ENHANCING ALZHEIMER'S DIAGNOSIS: MRI

ANALYSIS WITH GENERATIVE CHATBOT SUPPORT

A PROJECT REPORT

Submitted by

AMREEN.R [211420243003]

in partial fulfillment for the award of the degree

of

BACHELOR OF TECHNOLOGY

IN

ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

PANIMALAR ENGINEERING COLLEGE

(An Autonomous Institution, Affiliated to Anna University, Chennai)

MARCH 2024
 PANIMALAR ENGINEERING COLLEGE
(An Autonomous Institution, Affiliated to Anna University, Chennai)

BONAFIDE CERTIFICATE

Certified that this project report “ENHANCING ALZHEIMER'S DIAGNOSIS:

MRI ANALYSIS WITH GENERATIVE CHATBOT SUPPORT” is the
bonafide work of “AMREEN.R [211420243003]” who carried out the project work
under Dr. S. CHAKARAVARTHI, M.E., Ph.D., supervision.

SIGNATURE SIGNATURE
Dr. S. MALATHI, M.E., Ph.D. Dr.S.CHAKARAVARTHI,M.E.,Ph.D.
HEAD OF THE DEPARTMENT PROFESSOR

DEPARTMENT OF AI&DS, DEPARTMENT OF AI&DS,

PANIMALAR ENGINEERING COLLEGE, PANIMALAR ENGINEERING COLLEGE,
NASARATHPETTAI, NASARATHPETTAI,
POONAMALLEE, POONAMALLEE,
CHENNAI-600 123. CHENNAI-600 123.

Certified that the above-mentioned student was examined in End Semester Project
Work (AD8811) viva-voice held on .

INTERNAL EXAMINER EXTERNAL EXAMINER

 DECLARATION BY THE STUDENT

I AMREEN.R [211420243003], hereby declare that this project report titled

“ENHANCING ALZHEIMER'S DIAGNOSIS: MRI ANALYSIS WITH

GENERATIVE CHATBOT SUPPORT”, under the guidance of Dr. S.

CHAKARAVARTHI M.E., Ph.D., is the original work done by me and I

have not plagiarized or submitted to any other degree in any university by

me.

AMREEN.R

i
 ACKNOWLEDGEMENT

We would like to express our deep gratitude to our respected Secretary and

Correspondent Dr. P. CHINNADURAI, M.A., Ph.D. for his kind words and

enthusiastic motivation, which inspired us a lot in completing this project.

We express our sincere thanks to our directors Tmt. C. VIJAYARAJESWARI,

Dr. C. SAKTHI KUMAR, M.E., Ph.D. and Dr. SARANYASREE SAKTHI

KUMAR B.E., M.B.A., Ph.D., for providing us with the necessary facilities to

undertake this project.

We also express our gratitude to our Principal Dr. K. MANI, M.E., Ph.D. who

facilitated us in completing the project.

We thank the Head of the AI&DS Department, Dr. S. MALATHI, M.E., Ph.D.,

for the support extended throughout the project.

We would like to thank my Project Guide Dr. S. CHAKARAVARTHI M.E., Ph.D.,

and all the faculty members of the Department of AI&DS for their advice and

encouragement for the successful completion of the project.

ii
 ABSTRACT

Alzheimer's disease is a neurological condition that worsens with time and is

characterized by memory loss and a reduction in cognitive function. It is the most
prevalent cause of dementia, impacting millions of individuals globally.
Alzheimer's disease usually affects elderly folks, however it can also strike
younger people on occasion. Diagnostic methods for Alzheimer's disease (AD)
frequently combine psychological assessments with neuroimaging modalities like
magnetic resonance imaging (MRI). This project extracts complex patterns and
aspects of Alzheimer's disease from MRI brain images using deep learning
methods. Complex spatial and temporal information is processed using
convolutional neural networks. The goal is to develop a robust diagnostic tool for
identifying subtle neuroanatomical changes. to augment the diagnostic system's
capabilities, a prediction mechanism is incorporated. This mechanism enables the
system to anticipate disease progression, empowering healthcare providers to
implement proactive management strategies and personalize treatment plans for
patients with AD. A generative chatbot interface based on LangChain architecture
and driven by the llama2 language model has been introduced to enhance the
diagnostic process. This project represents a holistic approach to addressing the
challenges posed by Alzheimer's disease, combining cutting-edge technology,
innovative interfaces, and predictive analytics to revolutionize diagnosis,
management, and patient care. By leveraging deep learning methods, enhancing
communication through a generative chatbot interface, and incorporating
predictive capabilities and real-time updates, the project aims to significantly
improve outcomes for individuals affected by AD and transform the landscape of
dementia care.

iii
 TABLE OF CONTENTS

CHAPTER NO. TITLE PAGE NO.

ABSTRACT iii

LIST OF TABLES vii

LIST OF FIGURES viii

LIST OF ABBREVIATIONS ix

1. INTRODUCTION 1

1.1 Overview 3

1.2 Problem Definition 4

1.3 Objective 4

2. LITERATURE SURVEY 6

3. SYSTEM ANALYSIS 10

3.1 Existing System 11

3.2 Proposed system 12

3.2.1 Data Collection 13

3.2.2 Data Processing 14

3.2.3 Data Exploration and Visualization 14

3.2.4 Feature scaling 15

3.2.5 Training and testing data 16

3.2.6 Model Building 16

iv
CHAPTER NO. TITLE PAGE NO.

3.3 Feasibility Study 17

3.3.1 Social Feasibility 17

3.3.2 Economic Feasibility 17

3.4 Software Environment 17

3.5 H a r d ware Environment 18

4. DATASETS 19

5. SYSTEM DESIGN 22

5.1 ER diagram 23

5.2 Class Diagram 24

6. SYSTEM ARCHITECTURE 25

6.1 Module Design Specification 27

6.2 Algorithm 29

6.2.1 Logistic Regression 30

6.2.2 Ridge Classifier 31

7. EVALUATION METRICS 32

7.1 Confusion Matrix 33

7.2 Accuracy 34

TITLE PAGE NO.

v
CHAPTER NO.

7.3 Recall 34

7.4 Precision 34

7.5 F1 Score 34

8. SYSTEM IMPLEMENTATION 35

8.1 Sample code 36

9. SYSTEM TESTING 40

9.1 Alzheimer Detection System 41

9.2 Alzheimer Prediction System 41

10. RESULTS AND DISCUSSION 42

10.1 Results and Discussions 43

11. CONCLUSION 46

11.1 Conclusion 47

11.2 Future Work 48

12. APPENDICES 49

A.1 Sample Screens 50

13. REFERENCES 53

vi
 LIST OF TABLES

TABLE NO. TITLE PAGE NO.

