Dementia, particularly Alzheimer’s disease and its prodromal stage, mild cognitive impairment (MCI), is a major global health challenge. Early detection of MCI is crucial because it often precedes irreversible neurodegeneration, yet distinguishing it from later-stage dementia remains difficult due to overlapping symptoms and subtle early changes in brain function.
This book, "Graph Convolution Networks for EEG-Based Brain Network Biomarkers in Alzheimer’s Disease Progression", addresses this challenge by proposing analytical frameworks that reveal the underlying pathophysiological mechanisms of dementia-related disorders. Electroencephalography (EEG), with its accessibility and high temporal resolution, offers a practical window into neural activity, but its full potential emerges only when interpreted from a network-centric perspective.
Adopting a complex network approach, this work investigates EEG-derived brain functional networks (BFNs) in dementia. Using the Phase Lag Index (PLI) as the core connectivity metric, it constructs frequency-specific functional networks and applies a data-driven thresholding technique for robust, unbiased topology estimation. Quantitative and statistical network analyses show that graph-theoretic measures such as rich-club organization, transitivity, and assortativity provide effective biomarkers for differentiating MCI, Alzheimer’s disease, and vascular dementia.
Building on these insights, the BFNs are then used as structured graph inputs to a Graph Convolution Network (GCN) model. Integrating network neuroscience with deep learning, the proposed GCN framework achieves high classification accuracy (around 95%), highlighting the power of graph-learning methods for dementia staging.
Combining methodological rigor, theoretical depth, and practical evaluation, this book presents a unified framework for EEG-based brain network biomarker discovery. It is intended for researchers, clinicians, and students in computational neuroscience, biomedical signal processing, machine learning, and neurodegenerative disease research, and aims to contribute to earlier detection, better tracking, and deeper understanding of Alzheimer’s disease progression.
Table of Contents
1 Introduction
1.1 Study Background
1.1.1 Complex Networks in Medicine
1.1.2 Brain Functional Network
1.1.3 Dementia and Neurological Disorders: A Complex Network Perspective
1.2 Motivations
1.3 Problem Statements
1.4 Contributions
1.5 Organization
2 Literature Review
2.1 Complex Network Theory and its Interdisciplinary Applications
2.1.1 Graph theory approach
2.1.2 Connectivity in brain networks
2.1.3 Nodes/Vertices definition in brain network construction
2.1.4 Edges definition in brain network construction
2.1.5 Graph theoretical metrics for brain network analysis
2.2 Functional Brain Networks from Electrophysiological Signals
2.2.1 EEG as a tool for brain connectivity
2.2.2 Preprocessing and epoching strategies
2.2.3 Functional connectivity measures
2.3 Applications to Dementia-Related Disorders
2.3.1 Forms of dementia: A multifaceted spectrum
2.3.1.1 Alzheimer’s Disease (AD)
2.3.1.2 Vascular Dementia
2.3.1.3 Lewy Body Dementia (LBD)
2.3.1.4 Frontotemporal Dementia (FTD)
2.3.1.5 Mixed Dementia
2.3.2 Mild cognitive impairment and progression to AD
2.3.2.1 Mild cognitive impairment (MCI)
2.3.3 Challenges in dementia diagnosis
2.3.3.1 Diagnosing dementia
2.4 Identification of Dementia Disorders using Learning Techniques
2.4.1 Machine learning approaches for dementia recognition
2.4.2 Deep learning approaches for dementia recognition
2.4.3 Graph convolution network for dementia recognition
2.5 Summary
3 EEG-Based Brain Functional Network Analysis for Differential Identification of Dementia-Related Disorders and Their Onset
3.1 Overview
3.2 Materials and Methods
3.2.1 Phase lag index, PLI
3.2.2 Identification of band-specific most significant connections
3.2.3 Threshold selection and brain functional network formulation
3.2.4 Threshold validation
3.2.5 Brain functional network quantification
3.2.5.1 Rich-club coefficient:
3.2.5.2 Transitivity (functional segregation):
3.2.5.3 Assortativity coefficient (network resilience):
3.2.6 Graph convolution network
3.2.6.1 Graph learning
3.2.6.2 General representation of graph
3.2.6.3 Spectral graph filtering
3.2.6.4 Model architecture and initialization
3.3 Results
3.3.1 Phase lag index
3.3.2 Most significant electrode(s) and most significant connections identification
3.3.3 Threshold selection
3.3.4 Network topology quantification
3.3.5 Graph convolution network classification
3.3.5.1 Implementation details
3.3.5.2 Model performance
3.4 Discussion
3.5 Summary
4 Conclusions
Research Objectives and Core Themes
The primary objective of this book is to integrate complex network theory with deep learning models to analyze and identify neurological disorders, specifically focusing on Alzheimer’s disease, mild cognitive impairment (MCI), and stages of neuropathic pain. The work seeks to provide an analytical framework for biomarker discovery by processing EEG-derived brain functional networks through graph convolutional networks (GCNs) to achieve high diagnostic precision.
