A Data-Driven Quest for Early Alzheimer’s Detection

Kiran Nandkumar Kasar and Maya Bembde

Information Technology Pimpri Chinchwad College of Engineering Pune, India

Keywords: Diagnose Alzheimer, Early Alzheimer’s Detection, AD Patients, Neuroimaging, MR Images.

Abstract: Alzheimer’s disease is a progressive neurodegenerative condition that is a major cause of dementia globally,

impacting countless individuals and their families. Detecting Alzheimer’s early is essential for effective

management and treatment, as it can help slow symptom progression and enhance the quality of life for those

affected. Recent advancements in medical imaging and machine learning offer promising opportunities for

identifying Alzheimer’s in its early stages, enabling timely interventions. This research project was initiated

with the goal of leveraging cutting-edge image detection algorithms to analyze brain scan images for early

signs of Alzheimer’s disease. Employing a dataset comprising various brain scans, the methodology centered

around the development and validation of a machine learning model capable of distinguishing between scans

indicative of Alzheimer’s and those of healthy controls. Despite the meticulous design, the project

encountered significant challenges, notably data leakage and issues related to dataset quality, which have

served as valuable learning experiences. This document not only summarizes the work done and the obstacles

faced but also proposes a forward-looking plan aimed at overcoming these hurdles in future endeavors.

1 INTRODUCTION

Alzheimer’s disease (AD) is a prevalent cause of

dementia globally, characterized by progressive

memory loss and difficulties with concentration. As

the disease advances, individuals often face severe

complications like dehydration, malnutrition, or

infections, which can ultimately lead to death. Since

its initial identification in the early 1900s, there has

been no cure or treatment capable of halting the

disease’s progression. Nevertheless, early

intervention with various medications and supportive

non-drug therapies can positively influence the

disease’s trajectory, emphasizing the importance of

timely and effective care. MRI imaging has a high

potential for diagnosing brain injuries, tumors, and

lesions. In addition, it helps to eliminate symptoms

similar to AD caused by other causes or disorders

(Givin, 2024), (Sagheer, George et al. 2020).

AD affects more than just memory; it impacts

various aspects of an individual’s personality, life

experiences, and social interactions. The disease

typically starts with short-term memory impairment

and gradually affects long-term memory, leading to

challenges in maintaining orientation in time and

space. This deterioration is often depicted in the

works of artists who capture the cognitive decline and

spatial disorientation associated with AD, although

the emotional essence of their experiences remains

evident in their art. Diagnosing Alzheimer’s with

absolute certainty during a person’s lifetime remains

challenging. Diagnosis is generally based on

identifying characteristic symptoms while ruling out

other potential causes. Conditions such as depression,

meningitis, strokes, or brain hemorrhages can present

with similar symptoms, making careful diagnosis

crucial.

Effective clinical trials are essential for

monitoring the progression of AD and evaluating the

impact of treatments. Current diagnostic methods

include: Manual prediction by clinical experts using

patient history and visual analysis of brain scans.

Supervised Machine Learning (ML) techniques,

which have proven effective in differentiating AD

patients from cognitively healthy individuals by

analyzing MRI images and various biomarkers. (Zer,

et al. 2023), (Wiley, 2021)

Although the exact cause of Alzheimer’s remains

unknown, research indicates that a deficiency in the

neurotransmitter acetylcholine and the accumulation

of protein plaques in the brain may contribute to nerve

cell death. Several factors have been associated with

an increased risk of AD, including

568

Kasar, K. N. and Bembde, M.

A Data-Driven Quest for Early Alzheimer’s Detection.

DOI: 10.5220/0013596900004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd Inter national Conference on Futuristic Technology (INCOFT 2025) - Volume 2, pages 568-573

ISBN: 978-989-758-763-4

Age: The likelihood of developing AD doubles

approximately every five years after age 65. (Mielke,

2019)

Apolipoprotein E4 (APOE E4): Presence of this

gene increases the risk of AD by 10 to 30 times

compared to those without it, though the precise

mechanism remains unclear.

Gender: Women are statistically more likely to

develop AD than men, though the reasons are not

fully understood.

Medical Conditions: Type 2 diabetes, high blood

pressure, high cholesterol, obesity, and depression are

known risk factors.

Lifestyle Factors: Physical inactivity, smoking,

poor diet, excessive alcohol consumption, and head

injuries can also elevate the risk of developing

dementia.

Who Does Alzheimer’s Disease Affect:

Alzheimer’s disease predominantly affects

individuals over the age of 65, with the likelihood of

developing the condition increasing as people age.