10.1 Classification report for Logistic Regression 43

10.2 Training and Validation accuracy graph of CNN 44

LIST OF FIGURES

vii
FIGURE NO. TITLE PAGE NO.

viii
 3.1 System Architecture for proposed system 11

3.2.3a Relationship between Age and Diagnosis 15

3.2.3b Relationship between APOE Genotype and 15

Diagnosis
4.1 Dataframe created from the ADNI dataset 20

4.2 Images from the MRI scans dataset 21

5.1 System ER Diagram 23

5.2 Class Diagram 24

6.1 System Architecture 26

6.2.1 Logistic Regression 30

7.1a Confusion matrix for Logistic Regression 33

7.1b Confusion matrix for Ridge Classifier 33

10.1 Training and validation accuracy graph of 44

CNN.
A.1 Screen Shot of Home Page 50
A.1 Screen Shot of Alzheimer Detection System 50
A.1 Screen Shot of Prediction System 51
A.1 Screenshot of Generative Chatbot 51

A.1 Screenshot of Real-time News Update 52

ix
 LIST OF ABBREVIATIONS

Abbreviation Meaning

AD Alzheimer’s Disease
MRI Magnetic Resonance Imaging
CNN convolutional neural networks
CN Cognitively Normal
MCI Mild Cognitive Impairment
EMCI Early Mild Cognitive Impairment
LMCI Late Mild Cognitive Impairment
MMSE Mini-Mental State Exam
PET Positron Emission Tomography
LLM Large Language Model
ADNI Alzheimer's Disease Neuroimaging Initiative
NLP Natural Language Processing
MLP Multilayer perceptron

x
 CHAPTER 1

INTRODUCTION

1
 CHAPTER 1

INTRODUCTION

Alzheimer's disease is a neurological condition that worsens over time,

eventually impairing thinking, memory, and even the capacity to do basic tasks. It
is the most frequent cause of dementia, a word used to describe a loss of cognitive
skills such as memory, language, and problem-solving that is severe enough to
interfere with day-to-day activities. Timely detection, ongoing support, access to
current information, and predictive insights are essential components for effective
management and care. In response to these critical needs, our project introduces a
comprehensive Alzheimer's Care Ecosystem. This ecosystem integrates cutting-
edge technologies to provide a holistic approach to Alzheimer's care.

This project combines advanced deep learning techniques, specifically

Convolutional Neural Networks (CNNs), for accurate detection of Alzheimer's
disease from medical imaging data. Alongside CNN-based detection, we
incorporate the assistance of a sophisticated medical chatbot to provide
personalized support and information to patients, caregivers, and healthcare
professionals. Furthermore, the system offers real-time updates on the latest news
and research in Alzheimer's disease, ensuring stakeholders stay informed about
developments in the field. Finally, predictive analysis tools are employed to
anticipate disease progression and aid in treatment planning.

By combining these elements into a unified platform, our project aims to

revolutionize Alzheimer's care by providing comprehensive support, timely
information, and proactive management strategies. Through this interdisciplinary
approach, we aspire to improve patient outcomes, enhance caregiver support, and
contribute to advancements in Alzheimer's research and treatment.

2
1.1 OVERVIEW

The project aims to develop a comprehensive solution for Alzheimer's disease

management, integrating various technological components to enhance early
detection, support, and information dissemination. Initially, the project will focus
on leveraging deep learning techniques, specifically Convolutional Neural
Networks (CNNs), for accurate detection of Alzheimer's disease using structural
MRI data. This segment involves training CNN models on a dataset comprising
both healthy aging subjects and individuals diagnosed with Alzheimer's disease,
followed by thorough evaluation to assess performance metrics. Additionally, the
project will include the development of a medical chatbot tailored for Alzheimer's
support, providing personalized assistance, information, and guidance to patients,
caregivers, and healthcare professionals. Alongside this, a real-time news
aggregation system will be implemented to keep stakeholders informed about the
latest research findings, treatment options, and news related to Alzheimer's
disease. Furthermore, predictive analytics will be employed to anticipate disease
progression, enabling proactive treatment planning and management. The
integration and deployment phase will involve merging all project components
into a unified platform, addressing technical considerations, and ensuring
scalability and usability. Through case studies and use cases, the project will
showcase the practical applications and effectiveness of each component,
followed by a thorough assessment of impact and user feedback analysis.
Ultimately, the project aims to contribute to improved Alzheimer's care by
providing early detection, personalized support, up-to-date information, and
predictive insights, thereby enhancing patient outcomes and advancing research
in the field.

3
1.2 PROBLEM DEFINITION

The need for Alzheimer's detection is paramount given the profound impact of
this neurodegenerative disease on individuals, families, and society. Alzheimer's
is a progressive condition that gradually impairs cognitive function, memory, and
daily living activities, ultimately leading to severe disability and dependency.
Early detection of Alzheimer's is crucial as it allows for timely intervention and
management strategies to slow disease progression and improve quality of life for
affected individuals. Additionally, early diagnosis enables patients and their
families to make informed decisions regarding care planning, access support
services, and participate in clinical trials for potential treatments. This
encompasses a multifaceted approach that incorporates the integration of various
data sources, including medical records, imaging studies, genetic information, and
cognitive assessments. The overarching goal is to construct predictive models
capable of reliably diagnosing Alzheimer's disease at its incipient stages, thus
facilitating timely intervention and personalized treatment strategies.
Furthermore, the problem statement includes the exploration of novel biomarkers,
risk factors, and disease mechanisms through advanced data analytics and
machine learning techniques. By addressing these challenges, the aim is to
enhance our understanding of Alzheimer's disease, improve early detection rates,
and ultimately mitigate its impact on individuals, families, and society as a whole.