- Theoretical foundations of complex network analysis in medicine.
- Methodologies for EEG-based functional network construction using Phase Lag Index (PLI).
- Data-driven thresholding techniques for robust network characterization.
- Application of Graph Convolutional Networks for differential classification of dementia stages.
- Evaluation of network-based biomarkers for tracking neurodegenerative progression.
Excerpt from the Book
3.2.2 Identification of band-specific most significant connections
Brain Functional Networks (BFNs) are composed of nodes and connections, with varying centrality values indicating node contributions to network properties. Among these connections, certain ones, referred to as Most Significant Connections (MSCs), play a pivotal role in identifying network characteristics. In this study, MSCs are defined as those revealing substantial variation in functional connectivity measures between a neurological state and a reference state (see Fig. 3.1). Identifying MSCs involves considering node importance based on eigenvector centrality measures, a method widely used to pinpoint the most crucial nodes in a network. Eigenvector centrality accounts for both the number of connections a node possesses and the centrality of its connected nodes [88].
In this study, we introduce the algorithm for identifying ’n’ significant electrodes, outlined in Algorithm 1. The algorithm incorporates the concept of PLIsub(band), representing subject-specific PLI matrices from dementia-related groups, and PLIref(band), indicating the reference PLI matrix derived from the average PLI matrix of the normal control (NC) group. Utilizing the average PLI matrix of the NC group as a reference serves multiple purposes within our framework:
Establishing a Reasonable Baseline for Comparison: This choice provides a robust baseline, representing expected connectivity patterns in a healthy population. It enables quantifying differences between the connectivity of dementia-related subjects and that of the average NC group.
Quantifying Differences in Connectivity: The reference PLI matrix facilitates quantifying differences in connectivity, essential for identifying and characterizing abnormalities in the brain networks of dementia patients. This step is crucial for diagnosis and understanding disease mechanisms in conditions like MCI, AD, and VD.
Chapter Summaries
1 Introduction: Provides a comprehensive background on complex network theory as a framework for understanding brain functional networks and diagnosing dementia-related disorders.
2 Literature Review: Explores current scientific advancements in network neuroscience, detailing graph-theoretical measures and classification algorithms used in clinical settings.
3 EEG-Based Brain Functional Network Analysis for Differential Identification of Dementia-Related Disorders and Their Onset: Presents the primary methodological framework, including threshold selection, BFN construction, and the application of GCN-net to classify dementia stages.
4 Conclusions: Summarizes the key findings of the study, emphasizing the importance of methodological rigor and the potential of integrated deep learning approaches in clinical diagnostics.
Keywords
Dementia, Alzheimer’s Disease, Mild Cognitive Impairment, EEG, Brain Functional Networks, Graph Convolution Networks, Network Neuroscience, Phase Lag Index, Eigencentrality, Network Topology, Biomarkers, Machine Learning, Deep Learning, Neuropathic Pain, Functional Connectivity
Frequently Asked Questions
What is the core focus of this publication?
The book focuses on using complex network theory and deep learning to model and identify neurological disorders like dementia and neuropathic pain through the analysis of EEG-based brain functional networks.
Which specific conditions are analyzed in this research?
The research primarily analyzes Alzheimer’s disease (AD), mild cognitive impairment (MCI), vascular dementia (VD), and, in an extended application, chronic neuropathic pain.
What is the primary goal of the developed framework?
The primary goal is to provide a unified, analytical framework that uses EEG data and graph convolutional networks to detect subtle network disruptions for earlier and more accurate dementia diagnosis.
Which scientific method is central to this study?
The study employs a graph-theoretical approach, constructing brain functional networks using the Phase Lag Index (PLI) as a connectivity measure and leveraging Graph Convolutional Networks (GCNs) for automated classification.
What topics are covered in the main body of the work?
The main body covers literature on network neuroscience, the detailed formulation of BFNs, data-driven thresholding techniques, and the implementation and performance evaluation of GCN-based diagnostic models.
Which keywords define this work's academic scope?
Key terms include dementia, Alzheimer’s disease, brain functional networks, EEG, graph convolutional networks, and network neuroscience.
How does the proposed thresholding technique improve diagnosis?
It provides a data-driven, unbiased method that preserves significant connections based on eigencentrality, allowing for more reliable topological comparisons between patient groups and healthy controls.
What do the results indicate regarding delta band performance?
The study found that the delta frequency band yielded the highest accuracy (95.07%) and F1-scores, suggesting that it is particularly significant for identifying the onset of dementia-related disorders.
- Arbeit zitieren
- Abdulyekeen T. Adebisi (Autor:in), Kalyana C. Veluvolu (Autor:in), 2025, EEG-Based Brain Network Biomarkers in Alzheimer's Disease Progression, München, GRIN Verlag, https://www.grin.com/document/1676601
