Predicting Early Stages Of Alzheimer's Disease And Identifying Key Biomarkers Using Deep Artificial Neural Network And Ensemble Of Machine Learning Methodologies

Jul 16, 2026

Alzheimer's disease (AD) is a slowly progressing neurological disorder that begins with mild memory loss and later impairs everyday functioning by affecting brain regions responsible for thinking, memory, and language. Disease onset and symptom progression are highly unpredictable, while dementia is often mistaken for normal aging, increasing adverse outcomes. The study proposes an automatic early-stage diagnostic model using clinical, neuropsychological, and radiological measures. It addresses missing data via iterative imputation, class imbalance via Borderline SVM-SMOTE, and selects informative features through wrapper and embedded methods.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages233

Reading Time9 h 47 min

Chapter 1

Introduction

Chapter 2

Literature Review

PREDICTING EARLY STAGES OF ALZHEIMER'S DISEASE AND IDENTIFYING KEY BIOMARKERS USING DEEP ARTIFICIAL NEURAL NETWORK AND ENSEMBLE OF MACHINE LEARNING METHODOLOGIES DEBOPRIYA GHOSH Master of Science (MSc), Thesis Report MARCH 2022 ABSTRACT Alzheimer's disease (AD) is a neurological disorder that begins slowly with some mild symptoms of memory loss and subsequently affects an individual’s ability to accomplish everyday activities. Alzheimer's disorder affects different areas of the brain that manage thinking, memory, and linguistics. It is the most common form of dementia, the cause of which little is known to scientists. Although it is most common in older people, it often affects younger people as well. The main problem with AD is that the onset, progression, and course of symptoms are highly unpredictable and variable. Along with physical and mental issues, patients with the disease and their families face social issues like isolation, unpredictability, fear, fatigue. There is also a huge economic impact of dementia relating to medical and social care expenditures. Many times, dementia is misdiagnosed as a sign of ageing by physicians, and the severity of the disease becomes more adverse when the person is actually detected with dementia. At present, no drugs are available to prevent this disease, but some symptom-delaying drugs are available that can slow the disease's progression to some extent. Therefore, an automatic early-stage diagnostic system will help physicians control the disease in its primitive phase by taking preventive measures. The purpose of the study is to propose a model which can accurately diagnose the different early stages of Alzheimer's Disease by analysing clinical information, neuropsychological and radiological test measures. This study also helps clinicians to understand the significant biomarkers associated with Alzheimer's disorder. A dataset has been collected from the Alzheimer’s Disease Neuroimaging Initiative study data for this research work. This dataset contains missing values, which will be treated with an iterative imputation technique. Being a healthcare dataset, a class imbalance problem exists in this, which is addressed by the borderline SVM-SMOTE algorithm. Feature selection needs to be performed as fitting a machine learning model with irrelevant features leads to inaccurate detection. This study uses wrapper-based and embedded feature selection techniques to obtain the most significant feature set. A train-test split needs to be carried out on selected predictors, and feature scaling is done to make the features scale free. After completing all data preprocessing tasks, a stacking-based ensemble model, comprised of Logistic Regression, Extra Tree, Bagging KNN and LightGBM classifiers as a base estimator, is designed. A deep learning model (ANN) is also implemented on the same dataset, which may increase the accuracy. Then, a comparative analysis is performed among the above models by evaluating different performance metrics, namely, precision, recall, f1-Score, and AUC-ROC, to determine the best classifier along with the most important biomarkers for early diagnosis of AD. TABLE OF CONTENTS ABSTRACT .............................................................................................................................................. ii LIST OF TABLES ................................................................................................................................... vii LIST OF FIGURES .................................................................................................................................. ix LIST OF ABBREVIATIONS .................................................................................................................... xii CHAPTER 1 ............................................................................................................................................ 1 INTRODUCTION ................

FAQs

What is the main goal of the study on Alzheimer's disease?

To build a model that accurately diagnoses different early stages of Alzheimer's disease and helps identify significant associated biomarkers using clinical, neuropsychological, and radiological data.

How does the study handle missing values and class imbalance in the dataset?

Missing values are treated with an iterative imputation technique, and class imbalance is addressed using the Borderline SVM-SMOTE algorithm.

Which modeling approaches are compared for early AD diagnosis?

The study compares a stacking-based ensemble model using Logistic Regression, Extra Tree, Bagging KNN, and LightGBM, alongside a deep learning ANN model, evaluated with precision, recall, F1-score, and AUC-ROC.

Jul 16, 2026

Upload by Ava Thompson