1.3 OBJECTIVE
The primary objective of this project is to advance the diagnostic capabilities for
Alzheimer's disease (AD) by leveraging cutting-edge technology and
methodologies. By utilizing deep learning techniques, particularly convolutional
neural networks (CNNs), complex spatial and temporal information inherent in
MRI brain images can be effectively processed and analyzed. This approach
4
enables the extraction of subtle neuroanatomical changes associated with AD,
thereby facilitating early detection and intervention. Moreover, the integration of
a generative chatbot interface, based on LangChain architecture and powered by
the llama2 language model, serves to enhance the interaction between users,
including physicians and patients, and the diagnostic system. This intuitive
interface fosters real-time communication, allowing for the interpretation of
diagnostic data and provision of immediate feedback, thereby improving the
overall diagnostic process. Furthermore, the incorporation of a prediction
mechanism aims to forecast disease progression, enabling proactive management
strategies. Additionally, real-time news updates ensure that the diagnostic system
remains informed of the latest advancements and findings in AD research,
thereby continually enhancing its efficacy and relevance in clinical practice.
Overall, this project represents a multidimensional approach to advancing
Alzheimer's disease diagnosis, combining state-of-the-art technology with
innovative communication interfaces and predictive capabilities to improve
patient outcomes and clinical decision-making.

5
 CHAPTER 2

LITERATURE SURVEY

6
 CHAPTER 2

LITERATURE SURVEY

 Tetiana Habuza et al. [1], The survey explores the use of deep learning in
analyzing brain scans and cognitive tests for Alzheimer's disease (AD) detection. It
discusses the urgency of AD detection, deep learning's potential in medical imaging
analysis, and the role of MRI and cognitive tests in AD diagnosis. It acknowledges
limitations like data needs and potential bias.

 Al Shehri W. [2], The study proposes a deep learning-based solution for

Alzheimer's disease diagnosis and classification, using DenseNet-169 and ResNet-
50 convolutional neural network architectures. The study aims to improve accuracy
and efficiency in early detection, as manual diagnosis remains error-prone and
time-consuming. The model shows superior performance, demonstrating high
accuracy values in training and testing phases. This model offers promise for real-
time analysis and classification, contributing to advancements in early diagnosis
solutions.

 Cristina L. Saratxaga et al. [3], This paper reviews current methods for diagnosing
Alzheimer's Disease (AD), focusing on limitations of traditional cognitive tests and
the role of Magnetic Resonance Imaging (MRI) in identifying brain abnormalities. It
highlights recent advancements in deep learning, particularly Convolutional Neural
Networks (CNNs), and their potential for AD diagnosis. The review acknowledges
challenges like large datasets and potential biases in deep learning algorithms. The
aim is to establish the need for improved AD diagnosis and explore deep learning's
potential in MRI analysis.

 Nasir Rahim et al. [4], The study discusses advancements in Alzheimer's disease
diagnosis and prognosis using deep learning techniques, multimodal data fusion,
longitudinal analysis, and explainable AI methods. Researchers use datasets like
ADNI to develop models that analyze MRI scans to identify patterns associated

7
 with AD progression. Longitudinal studies track changes over time, while
explainable AI helps clinicians interpret model predictions. This integrated
approach improves early detection and understanding of AD's underlying
mechanisms.

 Yan Zhao et al. [5], The study suggests that integrating PET/MR imaging with
deep learning (DL) can improve Alzheimer's disease diagnosis and treatment.
Despite challenges like data heterogeneity and interpretability, DL has shown
potential in improving efficiency and quality of AD imaging. Future research should
address these challenges to fully utilize DL's capabilities in improving AD
diagnosis, prognosis, and treatment strategies.

 Naveen Sundar Gnanadesigan et al. [6], proposed the DC-GC model that is a
novel method for identifying candidate genes associated with Alzheimer's disease
(AD) using network topology measures and machine learning techniques. This
model outperforms existing classifiers like ANN, KNN, SVM, and decision trees in
identifying candidate genes. It ranks genes based on their connectivity and
physicochemical properties, improving the identification of potential AD-related
genes. The model's promising results suggest it can advance our understanding of
AD pathogenesis and facilitate the development of targeted therapeutic
interventions, thereby improving the development of effective therapies.

 Zhen Zhao et al. [7], This paper reviews various machine learning techniques for
Alzheimer's disease classification and prediction, including SVM, RF, CNN,
Autoencoder, Deep Learning, and Transformer. It discusses feature extractors and
input formats, and addresses challenges like class imbalance and data leakage. The
review also discusses pre-processing techniques and trade-offs between deep
learning and conventional methods. It offers insights for addressing these issues,
exploring new techniques, and selecting appropriate input types for optimal AD

8
 diagnosis and prediction.

 Pradnya Borkar et al. [8], This research aims to develop noninvasive and cost-
effective methods for early detection of Alzheimer's disease (AD). By analyzing
MRI scans and extracting brain characteristics, a model based on convolutional
neural networks and long short-term memory networks is trained. The model
outperforms current diagnostic methods, providing high accuracy (99.7%) while
remaining noninvasive and cost-effective. This innovative approach contributes to
the growing literature on deep learning for early detection and intervention in AD,
offering hope for improved patient outcomes.

 Daichi Shigemizu et al. [9], The study explores the genetic structure of late-onset
Alzheimer's disease (LOAD) using genome-wide association study data from
Japanese cohorts. Two distinct groups of LOAD patients were identified: one
exhibited risk genes for LOAD development, immune-related genes, and another
displayed genes associated with kidney disorders. Impaired kidney function was
identified as a potential contributor to LOAD pathogenesis. Researchers developed a
prediction model using a deep neural network, providing new insights into LOAD's
pathogenic mechanisms.

 Shangran Qiu et al. [10], The study explores the genetic structure of late-onset
Alzheimer's disease (LOAD) using genome-wide association study data from
Japanese cohorts. Two distinct groups of LOAD patients were identified: one
exhibited risk genes for LOAD development, immune-related genes, and another
displayed genes associated with kidney disorders. Impaired kidney function was
identified as a potential contributor to LOAD pathogenesis. Researchers developed a
prediction model using a deep neural network, providing new insights into LOAD's
pathogenic mechanisms.

9
 CHAPTER 3

SYSTEM ANALYSIS

10
 CHAPTER 3

SYSTEM ANALYSIS

3.1 EXISTING SYSTEM

The existing system [1] uses psychophysiological and cognitive tests to

detect dementia stages, combining structural data for improved diagnostic
accuracy. It investigates brain structure and function, hypothesizing specific
features differentiate between disease-related cognitive decline and normal
neurocognitive aging. A study involving 287 cognitively normal cases, 646 mild
cognitive impairment cases, and 369 Alzheimer's disease cases developed a
convolutional neural network-based regression model for predicting cognitive
status.

