Jiangxi Puzhuo Medical Equipment Co., Ltd

Alzheimer's disease (AD), as the most common neurodegenerative disease, has long suffered from a disconnect between its clinical cognitive stratification and pathological biomarker assessment systems, which has become a major obstacle to precision diagnosis and treatment. Electroencephalography (EEG), with its advantages of being non-invasive and capable of reflecting brain electrical activity in real time, brings new hope to solving this challenge.

A multi-center study published in Clinical Neurophysiology successfully constructed and validated a resting-state EEG-based AD stratification model, achieving precise associations among clinical symptoms, pathological biomarkers, and the therapeutic effects of repetitive transcranial magnetic stimulation (rTMS), providing a brand-new unified framework for the personalized management of AD.

I. Multi-Center Large-Sample Study: Building the Core EEG Stratification Model

This study, jointly conducted by multiple institutions including Southern University of Science and Technology, Shenzhen People's Hospital, and Tianjin Huanhu Hospital, enrolled 308 AD patients, 176 Parkinson's disease (PD) patients, and 181 healthy controls, and was a multi-center retrospective validation study. With thalamocortical circuit integrity as the core focus, the research team performed refined analysis of resting-state EEG data using the irregular resampling auto-spectral analysis method to separate the fractal and oscillatory components of the neural power spectrum, ultimately defining four distinct EEG oscillatory patterns (A, B, C, D) and constructing a novel AD stratification model.

Pattern A: Dominant peak frequency and lower bound >8Hz

Pattern B: Dominant peak >8Hz, lower bound ≤8Hz

Pattern C: Dominant peak ≤8Hz, upper bound >8Hz

Pattern D: Both dominant peak and upper bound ≤8Hz

These four patterns were classified based on the numerical characteristics of alpha peak frequency and spectral boundary values, forming a gradient of thalamocortical denervation severity from mild to severe:

Patterns A and B represent mild abnormalities, while Patterns C and D represent severe abnormalities, precisely corresponding to the disease progression from normal cognition to mild cognitive impairment (MCI) and then to dementia. The study simultaneously enrolled PD patients as disease controls, further validating the specificity of this model for AD pathophysiology.

Figure 1:Electroencephalography (EEG) analysis pipeline. (a) EEG data were acquired from healthy controls and diagnosed Parkinson's/Alzheimer's disease patients recruited from three different research centers. (b–c) After data preprocessing, key neural oscillatory patterns (A, B, C, D) were extracted.

II. Three Core Findings Validating the Clinical and Pathological Value of EEG Stratification

Precisely Capturing the Cognitive Decline Continuum, Matching Clinical Disease Stages

The study confirmed that each EEG oscillatory pattern was highly correlated with cognitive scores (MoCA, MMSE) in AD patients: patients with Patterns A and B had significantly higher cognitive scores than those with Patterns C and D, and the boundary between Patterns B and C corresponded to an MMSE score of 22, precisely aligning with the critical clinical transition point from MCI to dementia, consistent with the eligibility criteria for current AD-targeted therapies.

EEG patterns in healthy controls were almost exclusively of Types A and B, with no Type C or D presentations, indicating that Types C and D are characteristic EEG changes of cognitive decline in AD patients, which also makes EEG stratification an objective metric for tracking cognitive progression in AD.

Associating Core Pathological Biomarkers, Anchoring the Key Node of Tau Hyperphosphorylation

To validate the pathological validity of the model, cerebrospinal fluid (CSF) testing was performed on 100 AD patients to analyze the distribution of core AD biomarkers including Aβ42, t-tau, and p-tau. Results showed that EEG patterns were highly matched with the AD pathological AT(N) classification: Pattern B was predominantly A+T- (amyloid abnormality only), while Pattern C was overwhelmingly dominated by A+T+ (amyloid + tau abnormality), with tau hyperphosphorylation emerging as the core biological marker distinguishing mild from severe EEG patterns.

Additionally, Patterns C and D achieved over 90% accuracy for clinically confirmed AD, far exceeding Patterns A and B, confirming that EEG stratification can precisely reflect the underlying pathological progression of AD, achieving a direct association between brain electrical activity and molecular pathology.

