Towards Diagnosis of Dementia - Microstate Analysis of EEG Signals and Classification using Machine Learning

Aug 3, 2026

Dementia is a neurodegenerative disorder marked by cognitive decline that creates major pressure on healthcare systems worldwide. Timely and accurate detection supports better intervention and management strategies. This thesis evaluates electroencephalography (EEG), a noninvasive and cost-effective technique, and investigates EEG microstates as features for traditional machine learning to detect dementia and related stages. Two pipelines are developed: conventional classification from microstate features and deep learning classification from topographic images derived from GFP peaks. Models including SVM, Random Forest, and XGB are used, achieving strong multiclass and binary results, while a lightweight five-layer 2D CNN further improves performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages72

Reading Time3 h 1 min

AccessFree to download

Abstract

Acknowledgments

Introduction

Electroencephalography

Foundations and Theory

Literature Review

Theoretical Foundation

Data

Dataset

Dataset Features

Methodology

Data Processing

Signal Preprocessing Activities

Determining Microstates

Extracting Topographic Images for Deep Learning

Machine Learning Models

Discussion and Results

Towards Diagnosis of Dementia: Microstate Analysis of EEG Signals and Classification using Machine Learning Mohammad Mehedi Hasan Thesis submitted for the degree of Master in Applied Computer and Information Technology-ACIT (Data Science) 30 credits Department of Computer Science Faculty of Technology, Art and Design Oslo Metropolitan University — OsloMet Spring 2024 Towards Diagnosis of Dementia: Microstate Analysis of EEG Signals and Classification using Machine Learning Mohammad Mehedi Hasan © 2024 Mohammad Mehedi Hasan Towards Diagnosis of Dementia: Microstate Analysis of EEG Signals and Classification using Machine Learning [http://www.oslomet.no/](http://www.oslomet.no/) Printed: Oslo Metropolitan University — OsloMet Abstract Dementia is a neurodegenerative disorder characterized by cognitive decline which presents significant challenges to healthcare systems around the world. Early and accurate detection is important for effective intervention and management. Electroencephalography (EEG) provides a noninvasive and cost-effective technique for assessing neurological conditions. This thesis explores the use of EEG microstates as features for traditional machine learning to detect dementia. The study develops two machine learning techniques: one employing traditional machine learning for microstate features and the other training a deep learning model using topoplot images based on GFP peaks. EEG microstates are quasi-stable states representing transient functional brain activities that provide valuable insights into brain activity associated with dementia. By leveraging both the temporal and spatial resolution of EEG microstates, the analysis reveals patterns linked to dementia. Traditional machine learning models such as Support Vector Machines (SVM), Random Forests (RF), and eXtreme Gradient Boosting (XGB) were deployed to classify Dementia, Mild Cognitive Impairment (MCI), and healthy individuals (Normal) . In addition, we proposed extracting topographic images from EEG signals and utilizing a deep learning model to classify EEG recordings. The proposed five-layer 2D Convolutional Neural Network (CNN) is a simple, lightweight model. The CNN kernel captures spatial features from images, enhancing the model’s ability to differentiate between dementia-related brain dysfunction and healthy brain activity. The model aims to classify EEG signals into three categories: Dementia, MCI, and healthy individuals. Furthermore, we compared our model with baseline results from previous work conducted on the publicly available CAUEEG dataset. Our findings demonstrate the effectiveness of the traditional model using microstate features, achieving 74 % accuracy for classifying Dementia, MCI, and Normal, and 76% for binary classification. The proposed deep learning model achieved 82% accuracy for multiclass classification. Overall, this thesis contributes to dementia diagnosis by leveraging EEG microstate analysis and advanced machine learning techniques. Acknowledgments I am profoundly grateful to my supervisor Rabindra Khadka for his continuous support, and guidance throughout the period; and providing me the flexibility to work on EEG signal analysis. I would like to also thank Prof Anis Yazidi and Prof Pedro Lind for inspiring me to work in the field of computer vision and for their valuable feedback and suggestions. I am also deeply grateful to all of my university professors, also like to thank all my classmates for being so cooperative. Nevertheless all my gratitude and thanks to Oslo Metropolitan University for giving me this opportunity to study in this wonderful environment. I acknowledge that the research has benefited from the work done by many other researchers in the field of neuroscience particularly in EEG microstate. Finally, I would like to dedicate this thesis work to my parents and family for always being there for me with complete love and support. Contents Abstract i

FAQs

Why is early detection of dementia important in this thesis?

Early and accurate detection enables more effective intervention and management, addressing major challenges for healthcare systems. The thesis focuses on improving diagnostic capability using EEG analysis.

How are EEG microstates used for dementia classification?

EEG microstates are treated as quasi-stable representations of transient functional brain activity. The thesis extracts microstate features and trains traditional machine learning models to classify Dementia, MCI, and Normal subjects.

What deep learning approach is proposed for EEG-based diagnosis?

Topographic images are extracted from EEG signals using GFP peaks. A lightweight five-layer 2D CNN is trained to classify EEG recordings into three categories: Dementia, MCI, and healthy individuals, and is compared with baseline work on the CAUEEG dataset.

Aug 3, 2026

Upload by Theodore