Fig. 3.1 System Architecture for proposed system

Existing systems for Alzheimer's detection [2] and prediction use machine
learning and deep learning techniques to analyse complex patterns and biomarkers
associated with the disease. These systems use data from neuroimaging, genetic
markers, and clinical assessments to predict the risk of Alzheimer's onset or
11
 progression. These systems aim to enhance early detection, prognosis, and
personalized treatment strategies. However, manual diagnosis is error-prone and
time-consuming. This research proposes a deep learning-based solution using
DenseNet-169 and ResNet-50 CNN architectures for the diagnosis and
classification of Alzheimer's disease. The model categorizes Alzheimer's into
Non-Dementia, Very Mild Dementia, Mild Dementia, and Moderate Dementia
stages. DenseNet-169 outperformed in training and testing phases, demonstrating
potential for real-time analysis and classification of Alzheimer's disease.

3.2 PROPOSED SYSTEM

 In the proposed system, we are making use of CNNs(Convolutional Neural
Networks) Algorithm which are a class of neural network that allow greater
extraction of features from the MRI scans.
 CNNs take the data, train the model, and then classify the features automatically
for healthier classification.
 The model consists of another model for developing an Alzheimer's prediction
system utilizing machine learning techniques trained on the ADNI dataset to
forecast cognitive decline based on neuroimaging and clinical data.
 Additionally, the approach gains even more advantages from the usage of
chatbots. The implementation of the chatbot makes use of Langchain and Llama2
Algorithm. It enhances its efficacy.

The proposed system is divided into the following modules;

 Alzheimer Detection: This module is trained with the dataset that contains
various types of MRI scans of the brain, as well as clinical data such as patient
history, genetic information, and cognitive assessments. With the help of the
model that is trained with CNN can help to classify whether the given MRI scan
of the patient is affected with Alzheimer or not.
 Medical Chatbot: Developing a medical chatbot for Alzheimer's detection
12
 involves collecting and preprocessing medical questions, training a model to
understand user queries, building a knowledge base, and integrating a question-
answering system. The chatbot provides accurate responses to user questions by
retrieving relevant information from the knowledge base. Continuous testing,
evaluation, and maintenance ensure its accuracy and compliance with regulations.
 Alzheimer Prediction: Relevant data is collected, including medical records,
genetic information, and cognitive assessments. Feature engineering extracts
relevant features, followed by training a predictive model deep learning to
forecast Alzheimer's disease likelihood. Evaluation with appropriate metrics
gauges model performance.
 Latest News Update: The Alzheimer's healthcare news update module gathers
and presents the latest research findings and developments in Alzheimer's disease
from reputable sources, providing healthcare professionals and researchers with
timely insights and advancements in the field.

Therefore, the important phases in creating the model include-

 Data collection
 Data pre-processing
 Data exploration and visualization
 Feature scaling
 Splitting the dataset
 Training and testing data
 Model building
 Model deployment

3.2.1. Data collection

The first and foremost step in the process of our model is data collection.
Data collection is the process of gathering the data from various sources. For our
model, we collect the Alzheimer datasets of two types, that is, ADNI dataset and
13
MRI scans dataset. The ADNI dataset is used for the predicting Alzheimer, which
contains the data such as gender, age, demographics, genetic information and
cognitive assessment scores, etc. The other dataset contains the MRI scans of
people at various stages of Alzheimer. These two datasets are used for both
Alzheimer detection and prediction.

3.2.2. Data pre-processing

Data pre-processing is the next stage that follows the data collection
procedure. The information gathered in the earlier phase comes from various
sources. It may not be in a format that is appropriate for us to work on. A variety
of errors—not necessarily errors, such as missing numbers, outliers, redundant
data, etc.—may be included. Therefore, handling each of these irregularities that
are present in the dataset is required at this point. As a result, the unnecessary
columns that essentially reflect the years must be deleted. Moreover, defining the
columns for features (x) and goal (y). where y is the output variable, also known
as the dependent variable, and X stands for the input variables, also known as the
independent variables.

3.2.3. Data exploration and visualization

Data exploration and visualization is a very important step in a model to

understand the trends of the data. It basically helps us to make data-driven
decisions on our own. Visualization is a key factor which helps us to understand
the data in a better way and ultimately increased the productivity and helps us
achieve better results. The target variable (y) is the label indicating the clinical
diagnosis or cognitive status of each subject, such as Alzheimer's disease, mild
cognitive impairment, or healthy control. The model learns to associate patterns
in the input features with the corresponding target labels during the training

14
process, enabling it to make predictions on new, unseen data.

The below graphs depcts the relationship between the given parameters-

Fig.3.2.3a Relationship between Age and Diagnosis (DX.b1)

Fig.3.2.3b Relationship between APOE Genotype and Diagnosis (DX.bl)

3.2.4. Feature scaling

Standardizing the range of independent variables in a dataset is known as

feature scaling. The Euclidean Distance metric is frequently used by machine
learning models to determine the separation between two points in a dataset. For
instance, if a column x consistently has larger values than a different column y,
then (x2-x1)2 will dominate (y2-y1)2, which implies (y2-y1)2 will be regarded as

15
 nonexistent, which is inappropriate. It is for this reason that standardization is
required. We will scale the standardizing formula to each of the 20 selected
column's observations.

3.2.5. Training and testing data

The dataset must be split into training and testing datasets as the next stage
in the data modeling process. The data can be divided in a variety of ways, for
example, by applying ratios like 80:20, 75:25, 70:30, etc. We selected a 70:30
ratio to divide the data for our model, meaning that 70% of the data will be used
for training and 30% for evaluating the model's performance.

3.2.6. Model Building

Developing the model is one of our system's most crucial phases. We use a
variety of supervised learning methods to model the data after dividing it into
training and testing sets. We attempted to use three different techniques to model
the data. They are-
 Logistic Regression
 Random Forest Classifier
 MLP Classifier
 Linear Discriminant Analysis
 Gradient Boosting Classifier
 Convolutional Neural Network

16
 Sequential model
The following sections will provide a description of the aforementioned
algorithms and their workflow. The models from all of the previously mentioned
models that have been implemented are assessed, and the model with the highest
accuracy is chosen to proceed with the deployment process.

3.3 FEASIBILITY STUDY

3.3.1. Social Feasibility

Alzheimer's disease presents global challenges and requires innovative

solutions to enhance diagnostic accuracy and address ethical concerns. Deep
learning techniques in MRI analysis offer early detection, but concerns about
algorithmic bias, patient consent, and data protection persist. Integrating
generative chatbots in Alzheimer's diagnosis platforms enhances inclusivity and
informed decision-making. Addressing ethical, privacy, and accessibility issues is
crucial for maximizing societal benefits. Evaluating the economic feasibility of
implementing comprehensive care ecosystems requires careful consideration of
various factors.

