Alzheimer Detection System

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 04 | Apr 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1473
Alzheimer Detection System
Hrutvik Rane, Swarali More, Ghanshyam Patel, Maitrey Phatak, Charmi Chaniyara
Hrutvik Rane, Atharva College of Engineering, Maharashtra, India
Swarali More, Atharva College of Engineering, Maharashtra, India
Ghanshyam Patel, Atharva College of Engineering, Maharashtra, India
Maitrey Phatak, Atharva College of Engineering, Maharashtra, India
Charmi Chaniyara, Atharva College of Engineering, Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Neurodegeneration in addition to poor
communication between neuron synapses lead to Dementia
and Alzheimer’s disease. Alzheimer's disease (AD), the most
common form of dementia, damages the brain, resulting in
impaired memory and ability to perform daily tasks due to
damage to the brain. With the help of MRI (Magnetic
Resonance Imaging) scans of brain images, with the help of
artificial intelligence (AI) technology, we can diagnose and
predict the disease and classify AD patients to determine if
they will develop this deadly disease. future. To be. The main
goal of all these actions is to save time and money for
radiologists, doctors and nurses and to develop better
predictive tools, information and diagnostics to assist
patients with this disease. Recently, the usage of deep
learning algorithms has been increasingly helpful in
diagnosis of AD. This is because DL algorithms work on
large datasets. In the paper, we have made use of
convolution neural network to work with early detection
and classifying of the disease. CNNs are popular because of
their excellent performance in machine learning using a
wide range of information.
Key Words: Neurological disorder, Alzheimer’s disease,
Deep learning, MRI, Convolutional neural network, Brain
imaging.
1.INTRODUCTION
The number of dangerous diseases has increased in recent
years due to demographic shifts in developing and
developed countries [1]. Except for some medications that
halt the growth of the disorders, effective therapies for
dementia and Alzheimer's disease are still elusive despite
advancements in medical science. Therefore, preventing
the spread of illnesses into their severe stages depends
greatly on early detection [1,2]. Some of the serious
diseases that have received a lot of attention in the mental
health field are dementia and Alzheimer's disease. This is
because of its prevalence in the elderly and its negative
impact on the elderly's ability to perform daily tasks.
Dementia is memory loss or impairment that prevents
mental health from being maintained due to aging or
illness. It is characterized by changes in mental and
behavioral disorders or stroke. It is a syndrome that
includes impaired memory, behavior and thinking and the
loss of ability to perform daily activities [3,4]. Reports
from World Health Organization (WHO) state that around
47 million people across the world live with dementia. It
could reach 82 million by 2030. The root cause of
dementia is neurodegeneration and poor connections in
the brain, which leads to poor decision making skills. Non-
neurodegenerative mechanisms cause vascular dementia.
Alzheimer's disease (AD) is one of the most common and
common forms of dementia, accounting for 60% to 70% of
dementia cases. Age is a risk factor for AD, especially in
people over the age of 65. AD is more commonly found in
women than men. However, the aetiology of AD has not
been correctly determined by the medical personnel.. The
main idea is based on the combination of extracellular Aβ
peptide and hyperphosphorylated tau protein in brain
cells. These two patterns are biomarkers called amyloid
plaques (aggregation of beta-amyloid fragments of
neurons) and tangles (intracellular accumulation of tau
protein in the form of twisted filaments).
1.1 Common Symptoms
Memory Loss:
The most common symptom of Alzheimer's disease is
memory loss. These include forgetting recent events,
forgetting names and faces, misplacing items, and
repeating questions.
Difficulty in planning and problem solving:
Alzheimer's patients often have problems planning and
solving problems. This can cause problems with tasks such
as paying bills, managing finances, and completing daily
tasks.
Speech Problems:
Alzheimer's patients may have difficulty finding the right
words to express themselves or to understand what others
are saying.
Mood and behavioural changes:
Alzheimer's disease can cause changes in mood and
behaviour, such as depression, anxiety, irritability, and
apathy.