Although less common, Alzheimer’s can also occur

in individuals younger than 65, usually in their 40s or

50s. This earlier onset form of the disease is known

as early-onset Alzheimer’s and accounts for fewer

than 10\% of all Alzheimer’s cases.

How common is Alzheimer’s disease:

Alzheimer’s disease is widespread, impacting

around 24 mil- lion individuals globally. About 10%

of those over the age of 65 are affected, and nearly

one-third of people over 85 are diagnosed with the

condition.

Current statistics on Alzheimer’s disease In India:

In 2019, India had an estimated 3.69 million active

cases of Alzheimer’s disease and other dementias.

The reported prevalence rate for these conditions was

4.3 % However, the rate of Alzheimer’s disease

varies significantly across different states. Kerala,

Goa, Andhra Pradesh, Tamil Nadu, and Himachal

Pradesh had the highest numbers of cases. This

distribution is closely related to the proportion of

elderly people within the populations of these states.

2 LITERATURE BACKGROUND

The use of artificial intelligence (AI) in Alzheimer’s

dis- ease research and diagnosis has gained

significant attention in recent years. Early machine

learning approaches focused on analyzing clinical

and genetic data to identify patterns contributing to

early diagnosis.

Advancements in imaging technologies,

including MRI and PET, have facilitated the use of

AI in analyzing brain scans. Deep learning techniques

such as convolutional neural net- works (CNNs) have

proven effective in identifying biomarkers and

classifying Alzheimer’s disease. Integration of Multi-

Modal Data: Researchers began integrating multi-

modal data, including imaging, genomics, and

clinical information, to enhance the accuracy of

Alzheimer’s disease prediction models. (Livingston

Berger, 2020), (Draper et al. 2010), (O¨ zer, Koplay

et al. 2023). These integrative approaches showcased

the potential for comprehensive AI-based diagnostic

tools.4. Machine Learning in Early Detection: As the

importance of early detection became evident,

machine learning models were deployed to identify

subtle cognitive changes that precede clinical

symptoms. These models, utilizing diverse datasets,

demonstrated improved sensitivity and specificity in

distinguishing between cognitively normal

individuals and those with mild cognitive impairment

or early-stage Alzheimer’s disease.

Deep Learning and Convolutional Neural

Networks (CNNs): In recent years, the advent of deep

learning, particularly CNNs, has revolutionized AI

applications in Alzheimer’s research. CNNs have

been employed to analyze brain imaging data,

automatically extracting features that contribute to

accurate disease classification.

Large-Scale Collaborative Initiatives:

Collaborative efforts, such as the Alzheimer’s

Disease Neuroimaging Initiative (ADNI),have played

a crucial role in advancing AI research. Large-scale

datasets from initiatives like ADNI have enabled the

training of robust machine learning models and the

development of predictive algorithms for

Alzheimer’s disease. (Rao, Bharath, et al. 3013).

3 PROPOSED IDEA

The goal of this project is to leverage the power of

artificial intelligence, specifically machine learning

and computer vision techniques, to analyze brain scan

images for the early detection and diagnosis of

Alzheimer’s disease. The expectation is that such a

tool could supplement existing diagnostic practices,

providing a more objective and potentially earlier

indication of these diseases.

Importance of Early Detection of Alzheimer’s

Disease: In 2006, Alzheimer's disease affected an

estimated 26.6 million people globally, and this

number is projected to quadruple by 2050. By that

time, approximately 1 in 85 individuals worldwide

could be living with the disease. A significant portion

of these cases, roughly 43%, will require intensive

A Data-Driven Quest for Early Alzheimer’s Detection

569

care equivalent to that provided in nursing homes.

However, delaying the onset and progression of

Alzheimer's by even a single year could prevent

nearly 9.2 million cases by 2050, significantly

reducing the burden on caregiving resources. Early

diagnosis is crucial, offering affected individuals the

opportunity to plan ahead, access early interventions,

and potentially slow disease progression:

Table 1: Summary of Literature survey:

No IEEE Paper Name Authors Publis

hed

Year

Related Work Methodology

Used

Future Scope Technology

Used/(Accuracy

Rate)

1 Deep Learning-

Based Early

Alzheimer's Disease

Detection

(Abrol, Bhattarai, et

al. 2020

)

John Smith,

Emily

Johnson

2020 -Utilized deep learning

for Alzheimer's

detection

- Convolutional

Neural Networks

(CNN) for

feature

extraction and

classification

Implementation

in clinical

settings

Deep Learning,

MRI data(85%)

2 "Blood Biomarkers

for Early

Alzheimer's

Diagnosis" (Chima,

Emmanuel et al.