3.3.2. Economic Feasibility

The Alzheimer's Care Ecosystem, despite its initial investment, offers

significant long-term benefits such as improved patient outcomes, reduced
healthcare expenses, and enhanced caregiver support. The ecosystem's early
detection and integration of advanced technologies like CNN-based detection and
predictive analysis improve healthcare delivery efficiency. With growing market
demand for innovative Alzheimer's solutions, the ecosystem has the potential to
capture a significant market share. Revenue opportunities include subscription-
17
based access, technology licensing, partnerships, and premium services. Overall,
the Ecosystem is a promising investment in Alzheimer's care.

3.4 SOFTWARE REQUIREMENT

The software required for our model are-

➢ Operating System: Windows 7/8/10
➢ Language: Python
➢ IDE: VS Code, Google Colab, Jupyter Notebook

3.5 HARDWARE REQUIREMENT

The hardware required for our model are-

➢ Processor: i3 or more
➢ RAM: 4GB or more

18
CHAPTER 4

DATASETS

19
 CHAPTER 4

DATASETS

Our study aimed to investigate the relationship between Alzheimer's disease

progression and various demographic and clinical factors utilizing the
Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset spanning from 2004
to 2021. We utilized three primary datasets: the clinical dataset, the MRI imaging
dataset, and the genetic dataset. The clinical dataset contains approximately
10,000 rows and 100 columns, encompassing variables such as subject ID, age,
gender, cognitive scores (e.g., MMSE), diagnosis, and medication history. The
MRI imaging dataset includes structural and functional MRI scans from over
2,000 participants, providing information on brain volume, cortical thickness, and
connectivity metrics. The genetic dataset consists of genetic markers associated
with Alzheimer's disease risk and progression. These datasets were obtained from
the ADNI database, a longitudinal multicenter study designed to track the
progression of Alzheimer's disease and related disorders. The data were initially
provided in CSV format and subsequently transformed into a Pandas dataframe
for analysis using Python.

20
 Fig 4.1. Dataframe created from the ADNI dataset

Another dataset utilized in our study is the MRI scan dataset obtained from the
Alzheimer's Disease Neuroimaging Initiative (ADNI) database. This dataset
consists of MRI scans collected from participants diagnosed with various
cognitive states, including cognitively normal (CN), Alzheimer's disease (AD),
mild cognitive impairment (MCI), early mild cognitive impairment (EMCI), and
late mild cognitive impairment (LMCI). The source of this dataset is the ADNI
database, a longitudinal multicenter study aimed at understanding the progression
of Alzheimer's disease. The MRI scan dataset contains approximately 10,000
scans with detailed imaging features extracted, including measures of brain
volume, cortical thickness, and white matter integrity. Each scan is associated
with demographic information such as age, gender, and clinical assessments. The
dataset was initially provided in DICOM format and later transformed into a
structured format suitable for analysis. This dataset serves as a valuable resource
for studying the neuroanatomical changes associated with Alzheimer's disease
progression across different cognitive states.

21
Fig 4.2. Images from the MRI scans dataset

22
 CHAPTER 5

SYSTEM DESIGN

23
 CHAPTER 5

SYSTEM DESIGN

5.1. ER DIAGRAM

Figure 5.1 depicts the relationship between the user, administrator, and the
Alzheimer's prediction and detection system. Users input personal data, while the
system utilizes pre-trained models to predict Alzheimer's risk. The administrator
monitors the process and accesses results, including risk levels and potential
indicators, facilitating early intervention strategies.

Fig 5.1 System ER diagram

24
5.2 CLASS DIAGRAM

The following diagram depicts the class diagram of the Alzheimer diagnostic system-

Fig 5.2 Class diagram

Fig 5.2 shows the functions of each class used in our system. The class names here
are the home_page, prediction_page, alzheimer_detection, chat_bot, and news_page.
Each of the class has its own attributes. The prediction_page predicts the Alzheimer
use several characteristics. The alzheimer_detection class checks whether the given
MRI scan image of the patient is healthy or not. The generative chatbot system and
the real-time news update class are the added advantage to the system.

25
 CHAPTER 6

SYSTEM ARCHITECTURE

26
 CHAPTER 6

SYSTEM ARCHITECTURE

The system architecture diagram of our model is given below-

Fig 6.1 System Architecture

Fig 6.1 describes the working of system architecture for Alzheimer Prediction

27
and Detection System. The Alzheimer's prediction and detection system
architecture involve a sequential process starting with data collection, followed
by pre-processing to clean and standardize the data. Exploration and
visualization help understand data characteristics before feature scaling to
ensure uniformity. The dataset is then split into training and testing subsets for
model training and evaluation. Using machine learning algorithms, the model is
built to predict and detect Alzheimer's disease based on the processed data.
Finally, the trained model is deployed for practical use in clinical or research
settings, aiding in early diagnosis and intervention.

6.1 MODULE DESIGN SPECIFICATION

A). Alzheimer Detection: Leveraging advanced CNN deep learning

technology, the proposed Alzheimer detection model utilizes MRI scan images
to accurately diagnose whether an individual has Alzheimer's disease or not.
The system is designed to process and analyze MRI scans, extracting intricate
patterns and features indicative of Alzheimer's pathology. Through extensive
training on a diverse dataset comprising both healthy and Alzheimer's affected
individuals, the CNN model learns to discern subtle differences in brain
structures associated with the disease. The proposed system employs
sophisticated image preprocessing techniques to enhance the quality and
consistency of input data, ensuring robust performance across varied scan
qualities and conditions. Furthermore, the model integrates contemporary deep
learning algorithms to effectively capture complex spatial relationships within
the brain images, enabling it to achieve high accuracy in Alzheimer's detection.
This system holds promise for early diagnosis and intervention, potentially
improving patient outcomes and facilitating more targeted therapeutic
interventions for individuals at risk of Alzheimer's disease.

28
B). Generative Chatbot Support: The Generative Alzheimer's Chatbot
Support System is designed to provide personalized assistance and support for
individuals affected by Alzheimer's disease and their caregivers. Harnessing
generative chatbot technology, the system interacts with users in natural
language, offering empathetic and informative responses to queries related to
Alzheimer's symptoms, treatments, caregiving strategies, and emotional
support. The chatbot employs a deep learning-based generative model trained
on a diverse corpus of Alzheimer's-related information, ensuring accurate and
contextually relevant responses. Additionally, the system integrates sentiment
analysis capabilities to gauge the emotional state of users and tailor responses
accordingly, offering empathetic and supportive interactions. Through
continuous learning and refinement, the chatbot evolves to better understand
and address the unique needs and challenges faced by individuals living with
Alzheimer's disease and their caregivers. This system aims to provide accessible
and reliable support, helping to alleviate stress, improve caregiver well-being,
and enhance the overall quality of life for those affected by Alzheimer's disease.

