 p F/Waldχ2 p MMSEa70.930<0.001***70.161<0.001***1.4740.234MoCAa49.105<0.001***50.689<0.001***4.1480.050*ADAS-coga52.625<0.001***27.093<0.001***0.0050.943BNTa36.086<0.001***17.137<0.001***3.3000.079DST-forwarda10.803<0.001***8.3420.001***2.8380.102DST-backwardb22.308<0.001***10.9640.004**1.4560.228RAVLT-IRa33.134<0.001***13.419<0.001***0.3310.569RAVLT-totala51.906<0.001***24.919<0.001***1.2270.276TMT-A (sec) b25.674<0.001***31.895<0.001***0.0010.970TMT-B (sec) b25.674<0.001***31.504<0.001***0.0010.970NPI-Totalb9.7230.008**17.3050.004**0.0480.827 Comparison of the neuropsychological performance between the Sham group and the tACS group Cognitive function and neuropsychiatric symptoms at pre-stimulation, post-stimulation, and 10-week follow-up (12 weeks after baseline) in the Sham (blue) and tACS (red) groups. A Changes in MMSE score. B Changes in MoCA score. C Changes in ADAS-cog score. D Changes in BNT-Spontaneous score. E Changes in DST-Forward score. F Changes in DST-Backward score. G Changes in AVLT-Immediate recall score. H Changes in AVLT-Total score. I Changes in TMT-A score. J Changes in TMT-B score. K Changes in NPI-Total score. *p < 0.05, ** p < 0.01, *** p < 0.001, versus pre-stimulation; # p < 0.05, ## p < 0.01, between the Sham group and tACS group Results of the repeated-measures ANOVA (a) or generalized estimating equation (b) comparing the two groups in the baseline are reported. ** p < 0.01, *** p < 0.001. Differences in the time-varying EEG network patterns In the tACS group, no significant changes in RMT were observed before and after stimulation (pre-stimulation RMT: 57.88 ± 7.47; post-stimulation RMT: 56.94 ± 7.87; paired t-test, t = 1.747, p = 0.100). As illustrated in Fig. 4, there is a notable distinction in the dynamic connectivity of the time-varying EEG network at the pre- and post-stimulation in AD patients in the tACS group. Specifically, it demonstrated that information flow was dominated by enhancement throughout the observation period. The enhanced information flow originated from the bilateral temporal (T3, T4) and prefrontal (Fpz) region. There was no change in information flow in the early period (100 ms). At the middle stage (280 ~ 460 ms), enhanced information flow was induced from T3 and T4 to frontal (F7, Fz and F8) and parietal (P4 and Pz) regions. Subsequently, an enhanced information flow was induced from Fpz to the temporal (T4, T5) and parietal (Cz, C4) regions. At the late stage (770 ~ 1020 ms), the information flow decrease compared with that at the middle stage, with a predominant increase in information flow from posterior to anterior brain regions, specifically from occipital to prefrontal regions. Additionally, the information flow diminished from the left (Fp1, F7) to the right frontal (F4, F8) regions throughout the observation period. Topography of dynamic connections revealed by TMS-EEG from a paired-sample t test map comparing pre-stimulation and post-stimulation in tACS group. Time (millisecond): after single-pluse TMS; red lines: enhanced connections; green lines: diminished connections; blue arrows: the direction of information flow Differences in cluster analysis All FC matrices were clustered into 6 connectivity states (Fig. 5A), state 1 (8.82%), state 2 (32.21%), state 3 (28.93%), state 4 (8.91%), state 5 (5.53%) and state 6 (15.60%). State 1 presented a neutral connection pattern, which we called the relative segregated state. State 2 exhibits a relatively weak connectivity pattern, marked by a blend of positive intra-network connections (e.g., VN, LN, DMN, and FPN), and predominantly negative and neutral inter-network connections (e.g., negative between DMN and SMN, neutral between DMN and VN). There are relatively fewer positive inter-network connections (e.g., between DMN and LN, DMN and FPN). Consequently, we refer to this state as the sparse connectivity state. States 3, 4, and 5 display similar connectivity patterns, characterized by approximately equal proportions of positive, negative, and neutral connections. The distinguishing feature among these states is the gradual increase in positive intra-network connections (e.g., SMN and DAN) and the progressively stronger negative inter-network connections (e.g., between DMN and VAN, DMN and FPN) from state 3 to 5. We refer to these states as the mild local tight connection state, moderate local tight connection state and severe local tight connection state, respectively. State 6 presented a strong positive connection pattern, characterized by some strong positive connections within networks (e.g., VN, SMN, DMN and FPN) and predominantly positive and neutral inter-network connections (e.g., between DMN and FPN, between SMN and DAN, etc.). And we called it the global tightly connected state. The frequency and percentage of occurrences were higher for state 2. Changes of ALFF, FC and dFC at pre- and post-stimulation in the tACS group. A k-means clustering estimation of the 6 functional connectivity states recurring in all subjects at pre- and post-stimulation in the tACS group. The total number of occurrences and percentage of total occurrences for each state are shown under the subheadings. The thin color bar graph represents the functional connectivity, showing yellow and red as positive functional connectivity and blue as negative functional connectivity, with larger absolute values indicating stronger connectivity. The wide colored bar graphs represent the functional networks to which the independent components belong. B Frequency bar graphs of dFC at pre- and post-stimulation in the tACS group. C Mean dwell time bar graphs of dFC at pre- and post-stimulation in the tACS group Temporal properties of dFC states After stimulation, the frequency and mean dwell time of state 1 and 2 increased and state 3,4,5, and 6 decreased in 6 dynamic functional states. tACS group exhibited a markedly higher frequency in state 2 at the post-stimulation compared to the pre-stimulation (Fig. 5B, pre-stimulation < post-stimulation, p < 0.05). However, no significant differences were identified in mean dwell time within each state between the pre- and post-stimulation periods for the tACS groups (Fig. 5C, pre-stimulation < post-stimulation, p > 0.05). Adverse events All participants were continuously monitored by physicians for adverse events throughout the trial. To ensure the detection of potential adverse effects, structured questionnaires were administered following each tACS session and during follow-up visits, documenting any participant-reported conditions. During the trial, two participants in the tACS group reported transient tingling sensations during stimulation, and one participant experienced mild fatigue. No severe adverse events were observed in either group. Discussion In this study, we observed that exposure to 40 Hz tACS on the biparietal lobes enhanced cognitive function, with improved performance on global cognitive function, memory, attention, and language function domains. This effect was observed immediately after treatment and persisted throughout the 10-week follow-up period. By TMS-EEG data analysis, we observed (1) no alteration in information flows at the early stage, (2) enhanced information flows from the temporal and prefrontal regions at the middle stage, (3) augmentation in information flows from the posterior to anterior brain regions at the late stage, and (4) decreased information flows from the left to right frontal regions. Furthermore, through dFC analysis of fMRI data, we identified 6 recurring functional connectivity states, and the sparse connectivity state occurs the most frequently. In other words, there is an increase in sparse and weak intra-network connections within the DMN, accompanied by an increase of weak positive inter-network connections, particularly between the DMN and FPN after tACS treatment. Additionally, gamma tACS was well-tolerated and did not induce any notable side effects. The trial demonstrated the safety and efficacy of tACS for the treatment of AD, offering promise for future clinical applications. Gamma oscillations are well known for a key role in cortical signaling and cognitive function. Recent studies have supported the potential efficacy of 40 Hz gamma oscillations as a therapeutic strategy for AD [31, 37, 55–57]. Several studies have shown that modulating parietal neural activity improves memory and cognition in both healthy individuals and AD patients [46, 58–60]. Despite discrepancies in the results within the field, some studies, such as Sprugnoli et al. observed no statistically significant alterations in global cognitive performance following 2 or 4 weeks of temporal lobe 40 Hz tACS in patients with mild-to-moderate AD [30]. In addition, the improvement of tACS was not sustained at 12 weeks [31, 61–63]. A 12-week follow-up trial of tACS is currently underway, and its results have not yet been reported [64]. However, our study showed encouraging results. Specifically speaking, we selected the bilateral PPC as the target and demonstrated that 40 Hz tACS has a significant effect on improving cognition, particularly attention, language, and memory, with