Dementia-related disorders—including Alzheimer’s disease (AD), vascular dementia (VD), and mild cognitive impairment (MCI)—remain among the most challenging neurological conditions due to their overlapping clinical characteristics and complex neurocognitive manifestations. The difficulty in distinguishing these disorders continues to hinder therapeutic interventions, underscoring the need for analytical tools capable of revealing subtle but meaningful differences in brain network organization.
This book presents an in-depth exploration of brain functional network analysis as a means to address these challenges. Leveraging phase-based mutual information as the connectivity measure, we examine the functional architecture of the brain across healthy controls and individuals diagnosed with MCI, AD, and VD. Using a carefully selected reactive frequency band, we identify consistent patterns of hypoconnectivity across the dementia groups, providing insights into shared and divergent network alterations.
Two complementary computational frameworks form the core of this work: a frequent subgraph mining approach for the extraction of recurrent network motifs, and a minimum spanning tree (MST) method for structural characterization of both unweighted and weighted brain networks. The frequent subgraphs derived from dementia-specific networks are further applied for classification tasks, while MST-based features and additional graph-theoretic measures are used to delineate disease-specific network signatures.
Table of Contents
1 Introduction
2 Time-Frequency Mapping of Phase-based Mutual Information
2.1 Instantaneous Amplitude and Phase Extraction
2.1.1 Short-Time Fourier Transformation
2.1.2 Hilbert Transform
2.1.3 Continuous Wavelet Transform
2.2 Phase-based Mutual Information
2.3 Reactive Band Identification
3 Functional Connectivity Analysis
3.1 Loss of Functional Connection in Dementia Groups
3.2 Inter-Channel Connection Variation Analysis
4 Brain Functional Network Analysis
4.1 Unweighted Brain Network Analysis Using Frequent Subgraph Approach
4.1.1 Formation of Unweighted Brain Functional Network
4.1.2 Frequent Edges Search
4.2 Weighted Brain Network Analysis Using Minimum Spanning Tree Approach
4.2.1 Minimum Spanning Tree Formation
4.2.2 Network Topology Quantification
5 Results
5.1 Data Description
5.2 Instantaneous Amplitude and Phase Extraction
5.3 Synchronization Analysis in Time-Frequency
5.3.1 Phase-based Mutual Information in Time Frequency
5.3.2 Common Reactive Band
5.4 MI-based Functional Connectivity Analysis
5.4.1 Loss of Connection in Dementia Related Groups Compared to HC Group
5.4.2 Average inter-channel connectivity analysis
5.5 Unweighted Brain Functional Network Analysis
5.5.1 Common Threshold Identification
5.5.2 MI-Based Frequent Network Edges Search
5.5.3 Classification
5.6 Weighted Brain Functional Network Analysis
5.6.1 Minimum Spanning Tree Formation
5.6.2 Quantification of MST Topology
5.7 Discussion
6 Conclusion
Objectives and Research Themes
This work aims to enhance the diagnosis of dementia-related disorders—specifically Mild Cognitive Impairment (MCI), Alzheimer’s disease (AD), and Vascular dementia (VD)—by leveraging brain functional network analysis derived from EEG signals. The primary research goal is to identify distinct neurophysiological signatures and hypoconnectivity patterns that can accurately discriminate between healthy controls and various stages of dementia.
- Application of phase-based mutual information as a sensitive connectivity measure.
- Identification of a "reactive frequency band" to isolate significant network variations.
- Utilization of frequent subgraph mining and minimum spanning tree (MST) methodologies for structural network characterization.
- Development of automated classification frameworks to improve diagnostic precision using network-driven features.