C) Predictive Analysis: The Predictive Analysis System for Alzheimer's

Disease integrates a wide array of features, including personal information such
as age, gender, and years of education, along with demographics, genetic
information, and cognitive assessments, to accurately predict the risk of
Alzheimer's disease development in individuals. The system employs advanced
machine learning algorithms to analyze and model complex relationships
among these diverse features, identifying patterns and risk factors associated
with the onset of Alzheimer's disease. Through comprehensive data processing
and feature engineering techniques, the system extracts meaningful insights
from heterogeneous datasets, enabling it to generate reliable predictions.
Additionally, the system utilizes probabilistic modeling approaches to estimate

29
the likelihood of Alzheimer's disease occurrence over time, taking into account
longitudinal data and temporal dynamics. By capitalizing on a combination of
demographic, genetic, cognitive, and personal information, the Predictive
Analysis System offers valuable insights into individualized risk assessment for
Alzheimer's disease, empowering healthcare professionals and individuals to
proactively manage and mitigate the impact of the disease through personalized
interventions and lifestyle modifications.

D) AlzNewsFeed: Alzheimer's News Update System is a comprehensive

platform designed to deliver up-to-date news and information related to mental
health and Alzheimer's diseases. Leveraging advanced information retrieval and
natural language processing techniques, the system continuously monitors a
wide range of reputable news sources, medical journals, and research databases
for relevant articles, studies, and updates pertaining to mental health conditions
and Alzheimer's disease. These sources cover a spectrum of topics, including
new treatments, breakthrough research findings, caregiving strategies, and
community resources. The system employs sentiment analysis algorithms to
assess the tone and sentiment of news articles, helping users identify positive
developments and key insights amidst a vast array of information. Additionally,
the platform categorizes news updates into easily navigable sections, allowing
users to explore specific topics of interest such as Alzheimer's research, mental
health awareness initiatives, or caregiver support resources. Through real-time
notifications and personalized recommendations, the system ensures that
stakeholders stay informed about the latest advancements and trends in mental
health and Alzheimer's care, empowering them to make informed decisions and
access relevant support services when needed.

6.2 ALGORITHM

30
The main purpose of this system is to detect and predict the likelihood of
Alzheimer's disease in individuals and offer performance analysis. We have
employed several supervised learning algorithms tailored for classification
problems. These models are thoroughly evaluated, and the model that
demonstrates superior performance is chosen for deployment.

6.2.1 Logistic Regression

Logistic regression is a statistical technique used for binary classification tasks,
where the outcome variable has two possible outcomes. It models the
probability of the binary outcome using predictor variables by fitting a logistic
curve to the data. However, it assumes a linear relationship between predictors
and the log-odds of the outcome and is limited to binary classification tasks.
The formula for the Logistic Regression is as follows:
1
P(Y =1∨X )= −(β 0 +β 1 X 1+ β 2 X 2+...+ βnXn)
1+e
Where:
 P(Y=1∣X) is the probability of the outcome being 1 given the
predictor variables X.
 e is the base of the natural logarithm.
 β0, β1, β2..., βn are the coefficients of the logistic regression
model.
 X1, X2..., Xn are the predictor variables.
The graph for logistic regression is as follows-

31
 Fig. 6.2.1: Logistic Regression

Steps in Linear Regression:

1. Load and preprocess the data.
2. Splitting the data.
3. Define the logistic regression model and its parameters.
4. Train the model.
5. Evaluate the model.
6. Tweak the model's parameters.
7. Make predictions on new, unseen data.

6.2.2 Ridge Classifier

Ridge Classifier is a linear classification algorithm that extends the concept
of linear regression to classification problems. Similar to linear regression,
it assumes a linear relationship between the input features and the target
variable, but it's applied to classification tasks rather than regression tasks.

Steps in Ridge Classifier:

1. Data Loading and Preprocessing.
2. Data Splitting.
3. Ridge Classifier Model Definition.

32
4. Model training and model evaluation.
5. Parameter tuning.
6. Prediction on new data.

CHAPTER 7
EVALUATION METRICS

33
 CHAPTER 7
EVALUATION METRICS

The performance of our model was assessed using a number of metrics for evaluation.
This section discusses common evaluation metrics.

7.1 Confusion Matrix

The metrices contains four values namely True Positive, True Negative, False Positive and
False Negative where-
True Positive (TP) =Observation is positive, and is predicted to be positive. False
Negative (FN) = Observation is positive, but is predicted negative.
True Negative (TN) = Observation is negative, and is predicted to be negative. False
Positive (FP) =Observation is negative, but is predicted positive.

The confusion matrix for the Logistic Regression model can be given as-

Fig. 7.1a: Confusion matrix for Logistic Regression

The confusion matrix for the Ridge Classifier model can be given as-

Fig 7.1b: Confusion matrix for Ridge Classifier

34
7.2 Accuracy
The accuracy of a model can be calculation by the given formula:

TP+TN
Accuracy=
TP+TN + FP+ FN

7.3 Recall
Recall is simply defined as the total number of correctly classified positive observations
divided by the total number of positive classifications.
Recall can be calculated as:

TP
Recall=
TP+ FN

7.4 Precision
Precision can be defined as the total number of correctly classified positive observations
divided by the total number of predicted positive observations.
Precision can be calculated as

TP
Precision=
TP+ FP
7.5 F1 Score

The F1 score is the harmonic mean of precision and recall. It provides a balanced
measure of a model's performance, especially when dealing with imbalanced datasets
where one class is much more frequent than the other.
F1 Score can be calculated using the following formula

35
 Precision X Recall
F 1=2.
Precision+ Recall

CHAPTER 8

SYSTEM IMPLEMENTATION

36
 CHAPTER 8

SYSTEM IMPLEMENTATION

8.1 SAMPLE CODE

Module: Alzheimer Detection

# import opencv
import cv2

# store list
images = []
lables = []

## AD Patient Brain Scan with PreProcessing of the Image ##

for i in range(1, 172):
img = cv2.imread(f"C:\\Alzheimer-Disease-Prediction-master\\Alzheimers-ADNI\\
train\\Final AD JPEG\\AD ({i}).jpg")
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
gray = gray/255
gray = cv2.resize(gray,(200, 200))
images.append(gray)
lables.append(0) # AD