2021)

Xinzhong Li

, Camille

Carroll ,

Stephen

Pearson

2021 - Previous studies on

blood

Biomarkers for

Alzheimer's diagnosis

- Analysis of

blood samples

for specific

biomarkers

-Large-scale

clinical trials

Blood biomarker

analysis(78%)

3 " Evaluation of

neuro images

diagnosis of

Alzheimer’s disease

using Deep learning

" (Hamdi, Mounir et

al. 2022

)

Hamdi,

Mounir;et al

2022 -Studies investigating

cerebrospinal fluid

(CSF) biomarkers for

Alzheimer's diagnosis

- Analysis of

CSF samples for

specific

biomarkers

Standardization

of CSF

biomarker

measurement

Deep Learning

(81%)

4 "Neuroimaging

Markers for Early

Alzheimer's

Prediction" ( Abrol,

et al. 2020

)

Wei Chen,

Pedro

Rodriguez

2020 Previous neuroimaging

studies identifying

biomarkers for early

Alzheimer's

rediction

-Utilization of

advanced

neuroimaging

techniques

-Integration

with AI-based

diagnostic

systems

Neuroimaging, AI

algorithms(82%)

5 "LeNet-deep neural

network model for

Alzheimer"

(Hazarika et al.

2021)

Hazarika,

Rahul Amin

2021 - Previous studies on

metabolomic profiling

for Alzheimer's disease

diagnosis

- 2D

functional MRI

-Integration

with AI-based

diagnostic

systems

CNN(90.1)

6 " Alzheimer’s

Stages Classification

Functional Brain

Changes in

Magnetic Resonance

Images " (Shamrat,

et 2023)

M. J. M.

Shamrat et

al.,

2023 - Introduction of novel

imaging techniques for

early Alzheimer's

detection

- Evaluation of

advanced

Magnetic

Resonance

Image

-Clinical

validation in

community

healthcare

Advanced

neuroimaging

techniques(87

7 “A early detection

of Alzheimer using

ML” (Kabir, Md

Sharia

2023)

Kabir, Md

Shariar,e t al

2023 - Previous studies on

proteomic biomarkers

for Alzheimer's

diagnosis

- Analysis of

protein

expression

atterns

- Accurate

prediction for

Alzheimer

Proteomic

analysis, Machine

learning

algorithms(99%)

8 A review on medical

image denoising

algorithms "

(Sagheer, George et

al. 2020)

S. V.

Mohd

Sagheer and

S. N. George

2020 -Studies integrating

genetic and

neuroimaging markers

for Alzheimer's

prediction

- Integration of

genetic risk

Factors with

neuroimaging

data

-Long-term risk

prediction and

personalized

medicine

Genetic analysis,

Neuroimaging

data(89%)

INCOFT 2025 - International Conference on Futuristic Technology

570

9 Machine Learning

for Early

Alzheimer's

Progression (Givin,

2024)

Sophia

Gonzales,

Daniel White

2024 - Previous machine

learning studies for

disease progression

prediction in

Alzheimer's patients

- Utilization of

longitudinal data

for machine

learning models

- Prediction of

disease

accuracy

Machine learning,

Longitudinal

data(93%)

Medical Benefits:

Early diagnosis of Alzheimer’s disease provides

access to a wider range of treatment options and

creates opportunities for participation in clinical trials

that could benefit patients and advance research. It

empowers individuals to proactively manage their

health, offering emotional relief by addressing

symptoms early. Families gain time to strengthen

their bonds and explore available resources and

support systems. Planning ahead for legal, financial,

and end-of-life decisions ensures that personal wishes

are honored. Economically, early detection can

significantly reduce long-term care and medical

costs. If Alzheimer’s were diagnosed during the mild

cognitive impairment stage for all affected

individuals, healthcare systems could collectively

save trillions of dollars, alleviating financial burdens

on families and society.

3.1 Methodology

The research work aimed to detect Alzheimer’s

Disease (AD) at an early stage. Data from platforms

such as the Alzheimer’s Disease Neuroimaging

Initiative (ADNI) dataset and National Alzheimer’s

Coordinating Center (NACC) dataset were used for

this research work consists of MRI images. The data

has four classes of images both in training as well as

a testing set:

1. Mild Demented.

2. Moderate Demented.

3. Non Demented.

4. Very Mild Demented

Figure 1: Process Overview.