1.2 Neuroimaging Modalities:
The use of MRI to diagnose Alzheimer's disease has
been the subject of research for many years. MRI
(magnetic resonance imaging) is a non-invasive technique
that provides detailed information about the brain. It is
widely used in clinical practice in the diagnosis and follow-
up of Alzheimer's disease. Neurons and synapses are lost
in the brain, and some brain areas shrink. MRI can detect
these changes.
2. DATA USED
The data is being sourced from Alzheimer diagnosed
dataset and around 6000 images were used for the model
training. The dataset comprises of four types of classes
non-demented Alzheimer, very mild demented Alzheimer,
mild demented Alzheimer and moderate demented
Alzheimer.
Fig 1: Input Images
3. PROPOSED MODEL
In this section, we discuss our proposed model that
consists of CNN model and the following steps.
Fig 2: Proposed System Architecture
3.1 Data preprocessing
Preprocessing is used to improve image data by removing
unwanted distortions and improving certain views that
are important for further processing. Tagging is a type of
image manipulation that combines multiple scenes into a
single image. It aids in resolving issues with overlapping
pictures' size, contrast, and image rotation. Combining the
picture data from many photos and transforming them to
the same coordinate system is known as image
registration. It has several applications in clinical and
medical research. Images taken for medical purposes can
be collected from the same person at the same time using
multiple models, or from different persons using different
models. For optimum results, it's crucial to convert the
MRI pictures in the file to the same width and height as
they differ in size. Since the input image size of the CNN
model is 224×224 pixels, this research reduces the MRI
image to 224×224.
3.2 Convolution Neural Network
Convolution neural networks are a subset of deep neural
networks which make use of convolutional layers for
processing inputs for the included images. The
convolutional layers of CNN compute the output of
neurons connected to specific regions in the input and
apply convolutional filters to the input. It assists in
extracting spatial and temporal information from images.
A weight-sharing method is used in CNN's convolutional
layers to reduce the overall number of parameters.
3.2.1 Feature Extraction
This work uses a CNN-based model to extract key features
without human interference. The proposed architecture
consists of four convolutional layers, a max-pooling layer,
dropout, flatten and a fully connected layer. The output
ranges from non-demented to moderately demented.
3.2.2 Layers
Convolutions
Using a kernel size of 45*45*45, convolution operations
are performed on an image of 8-block size. There are two
convolutional layers employed, and the first filter has 32
3*3 kernels. The kernel's size denotes a neuron's receptive
field, reinforcing the neurons' local link to the prior
volume.
Rectified Linear Unit and Softmax
The Activation function of ReLU is defined [11] and the
Softmax function let the model to express the inputs as a
discrete probability distribution. In ReLUs the training
time is significantly faster as compare to sigmoid units and
hyperbolic tangent [12].
Pooling Layer
The aggregation function max-pooling is used to obtain
the maximum value, as determined by the kernel size,

input hxw size, and stride. The pooling approach
effectively summarises the outputs of adjacent groups of
the inputs in addition to reducing the inputs' dimensions
[9]. When the picture size is very huge, this layer really
performs down sampling, which minimises the spatial
dimensions while maintaining valuable information and
also reduces the number of parameters [10] in this work 1
max pooling layer is used.
Dropout
The output of neurons with a ratio dropout, or prob-ability
of r, is adjusted to 0 by the usage of dropout layers in the
hidden layers. The forward pass and the backpropagation
processes are not affected by the neurons that dropped
out. Two dropout layers, with ratios of 0.25 and 0.5, have
been added to the design that we suggest.
Full Connected Layer
The last layer, which we refer to as the FC layer, is fully
connected; each of its neurons is connected to the layer
above it, and it also enhances the training performance of
CNN models since we flattened our matrix into a vector
form and fed it into the fully connected layer. [19]. All
activations in the layer below it are fully connected to this
layer.
4. OUTPUT
The output consists of the MRI scan with the assigned
label.
Labels can be from very mild, mild, moderate and non
demented. Output can get verified by a neurosurgeon and
add in to acccuracy of model.
Fig 3: Output of the user Interface
5. RESULTS AND DISCUSSION
In this study, we employed MRI scan pictures that were
characterized as having moderate dementia, non-
dementia, or very mild dementia. We randomly selected
80% of the training data, while the remaining 20% were
used to validate the model.
The CNN network has several layers, including a
convolutional, activation, pooling, and fully connected
layer. The activation layer applies the Rectified Linear Unit
(ReLU) to increase the nonlinear properties in the CNN
model because of its training speed. The first layer is the
convolutional layer, which takes the input image using a
kernel (ReLU) or filter and identifies the relationship
between the image and their features (to identify whether
the image is of an Alzheimer's patient or Normal). Since
we flattened our matrix into a vector form and sent it into
the fully connected layer, the fully connected layer
ultimately enhances the training performance of the
models. Using MRI images, CNN was employed in this
study to identify and predict AD. at this model, we were
able to train and test the model using 6000 photos while
also achieving a test accuracy rate of 0.98% and a low
proportion of test loss at a rate of 0.0667. We used four
alternative epoch sizes throughout the model's testing and
training in order to compare the findings and determine
which one produced the most accurate outcome. With
respect to all three epochs, we improved test loss and
accuracy by employing 30 epochs. Table 1 depict the
output and the number of epochs used in CNN.
SN Epoch Size Test Loss Accuracy
1. 30 0.0667 0.9843
2. 20 0.0541 0.9789
3. 10 0.2385 0.9365
Table 1: Comparative Analysis of epoch size.
The accuracy and loss of the model's training and
validation are shown in the following graph. In the
following chart, training set is used to train the model,
while the validation set is used to assess the model's
performance.
Chart 1: Accuracy of validation and test data

