Process Design and Data-Driven Decision-Making for Brain Health Management - A Machine Learning Approach for Epilepsy Seizure Classification - Thesis

Aug 5, 2026

Upload by Riley

The study leverages visualization techniques and advanced machine learning models within digital care pathways to improve the categorization of epileptic episodes using EEG signal information. With epilepsy affecting over fifty million people worldwide and imposing significant burdens on healthcare systems, accurate diagnostic tooling is essential for effective care. The research builds a framework using Logistic Regression, SVM, KNN, Random Forest, and XGBoost combined with visual analytics, evaluating seizure versus multiple non-seizure conditions (eyes open/closed and tumor/healthy regions).

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages94

Reading Time3 h 57 min

AccessFree to download

1. Introduction

1.1 Background

1.2 Research problems and objectives

1.3 Research scopes

1.4 Significance of study

1.5 Expected contributions

1.6 Thesis outlines

2. Related Works

2.1 Epilepsy

2.2 Seizure detection and classification

Process Design and Data-Driven Decision-Making for Brain Health Management: A Machine Learning Approach for Epilepsy Seizure Classification University of Oulu Information Processing Science Master’s Thesis Md Mustofa Kamal Abstract The present study focuses on utilizing visualization techniques and sophisticated machine learning methods in digital care pathways for strengthening the categorization of epileptic episodes using electroencephalogram (EEG) signal information. Over fifty million people globally are affected by epilepsy, a prevalent neurological disease that places a heavy cost on healthcare systems. For this reason, accurate diagnostic tools are very necessary to provide optimal epilepsy care. This study aims to establish a comprehensive approach that utilizes the techniques of machine learning, such as Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) with visual analytics, for enhancing the categorization of epileptic seizures. Additionally, it facilitates a comprehensive evaluation of brain health by integrating cognitive abilities, brain imaging information, and factors associated with lifestyle. The main goal of this research is to develop visual analytics for decision-making and design processes to improve seizure detection accuracy by examining the distinctive features of non-seizure EEG episodes (brain tumor area, healthy part, eye closed, and eye open) and comparing these classes to seizure episodes. Binary classifiers are trained to evaluate the seizure class against each non-seizure class. The research study evaluates the consequences of incorporating non-seizure situations, namely seizure vs (eyes open and eyes closed) and seizure vs (tumor region and healthy region), and predicts tumors from brain healthy class on machine learning classification effectiveness. The study seeks to improve seizure detection and tumor accuracy by incorporating non-seizure EEG categories and examining its findings with those associated with classes. An interactive visual analytics dashboard is developed to guide and support medical practitioners in informed decision-making by displaying key metrics related to brain health, including cognitive function scores, imaging data, and lifestyle factors. The findings ofthis research contribute to personalized medication and advanced planning for epilepsy treatment, with the potential to significantly enhance patient health outcomes. Keywords Visual analytics, machine learning, digital care pathway, EEG, epilepsy seizure classification, brain health, LR SVM, RF, XGB, KNN, decision-making Supervisor Ph.D., Docent, Post-doctoral Researcher Pantea Keikhosrokiani Foreword First and foremost, I would like to start by conveying my deepest gratitude to Dr. Pantea Keikhosrokiani, my supervisor, whose invaluable advice, motivation, and constant encouragement have been crucial to my research endeavors. Her precise direction, insightful comments, and considerate mentorship have played a crucial role in shaping this thesis. Dr. Keikhosrokiani’s expertise and attention to detail set a high standard that continually inspired me to push the boundaries of my knowledge and skills. Her approachable and supportive nature fostered a collaborative and enriching research environment, for which I am truly grateful. Additionally, my sincere appreciation to my beloved parents and other family members for their unconditional support and love. Their relentless encouragement and faith in my capabilities instilled in me the confidence and drive to persist in my academic pursuits. I extend my profound gratitude for their sacrifices and contributions to this accomplishment. Lastly, my profound thanks to everyone who supported me since the beginning of this course of study, especially my friends, teachers, and fellow students. Your generosity, assistance, and suggestions have been instrumental, an

FAQs

What is the main goal of the thesis for epilepsy care?

To develop a comprehensive approach that uses machine learning and visual analytics to strengthen the categorization of epileptic episodes and improve seizure detection accuracy.

Which machine learning models are used for seizure and non-seizure classification?

Logistic Regression, Support Vector Machine, K-Nearest Neighbors, Random Forest, and eXtreme Gradient Boosting are used for evaluating seizure classification performance.

How does the study use non-seizure EEG categories in the evaluation?

Binary classifiers compare the seizure class against each non-seizure class, including eyes open/eyes closed and tumor region/healthy region, to assess their impact on classification effectiveness.

Aug 5, 2026

Upload by Riley