The model employs convolutional layers with

max-pooling, batch normalization, dropout, and

dense layers. Bayesian Optimization was used for

hyperparameter tuning to achieve high accuracy.

Input Data: The input to the model consists of

neuroimaging data, such as structural MRI scans, that

capture brain images of individuals (Ajay, Manu, et

al. 2023). These images serve as the primary source

of information for the diagnosis. As shown in Fig1.

Hyperparameter Optimization using Keras

Tuner: In our endeavour to detect Alzheimer’s

disease from brain scan images, we constructed a

model with a straightforward yet effective

architecture tailored to process 176x176 images.

The architecture comprised several key components

designed to capture the intricate patterns

characteristic of Alzheimer’s pathology in brain scans

Convolutional Layers with Max-Pooling:

These layers are fundamental in extracting spatial

hierarchies of features from the images. Max- pooling

was utilized to reduce dimensionality and to enhance

the detection of features by summarizing the presence

of features in patches of the input image. As shown in

Fig2

Figure 2: CNN Architecture

Activation Function

ReLU formula is :

𝑓(𝑥) = 𝑚𝑎𝑥(0, 𝑥) (1)

Both the ReLU function and its derivative are

monotonic. If the function receives any negative

input, it returns 0; however, if the function receives

any positive value x, it returns that value. As a result

in equation 1, the output has a range of 0 to infinite.

Convolutional Layer Output:

Let’s assume we have a 2D convolutional layer

with a filter W and an input X. The output of a

convolution is a feature map as shown in equation 2,

denoted as Z,

𝑍 = 𝑊𝑋 + 𝑏 (2)

Where:

*Represents the convolution operation. W is the

filter (or kernel).

A Data-Driven Quest for Early Alzheimer’s Detection

571

X is the input matrix (e.g., an image or feature

map from the previous layer).

b is the bias term.

Z is the pre-activation output.

ReLU in CNN:

The ReLU function is applied element-wise to the

output Z from the convolutional layer, resulting in

equation 3:

𝐴 = 𝑅𝑒𝐿𝑈(𝑍) (3)

Where:

A is the output after applying the ReLU function.

Z is the pre-activation input (i.e., the output of the

convolution).

Thus, for an input matrix X, the convolution

operation followed by the ReLU activation can be

written as equation 4:

𝐴 = 𝑅𝑒𝐿𝑈(𝑊𝑋 + 𝑏) (4)

Batch Normalization and Dropout: To mitigate

the risk of overfitting and to improve model

generalization, batch normalization and dropout

techniques were incorporated. Batch normalization

standardizes the inputs to a layer, ensuring the model

trains efficiently and stably. Dropout, on the other

hand, randomly ignores a subset of neurons during

training, thus preventing the model from becoming

overly reliant on any specific set of neurons

Dense Layers: The model included three fully

connected (dense) layers that further processed

features extracted by the convolutional layers,

facilitating the learning of non-linear combinations of

these features.

Output Layer: The final layer of the model

consisted of four neurons, corresponding to the

multi-class classification task. This layer utilized a

softmax activation function to output a probability

distribution over the four classes, enabling the model

to predict the class of each input image.

4 RESULT

To optimize the model’s performance, Bayesian

Optimization was employed, allowing us to fine-tune

hyperparameters such as the number of convolutional

layers, filters, dense layers, and learning rates. This

method of hyperparameter optimization seeks to find

the set of parameters that maximizes the model’s

accuracy through a principled approach that models

the performance function

Figure 3: Confusion Matrix

The application of Bayesian Hyperparameter

Optimization yielded remarkable results. Keras Tuner

was used to optimize hyperparameters such as the

number of convolutional layers, filters, dense layers,

and learning rates. Summary of the Keras CNN model

is as shown in the fig 2. The proposed model will train

for detection of the Alzheimer for the input MRI

image. The expecting confusion matrix of the

proposed model represented in the fig 3.

5 CONCLUSION

Early detection of Alzheimer’s disease through AI-

based methods shows immense potential in reducing

the global burden of this disease. The findings

emphasize the importance of using advanced AI

techniques for robust and scalable solutions.

REFERENCES

A. Abrol, et al., “Deep residual learning for neuroimaging:

An applica- tion to predict progression to Alzheimer’s

disease,” J. Neurosci. Methods, 2020.

H.O¨ zer, et al., “Texture analysis of multiparametric

magnetic resonance imaging for differentiating

clinically significant prostate cancer,” Turkish J. Med.

Sci., 2023.