the curative effect lasting up to 10 weeks in patients with AD. The effect is combined with and beyond the scope of previous tACS clinical trials. Not only does it complement previous research focused on improvements in global cognition, working memory, and episodic memory, but it also unveils, for the first time, the potential of parietal tACS in restoring typically impaired cognitive functions in AD patients, significantly expanding the current understanding and appreciation of tACS’s efficacy in the intervention of cognitive impairments. Notably, the sham group also exhibited improvements in cognitive tasks. Compared to baseline, MMSE scores at post-stimulation, ADAS-Cog scores in follow-up, as well as TMT-A, TMT-B, and NPI-Total scores in poststimulation and follow-up all showed some degree of improvement, though the extent was less pronounced than in the tACS group. On one hand, we think these improvements may be related to the potential learning effects inherent in repeated neuropsychological assessments over short intervals [65, 66], which aligns with previous reports of sham transcranial direct current stimulation (tDCS), improving memory function [50, 67–70]. On the other hand, the sham stimulation in our study was not entirely devoid of electrical current. During the initial and final minutes of each session, the device delivered ramp-up/ramp-down currents to the scalp to mimic the subjective sensation of tACS and minimize placebo effects. However, even such minimal currents might exert effects by psychological expectancy or neuromodulatory pathways [71, 72]. Future studies should adopt more rigorously designed sham controls. AD is regarded as a functional disease of brain with abnormal neural activity and connections in different brain regions. Human brain activity exhibits time-varying properties, and the functional connectivity network in the resting state also fluctuates over time [73]. While our study indicates that tACS can improve AD cognitive function symptoms by modulating PPC, the exact neurophysiological mechanisms remain to be elucidated. We used ADTF to analyze the TMS-EEG signals, and proved that there was abnormal connectivity in AD, with the temporal and prefrontal regions as core nodes, enhanced information flow from bilateral temporal and frontal regions, and enhanced long-distance information transmission from posterior to anterior brain regions. Temporal and frontal lobes are the core nodes of DMN and FPN, and their abnormal functions are potential pathogenesis of AD [74]. The diminution of long-distance connections potentially disrupts functional integration amongst diverse brain regions, which is the basis of cognitive impairment in AD [50, 75]. Our previous research shows that 40 Hz repetitive TMS (rTMS) enhanced the flow of information from the left posterior temporoparietal area to the frontal area and enhanced the dynamic connectivity of the anterior and posterior areas of the brain [50]. The results of this study confirm that 40 Hz tACS achieves similar effects, facilitating global brain coordination and function. Also, we revealed a concomitant attenuation of the information flow from the left to the right frontal regions. This decreased dynamic connectivity between bilateral frontal regions may reflect neural plasticity during adaptive changes in the brain, optimizing global functionality by pruning unnecessary connections. Additionally, this reduction may serve as a compensatory mechanism, prompting the brain to forge new connections in other regions to sustain or enhance function [76, 77]. We surmised 40 Hz tACS on bilateral PPC regulates brain activity, promotes the restoration of information flow trends in the normal direction, and ultimately, the functions of attention, execution, language, and memory are improved. In comparison to the static functional connectivity, dFC can more accurately reflect the time-varying pattern of brain activity, which helps to reveal deeper brain neural activity [78]. In recent research, the decrease of local dynamic functional connections and the decrease of whole brain network metastability in AD temporal lobe, superior frontal gyrus and DMN may lead to the decline of cognitive function, and the two are positively correlated [79, 80]. Schumacher et al. found that AD patients exhibited prolonged dwell times in weak connectivity states, with reduced intra-connectivity within the VN and SMN, and disrupted inter-network connections. Notably, there was an absence of anti-correlation between the DMN and task-positive networks, potentially