Differentiating Brain Network Damage Characteristics, Reflecting Disease Specificity

Through brain network localization analysis, the study found that even when AD and PD patients exhibited similar EEG patterns, the underlying brain network damage pathways were distinctly different, confirming the disease specificity of the EEG stratification model. The results showed that compared to normal controls (NC), both the AD and PD groups exhibited dorsal attention network impairment under Pattern A. However, between Pattern A and Pattern B, and between Pattern B and Pattern C, the AD and PD groups presented different network damage patterns.

◼ AD patients: The transition from Pattern A to B primarily affected the ventral attention network, while the transition from B to C was centered on visual network impairment; the attention network was the key area of early damage;

◼ PD patients: The visual network was the initial site of damage, subsequently affecting the ventral attention network, forming a stark contrast with AD.

The intact dorsal attention network function in healthy controls also indirectly indicates that early impairment of this network is a common feature of neurodegenerative diseases, providing a target for early screening.

Figure 4:Brain network variability across EEG patterns in different diseases, with no significant differences between Models C and D. P-value standards: * (p < 0.05), ** (p < 0.01), *** (p < 0.001), ns (p > 0.05).

III. Clinical Translation Value Highlighted, Supporting Precision Treatment and Multi-Center Validation

01 83% Consistency Matching Clinical Dementia Scores, Demonstrating Broad Generalizability

The research team conducted multi-center validation among 100 AD patients at Tianjin Huanhu Hospital, dividing the EEG stratification into mild (A+B) and severe (C+D) groups and comparing with clinical dementia rating (CDR) scores, achieving an overall consistency of 83%, with 78.2% of EEG mild patients corresponding to CDR<1 (normal/MCI), highly aligned with the treatment window for current AD disease-modifying therapies, confirming the cross-center clinical applicability of this model.

02 Precisely Predicting rTMS Treatment Outcomes, Optimizing Patient Selection Strategies

As an important non-pharmacological treatment for AD, rTMS shows considerable individual variability in efficacy, and EEG stratification provides a key basis for addressing this issue. Analysis of 46 AD patients receiving rTMS treatment revealed:

◼ Patients in the EEG mild group (A+B) showed significant improvement in MoCA scores after rTMS treatment, with a standardized mean difference of 0.465, indicating significant treatment benefit;

◼ Patients in the EEG severe group (C+D) showed no significant improvement in cognitive scores after treatment, and even experienced a slight decline.

In contrast, traditional MoCA score-based stratification could not effectively identify rTMS treatment responders, while regression breakpoint design analysis further confirmed that the treatment predictive value of EEG stratification was statistically significant, making it an important tool for guiding precision rTMS treatment.

Figure 6:EEG stratification model, Alzheimer's disease (AD) and Parkinson's disease (PD) progression, and repetitive transcranial magnetic stimulation (rTMS). Based on the low-risk category defined by EEG patterns, representing key periods of disease progression and potential intervention. (a) Risk stratification based on EEG patterns. (b) Relationship between AD biomarkers and EEG patterns. (c) Relationship between PD progression and EEG patterns. (d) Low-risk patients defined by EEG patterns represent the optimal intervention window for rTMS treatment.

IV. Research Significance and Outlook, Ushering in a New Era of Precision AD Management

The core value of this study lies in, for the first time, upgrading EEG from a simple brain electrical detection tool to a unified stratification tool connecting AD clinical symptoms, molecular pathology, and treatment response, breaking through the limitations of traditional stratification systems that "emphasize clinical over pathology, and detection over application." As a non-invasive, repeatable, and easily disseminated detection method, the establishment of the EEG stratification model makes dynamic monitoring, early intervention, and precision treatment of AD possible: it can provide clinicians with an objective basis for identifying critical intervention periods, and also support clinical trials in optimizing patient enrollment criteria and reducing heterogeneity.

The research team also noted that the current study still has limitations such as limited sample size and missing biomarker data for some patients. Future work needs to expand the sample size, include more neurodegenerative disease types (such as frontotemporal dementia and Lewy body dementia), and further refine the model. At the same time, combining EEG stratification with artificial intelligence and imaging technologies is expected to build a more precise multimodal AD assessment system.

This EEG-based AD stratification study provides entirely new approaches for the precision diagnosis and treatment of neurodegenerative diseases, and also represents a key step forward in the application of non-invasive EEG technology in clinical AD management. As this model is further optimized and clinically disseminated, it may in the future change the landscape of AD screening, diagnosis, and treatment, bringing more personalized and effective medical solutions to the broad population of AD patients.