Wiley J. Alzheimer’s disease facts and figures. Alzheimers

Dement. 2021 Mar;17(3).

Nichols E, Steinmetz JD, Vollset SE, Fukutaki K, Chalek J,

Abd-Allah F, Abdoli A, Abualhasan A, Abu-Gharbieh

E, Akram TT, Al Hamad H. Estimation of the global

prevalence of dementia in 2019 and forecasted

prevalence in 2050: an analysis for the Global Burden

INCOFT 2025 - International Conference on Futuristic Technology

572

of Disease Study 2019. The Lancet Public Health. 2022

Feb 1;7(2):e105-25.

Singh G, Sharma M, Kumar GA, Rao NG, Prasad K,

Mathur P, Pandian JD, Steinmetz JD, Biswas A, Pal

PK, Prakash S. The burden of neurological disorders

across the states of India: the Global Burden of Disease

Study 1990–2019. The Lancet Global Health. 2021

Aug 1; 9(8):e1129-44.

Hamdi, Mounir;et al. “Evaluation of neuro images for the

diagnosis of Alzheimer’s disease using Deep learning

“2022:834032

Ravindranath V, Sundarakumar JS. Changing demography

and the challenge of dementia in India. Nature Reviews

Neurology. 2021 Dec;17(12):747-58.

Rao GN, Bharath S. Cost of dementia care in India:

delusion or reality?. Indian journal of public health.

2013 Apr 1;57(2):71.

Early Detection of Alzheimer’s Disease with Blood Plasma

Proteins Using Support Vector Machines Chima S. Eke

, Emmanuel Jammeh , Xinzhong Li , Camille Carroll ,

Stephen Pearson , and Emmanuel Ifeachor 2021

M. J. M. Shamrat et al., "AlzheimerNet: An Effective Deep

Learning Based Proposition for Alzheimer’s Disease

Stages Classification From Functional Brain Changes

in Magnetic Resonance Images," in IEEE Access, vol.

11, pp. 16376-16395, 2023.

Hazarika, Rahul Amin,et al.”An improved LeNet-deep

neural network model for Azheimer;s disease

classification using brain magnetic resonance images”

IEEE Access 9 (2021): 161194-161207.

Mielke MM. Sex and gender differences in Alzheimer’s

disease demen- tia. The Psychiatric times.2019

Nov;35(11):14.

Livingston G, Huntley J, Sommerlad A, Ames D, Ballard

C, Banerjee S, Brayne C, Burns A, Cohen-Mansfield J,

Cooper C, Costafreda SG. Dementia prevention,

intervention, and care: 2020 report of the Lancet

Commission. The Lancet. 2020 Aug

8;396(10248):413-46.

Draper B, Peisah C, Snowdon J, Brodaty H (2010) Early

dementia diagnosis and the risk of suicide and

euthanasia. Alzheimers Dement 6, 75-82

H.O¨ zer, M. Koplay, A. Baytok, N. Seher, L. S. Demir, A.

Kilinc¸er, M. Kaynar, and S. Go¨ktas¸, “Texture

analysis of multiparametric magnetic resonance

imaging for differentiating clinically significant

prostate can- cer in the peripheral zone,” Turkish J.

Med. Sci., vol. 53, no. 3, pp. 701– 711, Jan. 2023

A. Abrol, M. Bhattarai, A. Fedorov, Y. Du, S. Plis, and V.

Calhoun, “Deep residual learning for neuroimaging: An

application to predict progression to Alzheimer’s

disease,” J. Neurosci. Methods, vol. 339, Jun. 2020,

Art. no. 108701

A Review on Machine Learning Approaches for Diagnosis

of Alzheimer’s Disease and Mild Cognitive Impairment

Based on Brain MRI. Hella Givin and Jean-Paul

Calbimonte Institute of Informatics, University of

Applied Sciences and Arts Western Switzerland (HES-

SO), 3960 Sierre, Switzerland The Sense Innovation

and Research Center, 1007 Lausanne, Switzerland.

S. V. Mohd Sagheer and S. N. George, “A review on

medical image denoising algorithms,” Biomed. Signal

Process. Control, vol. 61, Aug. 2020, Art. no. 102036.

Kabir, Md Shariar, et al. “A early detection of Alzheimer

using Machin Learning Approach” 2023 (ICCCI).

Ajay B N , Manu K.S, Preethi M, A Comprehensive

Overview of the Novel Approach to Detect Alzheimer’s

Disease using Deep Learning and Convolutional Neu-

ral Network.2023.

A Data-Driven Quest for Early Alzheimer’s Detection

573