Excerpt from the Book
1 Introduction
Ever since the of its exploration, the brain has been found to be arguably the most important organ in the nervous system [1]. It controls and coordinates the activities of other organs in the body system. It is also the most complex body organ consisting of billions of tiny cells called nerve cells which are the building blocks of the brain functionalities [2], [3]. The nerve cells are highly interconnected together in networks. In fact, the brain has several specialized regions designated for carrying out specific activities and functionalities [4], [5], [6]. Naturally, human brain is programmed into functional networks that actively enhance brain functioning such as the processing of sensory stimuli, cognition as well as some goal-directed tasks [7]. A lot of such networks have been identified using neuroimaging modalities at resting stage [8]. This has enabled the bridging of the gap in understanding the mechanisms behind some of those brain functionalities. Nevertheless, a lot of the underlie mechanisms of brain functionalities especially in terms of the various brain disorders are yet to be fully understood [9]. Therefore, the field of computation neuroscience continues to gain multiple interest day in, day out.
Neurological disorders are the abnormalities that affect the central nervous system as a result of loss of neurons and or glial cells [10]. Neural disorders could arise as a result of physical, environmental, genetic or some other factors [11]. Moreso, some of these disorders are age related, thus, they mostly affect people within a specific age group. Neurodegenerative disorders are the forms of neural disorders associated with neuronal loss [12]. Dementia are forms of neurodegenrative disorders that affect the aged adults whose ages are 65 years and above.
Summary of Chapters
1 Introduction: Provides an overview of brain complexity, the necessity of neuroimaging for studying dementia, and the book's scope regarding EEG-based network analysis.
2 Time-Frequency Mapping of Phase-based Mutual Information: Details spectral analysis techniques, specifically focusing on instantaneous amplitude and phase extraction to compute mutual information.
3 Functional Connectivity Analysis: Explores methods for measuring connection loss in dementia subjects and analyzes inter-channel connectivity variations.
4 Brain Functional Network Analysis: Discusses the implementation of frequent subgraph mining and minimum spanning tree approaches to quantify network topology.
5 Results: Presents the experimental findings, including data processing, threshold selection, classification performance, and statistical comparisons of MST metrics.
6 Conclusion: Summarizes the effectiveness of the proposed framework and its potential as a clinical tool for diagnostic support.
Keywords
Dementia, Alzheimer's disease, Mild cognitive impairment, Vascular dementia, EEG, Mutual information, Brain functional networks, Frequent subgraph mining, Minimum spanning tree, Graph theory, Neuroengineering, Hypoconnectivity, Spectral analysis, Feature extraction, Clinical diagnostics.
Frequently Asked Questions
What is the core focus of this research?
The research focuses on using computational neuroengineering to analyze brain functional networks from EEG data to help differentiate between dementia-related disorders.
Which specific conditions are analyzed?
The work investigates Mild Cognitive Impairment (MCI), Alzheimer’s disease (AD), and Vascular dementia (VD) in comparison with a Healthy Control (HC) group.
What is the primary diagnostic goal?
The goal is to improve diagnostic precision and clinical decision-making by revealing subtle, disease-specific network alterations that are currently difficult to distinguish.
What methodology is employed to analyze brain signals?
The research employs phase-based mutual information as a connectivity measure, combined with frequent subgraph mining and minimum spanning tree (MST) topological analysis.
How is the "reactive band" identified?
The reactive band is defined as the frequency range showing the maximum differential mutual information between dementia groups and the healthy control group.
What do the results regarding frequent edges suggest?
The results show that the number of frequent edges decreases with the severity of cognitive impairment, suggesting that frequent edge density serves as a biomarker for disease progression.
Why are MST metrics utilized in this study?
MST metrics provide a robust way to quantify network topology without the biases introduced by arbitrary thresholding in unweighted networks.
What was the outcome of the classification performance?
By using an ensemble of phase-based mutual information and phase locking value (PLV) features, the study achieved a classification accuracy of 96.47%.
- Quote paper
- Abdulyekeen T. Adebisi (Author), Kalyana C. Veluvolu (Author), 2026, EEG-Based Computational Neuroengineering for Dementia Diagnosis, Munich, GRIN Verlag, https://www.grin.com/document/1689170