6. CONCLUSIONS
Recent developments in biomedical engineering have
made the study and interpretation of medical pictures one
of the primary research fields [14], [15]. The usage and
use of DL is one of the factors contributing to this
advancement in the analysis of medical pictures [16]. In
the last year, DL has been mostly employed for
classification, and AI-based approaches are used to
automatically diagnose AD in its early stages to meet the
main objectives of doctors [17]. In order to identify AD
patients early, an automated framework and classification
for AD utilizing MRI images is crucial. In this research, we
use MRI scans to propose a convolutional neural network
classification approach for AD. 98% accuracy is a huge
accomplishment. A notable result was attained while
dealing with an epoch size of 30, with an accuracy rate of
98%, out of all the outcomes with various epochs.
Future work is something we anticipate and hope to
encourage. Consequently, the outcome might be further
enhanced by using deep convolutional neural networks,
which have recently demonstrated their usefulness in
neuroimaging studies. As a result, the algorithm's capacity
to identify AD would be greatly enhanced by the usage of
deep CNN and large MRI scan pictures. Additionally, this
deep learning technique offers invaluable information to
the researcher in order to diagnose various types of
diseases in addition to helping the doctor, carers,
radiologist, and patients who are afflicted with this
ailment.
REFERENCES
[1] Bhagtani A, Choudhury T, Raj G, Sharma M. An efficient
survey to detect Alzheimer disease using data mining
techniques. In: 2017 3rd International conference on
applied and theoretical computing and communication
technology (iCATccT). IEEE; 2017
[2] Simons S, Abasolo D, Escudero J. Classification of
Alzheimer’s disease from quadratic sample entropy of
electroencephalogram. Healthc Technol Lett 2015
[3] Sharma J, Kaur S. Gerontechnology-The study of
alzheimer disease using cloud computing. In: 2017
International conference on energy, communication, data
analytics and soft computing (ICECDS). IEEE; 2017
[4] La Joie R, Bejanin A, Fagan AM, et al. Associations
between [(18)F]AV1451 tau PET and CSF measures of tau
pathology in a clinical sample. Neurology. 2018
[5] Daza JC, Rueda A. Classification of Alzheimer’s disease
in MRI using visual saliency information. In: Computing
conference (CCC), 2016 IEEE 11th Colombian. IEEE; 2016.
[6] Chupin M, Gérardin E, Cuingnet R, Boutet C, Lemieux L,
Lehéricy S, Benali H, Garnero L, Colliot O. Fully automatic
hippocampus segmentation and classification in
Alzheimer’s disease and mild cognitive impairment
applied on data from ADNI. Hippocampus 2009
[7] C. Patterson, World Alzheimer Report 2018-the State
of the Art of Dementia Research: New Frontiers. London,
U.K.: Alzheimer’s Disease International, 2018
[8] D. Shen, G. Wu, and H. Suk, ‘‘Deep learning in medical
image analysis,’’ Annu. Rev. Biomed. Eng., vol. 19, pp. 221–
248, Jun. 2017.
[9] Y. Lecun, Y. Bengio, and G. Hinton, “Deep learning,”
Nature, vol. 521, no. 7553, pp. 436–444, 2015, doi:
10.1038/nature14539.
[10] A. Krizhevsky and G. E. Hinton, “ImageNet
Classification with Deep Convolutional Neural Networks,”
pp. 1–9.
[11] T. Wood, “Softmax Function Definition DeepAI.” 2020.
[12] G. Hinton, “Dropout : A Simple Way to Prevent Neural
Networks from Overfitting,” vol. 15, pp. 1929– 1958, 2014.
[13] Y. Li, D. Shi, B. Ding, and D. Liu, “Unsupervised Feature
Learning for Human Activity Recognition Using
Smartphone Sensors,” pp. 99–100, 2014.
[14] A. W. Salehi, P. Baglat, and G. Gupta, “Materials Today
: Proceedings Review on machine and deep learning
models for the detection and prediction of Coronavirus,”
Mater. Today Proc., 2020, doi:
10.1016/j.matpr.2020.06.245.
[15] K. Sethi, “Machine Learning Based Performance
Evaluation System Based On Multi-Categorial Factors,”
2018 Fifth Int. Conf. Parallel, Distrib. Grid Comput., pp. 86–
89, 2018.
[16] S. M. Anwar, M. Majid, A. Qayyum, M. Awais, M.
Alnowami, and M. K. Khan, “Medical Image Analysis using
Convolutional Neural Networks: A Review,” J. Med. Syst.,
vol. 42, no. 11, 2018, doi: 10.1007/s10916- 018-1088-1.
[17] K. Sethi, “Comparative Analysis of Machine Learning
Algorithms on Different Datasets,” no. Icic 2017, pp. 87–
91, 2018

Alzheimer Detection System

More Related Content

What's hot (20)

Similar to Alzheimer Detection System (20)

More from IRJET Journal (20)

Recently uploaded (20)

Alzheimer Detection System