linked to cognitive impairment [81]. Additionally, AD patients showed weakened functional connectivity between the DMN and FPN [82]. Our results of dFC analysis revealed 6 recurring dynamic functional network states after clustering, and the frequency and percentage of state 2 were the highest. We refer to state 2 as the sparse connectivity state, characterized by positive intra-network connectivity within the DMN, VN, and SMN, as well as positive inter-network connections. The positive connectivity is observed between the DMN and FPN after tACS treatment. However, the negative connectivity is observed between the DMN and each of the following networks: the SMN, DAN, and VAN. These findings corroborate that tACS can reverse abnormal dFC changes in AD patients, aligning with the results of prior rTMS research [83]. There is still sufficient reason to speculate that continuous 15-day biparietal 40 Hz tACS primarily improves cognitive function in AD patients by normalizing dFC, mainly by enhancing connectivity within DMN and between DMN and FPN. The two methods used in this study, namely ADTF and dFC, are both capable of analyzing the temporal dynamics of functional connectivity, enabling the capture of changes in brain functional connections across different time points [84–86]. TMS-EEG possesses high temporal resolution, enabling the detection of millisecond-scale changes in neuronal activity, albeit with limited spatial resolution. Conversely, fMRI offers superior spatial resolution by focusing on the localization of specific brain regions within functional networks, albeit with temporal resolution inferior to that of EEG. The combination of these provides a complementary approach to exploring brain changes. Our time-varying EEG network demonstrates enhanced dynamic connectivity between frontal-temporal and temporo-parietal regions after tACS, as compared to pre-intervention. Analysis of dFC further reveals enhanced dynamic connectivity within the DMN and between the DMN and the FPN. The methods serve as mutual validation and analyzed the potential therapeutic mechanism of tACS from the perspective of brain regions to the network. In summary, our findings suggest that 40 Hz tACS therapy has the potential to ameliorate cognitive symptoms in patients with AD through the reconstruction of dFC within the brain. Our study demonstrated that tACS did not alter the RMT, consistent with previous reports showing no significant effects of anodal tDCS, rTMS, or transcranial near-infrared stimulation on RMT [87–89]. From a neurophysiological perspective, RMT reflects changes in ion channel conductivity and neuronal membrane excitability [89, 90]. However, tACS is thought to exert its effects primarily through entraining neural oscillations, inducing synaptic plasticity, and modulating cortical activity and network connectivity [91]. In this study, the absence of RMT changes suggests that tACS is more likely to enhance connectivity and synaptic plasticity rather than alter membrane excitability. However, our study has some limitations. Firstly, the sample size was relatively small. Future studies should aim to conduct randomized controlled, double-blind trials to replicate and validate our findings in a large sample size and contexts encompassing different ethnicities and populations to be included. Further, it is necessary to conduct subgroup analyses based on clinical characteristics, such as onset state, duration, and severity state of disease. Secondly, the absent correlation between cognitive improvement and fMRI findings likely reflects temporal misalignment, as neuroimaging examination was performed 12–24 h post-intervention. We recommend minimizing the interval between fMRI scanning and the final intervention phase, ideally restricting it to within 1 h to capture immediate neural responses. Thirdly, the absence of standardized quantitative metrics (e.g., ADTF) in TMS-EEG analysis compromises the reproducibility of findings. Future studies should employ multimodal analytical frameworks combining time-domain analysis of TMS-evoked potentials, spectral decomposition of neural oscillations, and time-frequency characterization to fully elucidate cortical network dynamics. Fourthly, the lack of assessments beyond cognitive evaluations in the sham-stimulation control group has constrained our comprehensive understanding of the therapeutic efficacy of sham stimulation. We should consider incorporating broader and more multidimensional evaluation tools for participants, encompassing but not limited to neuroimaging (e.g., MRI, fMRI), neurophysiology (e.g., EEG, magnetoencephalography), and biomarker detection in the future. Fifthly, research should not be satisfied with just observing changes in brain activity in the resting state. The application of advanced neuroimaging modalities to investigate dynamic changes in neural network connectivity during both online states and task-specific conditions would provide valuable insights into the temporal and spatial dynamics of cognitive processing. Sixthly, the absence of biomarker data (e.g., measurement of amyloid and tau levels by imaging, CSF, or plasma assays) to confirm the underlying AD pathology in our patients represents a key methodological limitation. Future work should integrate such assays to confirm pathology and explore its correlation with clinical outcomes, enhancing the pathological interpretation. Lastly, despite the design features implemented to maintain blinding, its success was not formally assessed (e.g., with participant guess questionnaires). Thus, we rely on methodological inference rather than quantitative data to support blinding efficacy. Conclusions In conclusion, this study represents a preliminary exploration of the potential of tACS at 40 Hz in the PPC as a promising alternative or complementary treatment for AD patients, aiming to demonstrate the cognitive improvement effects of 40 Hz HD-tACS targeting biparietal lobes in AD patients, particularly in parietal-related cognitive functional domains. Our results suggest that modulating dynamic functional connectivity is a potential therapeutic mechanism of tACS, primarily enhancing connectivity within the DMN and between DMN and FPN. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements Not applicable. Abbreviations Alzheimer’s disease Alzheimer’s disease assessment scale-cognitive subscale Adaptive directed transfer function Auditory vocabulary learning test Anti-amyloid-beta Boston naming test Clinical dementia rating Dorsal attention network Dynamic functional connection Dorsolateral prefrontal lobe Default mode network Digit span Test Electroencephalogram Functional magnetic resonance imaging Frontoparietal network High definition-tACS Limbic network Mini mental state examination Montreal cognitive assessment National Institute on Aging AD Neuropsychiatric inventory Posterior parietal cortex Resting motor threshold Resting-state functional magnetic resonance imaging Somatomotor network Transcranial alternating current stimulation Transcranial magnetic stimulation with electroencephalography Trail making test Time-varying multivariate adaptive autoregressive Ventral attention network Visual network Author contributions XM: Writing – original draft, Project administration, Supervision, Funding acquisition. MY: Writing – original draft, Visualization, Methodology. YG: Data curation, Investigation, Resources. ZG: Investigation. JL: Formal analysis, Visualization. FY: Formal analysis, Visualization. SW: Investigation. HQ: Data curation, Resources. SH: Investigation, Resources. YL: Conceptualization, Writing – review & editing, Resources, Supervision.YW: Conceptualization, Writing – review & editing. All authors read and approved the final manuscript. Funding This work was supported by the Brain Science and Brain-like Intelligence Technology Research Projects of China [2021ZD0201800 and 2021ZD0201801]; the Medica science research subject of Hebei Provincial Health Commission [20231065]; the senile disease prevention project of Hebei Province [LNB202204]; and the “Spark” Research Project Outstanding Fund of the First Hospital of Hebei Medical University [XH202402]. Data availability The author confirms that all data generated or analysed during this study are included in this published article. The data supporting the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This study was approved by the local medical research ethics committee at the First Hospital of Hebei Medical University, Shijiazhuang, China. The consents were provided by all participants (ethics approval number 20220723). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xiaowei Ma and Mengwei Yuan contributed equally to this work. Contributor Information Yang Li, Email: 1459317247@qq.com. Yuping Wang, Email: 56100008@hebmu.edu.cn, Email: doctorwangyuping@163.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The author confirms that all data generated or analysed during this study are included in this published article. The data supporting the findings of this study are available from the corresponding author upon reasonable request.