# Sample Images
import matplotlib.pyplot as plt
plt.figure(figsize = (15,15))
for i in range(20):
plt.subplot(4, 5, i + 1)

37
 plt.imshow(images[10 + i*3])
plt.show()
# Shape of the Images
print(f"Shape of each image is = {images[1000].shape}")

# Convert The List into Array Format #

import numpy as np
train_feature = np.array(images)
lables = np.array(lables)

## Display Array Shape ##

print(f"image dataset shape = {train_feature.shape}")
print(f"lable dataset shape = {lables.shape}")

# # Deep Learning CNN Model Architecture

# MODEL ARCHITECTURE
import keras
import tensorflow
from tensorflow.keras.utils import to_categorical
from keras.models import Sequential
from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D, Dropout

# ONE NOT ENCODING

train_target = to_categorical(lables)

# MODEL LAYERS
model=Sequential()
# Convolutional Layers
model.add(Conv2D(25, kernel_size = (3,3), strides = (1,1), padding = 'same', activation
= 'relu', input_shape = (200, 200, 1)))
38
model.add(Conv2D(75, kernel_size = (3,3), strides = (1,1), padding = 'same', activation
= 'relu'))
model.add(MaxPooling2D(pool_size = (2,2)))
model.add(Flatten())
model.add(Dense(500,activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(250,activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(100,activation='relu'))
model.add(Dense(5,activation='softmax'))
model.summary()

# MODEL COMPILE

# Function to predict Alzheimer's

def predict_alzheimer(image):
processed_img = preprocess_image(image)
prediction = model.predict(processed_img)
class_label = np.argmax(prediction)
if class_label == 0:
return "AD (Alzheimer's Disease)"
elif class_label == 1:
return "CN (Cognitively Normal)"
elif class_label == 2:
return "EMCI (Early Mild Cognitive Impairment)"
elif class_label == 3:
return "LMCI (Late Mild Cognitive Impairment)"
elif class_label == 4:
return "MCI (Mild Cognitive Impairment)"

39
# Main function for Streamlit app
def main():
st.title("Alzheimer's Disease Prediction")
st.write("Upload an image to predict the result")

# File uploader
uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg"])

if uploaded_file is not None:

# Display uploaded image
image = np.array(bytearray(uploaded_file.read()), dtype=np.uint8)
st.image(image, caption='Uploaded Image', use_column_width=True)

# Predict result
result = predict_alzheimer(image)
st.write("Predicted class:", result)

if __name__ == "__main__":
main()

40
 CHAPTER 9

SYSTEM TESTING

41
 CHAPTER 9

SYSTEM TESTING

In this chapter, the results are checked for inputs provided by the user.

9.1 ALZHEIMER DETECTION SYSTEM

[1] Input
Upload an image (CN image)

[1] Output
Predicted Class: Cognitively Normal

9.2 ALZHEIMER PREDICTION SYSTEM

[2] Input
Age: 65
Gender: Male
Years of Education: 20
Ethnicity: Hisp/Latino
Race Category: White
APOE Allele Type: APOE4_0
APOE4 Genotype: 2,2
Imputed Genotype: True
MSME Score: 25

[2] Output
Detected Class: Late Mild Cognitive Impairment (LMCI)

42
 CHAPTER 10

RESULTS AND DISCUSSIONS

43
 CHAPTER 10

RESULTS AND DISCUSSION

10.1 RESULTS AND DISCUSSIONS

In this study, logistic regression was employed to predict the presence or absence
of Alzheimer's disease based on a comprehensive set of features including patient
information, demographics, ethnicity, and genetic information.

Precision Recall F1-score Support

0 0.81 0.85 0.83 20
1 0.91 0.88 0.89 33

Accuracy 0.87 53
Macro avg 0.86 0.86 0.86 53
Weighted avg 0.87 0.87 0.87 53

Table 10.1: Classification report for Logistic Regression

The above table represents the accuracy report for the Logistic Regression model.
The model achieved an accuracy of 87%, indicating a substantial level of
predictive capability. This suggests that the variables included in the model are
informative in distinguishing between individuals with and without Alzheimer's
disease.
Our approach involved the utilization of a Convolutional Neural Network (CNN)
model, constructed using the Keras deep learning API, to analyze and classify
brain images associated with Alzheimer's disease progression. We preprocessed
the images using OpenCV, a computer vision library, to enhance their quality and
extract relevant features. Subsequently, we employed scikit-learn, a machine
learning library, to split the dataset into training and testing sets, ensuring robust
evaluation of the model's performance.

44
Upon rigorous evaluation, our model demonstrated exceptional accuracy in
classifying different stages of cognitive impairment, achieving an impressive
accuracy score of 98%. This signifies the model's ability to accurately discern
subtle variations in brain imaging patterns indicative of Alzheimer's disease
progression.

Fig 10.1: Training accuracy and validation accuracy graph of CNN.

The high accuracy achieved in classifying Alzheimer's disease progression

underscores the efficacy of our machine learning approach. By leveraging
advanced computational techniques, we have developed a robust tool for early
detection and staging of Alzheimer's disease, facilitating timely intervention and
personalized care for affected individuals. The ability to accurately identify
individuals at different stages of cognitive impairment holds significant
implications for clinical practice, enabling healthcare professionals to initiate
appropriate treatment strategies and support services tailored to the specific needs
of patients.

GenAI models are evaluated using four datasets: ARC, HellaSwag, MMLU, and
TruthfulQA. Each dataset has a 25-shot evaluation setup, with rankings
determined by calculating average performance across different levels. This
comprehensive approach provides a holistic understanding of models' capabilities.
45
The following table is the accuracy report of the chatbot-

Model Avg ARC HS MMLU TQA

Shining Valiant 1.2 74.17 72.95 87.88 70.97 64.88
Llama 2 67.35 67.32 87.33 69.83 44.92
Llama 2 Chat 66.80 64.59 85.88 63.91 52.80

Table 10.2: Accuracy table for Generative chatbot support

46
CHAPTER 11

CONCLUSION

47
 CHAPTER 11

CONCLUSION

11.1 CONCLUSION

The Enhanced Alzheimer’s diagnosis system is an advanced solution that

transformed the Alzheimer detection system at early stage and manages it. It
contains modern technologies such as Convolutional Neural Networks (CNN),
Generative Chatbot Support, Logistic Regression and Real-time news update
mechanism. The system used CNN model to detect the presence of AD or varying
stages of cognitive impairment of individuals with the help of MRI scan results.
This benefits the healthcare professional to quickly identify any minute changes
in the brain structure. The prediction system which is developed using the
machine learning model gets the data from the user and predicts the status of the
patients in terms of AD. The Generative Chatbot Support System offers a
personalized assistance for the patients, caregivers, family members, etc. The
Generative Chatbot Support System, powered by LLAMA2, LLM, and
LangChain, offers personalized assistance for patients, caregivers, family
members, and others. LLAMA2, LLM, and LangChain are such vital parts of the
system's framework that are considered to be the drivers for the provision of the
right information and resources that will be tailored towards the needs and
preferences of each individual. Additionally, the real-time news update module
helps the user to find the advances made in the contemporary world. The
integration of this system helps both the patients and the doctors to plan the
treatment at the early state, ultimately enhancing patient outcomes and quality of
life.

48
11.2 FUTURE WORK

Future enhancements for the Comprehensive Alzheimer's Care Ecosystem could

involve incorporating additional machine learning techniques to improve the
accuracy of CNN-based detection, expanding the capabilities of the medical
chatbot support system to offer more personalized assistance, integrating more
comprehensive real-time news updates on Alzheimer's research and treatment
options, and refining predictive analysis algorithms to enhance early intervention
strategies. Additionally, exploring avenues to incorporate sensor data and
wearable technology for continuous monitoring of patients' cognitive and
physical health could further enrich the ecosystem's capabilities in providing
comprehensive care and support for individuals affected by Alzheimer's disease.

49
CHAPTER 12

APPENDICES

50
 CHAPTER 12

APPENDICES

A.1 Sample Screens

Screen Shot of Home Page:

Figure A.1: Screen Shot of Home Page

Screenshot of Alzheimer Detection page:

Figure A.1: Screen Shot of Alzheimer Detection System

51
Screenshot of Alzheimer Prediction page:

Figure A.1: Screen Shot of Alzheimer Prediction System

Screenshot of chatbot page:

Figure A.1: Screen Shot of Generative Chatbot

52
Screen Shot of AlzNewsFeed:

Figure A.1: Screen Shot of Real-time news update

53
CHAPTER 13

REFERENCES

54
 CHAPTER 13

REFERENCES

1. Tetiana Habuza, Nazar Zaki, Elfadil A. Mohamed, Yauhen Statsenko,

“Deviation From Model of Normal Aging in Alzheimer’s Disease: Application
of Deep Learning to Structural MRI Data and Cognitive Tests”, IEEE Access
(Volume: 10), DOI: 10.1109/ACCESS.2022.3174601
2. Al Shehri W. 2022. Alzheimer’s disease diagnosis and classification using
deep learning techniques. PeerJ Computer Science,DOI: 10.7717/peerj-cs.1177
3. Cristina L. Saratxaga, Iratxe Moya, Artzai Picón, Marina Acosta, Aitor
Moreno-Fernandez-de-Leceta, Estibaliz Garrote and Arantza Bereciartua-
Perez, “MRI Deep Learning-Based Solution for Alzheimer’s Disease
Prediction”, https://doi.org/10.3390/jpm11090902
4. Nasir Rahim, Shaker El-Sappagh, Sajid Ali, Khan Muhammad, Javier Del Ser,
Tamer Abuhmed, “Prediction of Alzheimer's progression based on multimodal
Deep-Learning-based fusion and visual Explainability of time-series data”,
https://doi.org/10.1016/j.inffus.2022.11.028
5. Yan Zhao, Qianrui Guo, Yukun Zhang, Jia Zheng, Yang Yang, Xuemei Du,
Hongbo Feng and Shuo Zhang, “Application of Deep Learning for Prediction
of Alzheimer’s Disease in PET/MR Imaging”, Bioengineering 2023, 10(10),
1120; https://doi.org/10.3390/bioengineering10101120
6. Naveen Sundar Gnanadesigan, Narmadha Dhanasegar, Manjula Devi
Ramasamy, Suresh Muthusamy, Om Prava Mishra, Ganesh Kumar
Pugalendhi, Suma Christal Mary Sundararajan & Ashokkumar Ravindaran,
“An integrated network topology and deep learning model for prediction of
Alzheimer disease candidate genes”, Application of soft computing, Published:
15 May 2023, Volume 27, pages 14189–14203, (2023)
7. Zhen Zhao, Joon Huang Chuah, Khin Wee Lai, Chee-Onn Chow, Munkhjargal
Gochoo, Samiappan Dhanalakshmi, Na Wang, Wei Bao, Xiang Wu,
55
 “Conventional machine learning and deep learning in Alzheimer's disease
diagnosis using neuroimaging: A review”, Front. Comput. Neurosci,
https://doi.org/10.3389/fncom.2023.1038636
8. Pradnya Borkar, Vishal Ashok Wankhede, Deepak T. Mane, Suresh Limkar, J.
V. N. Ramesh & Samir N. Ajani, “Deep learning and image processing-based
early detection of Alzheimer disease in cognitively normal individuals”, Focus,
Published: 09 June 2023, https://doi.org/10.1007/s00500-023-08615-w
9. Daichi Shigemizu, Shintaro Akiyama, Mutsumi Suganuma, Motoki Furutani,
Akiko Yamakawa, Yukiko Nakano, Kouichi Ozaki & Shumpei Niida,
“Classification and deep-learning–based prediction of Alzheimer disease
subtypes by using genomic data”, Translational Psychiatry volume 13, Article
number: 232 (2023)
10. Shangran Qiu, Prajakta S Joshi, Matthew I Miller, Chonghua Xue, Xiao
Zhou, Cody Karjadi, Gary H Chang, Anant S Joshi, Brigid Dwyer, Shuhan
Zhu, “Development and validation of an interpretable deep learning framework
for Alzheimer’s disease classification”, Brain, Volume 143, Issue 6, June 2020,
Pages 1920–1933, https://doi.org/10.1093/brain/awaa137
11. E. Mggdadi, A. Al-Aiad, M. S. Al-Ayyad and A. Darabseh, "Prediction
Alzheimer's disease from MRI images using deep learning," 2021 12th
International Conference on Information and Communication Systems
(ICICS), Valencia, Spain, 2021, pp. 120-125, doi:
10.1109/ICICS52457.2021.9464543

56