You are on page 1of 8

Abasyn University Islamabad Campus

Final Year Project Proposal

For

Alzheimer’s disease Detection using Deep


Learning

By

Sayyad Hussain AUIC-19FALL-BSSE-3361


Bilal Shah AUIC-19FALL-BSCS-3472

Supervisor
Dr. Amjad Khan

Final Year Project proposal submitted to the Department of Computing, Abasyn


University Islamabad Campus in partial fulfillment of the requirements for the
degree of Bachelor of Science in Software Engineering (BSSE).

Department of Computing
Abasyn University Islamabad Campus
Fall 2022
Table of Contents

Table of Contents........................................................................................................................................2
1. Introduction.............................................................................................................................................3
2. Literature Review................................................................................................................................3
2.1. Nora...........................................................................................................................................4
2.2. PSYCOM......................................................................................................................................4
3. Problem Statement.............................................................................................................................4
4. Project objective.................................................................................................................................5
5. Significance of the Project..................................................................................................................5
6. Project Methodology..........................................................................................................................5
6.1. Convolutional Neural Networks (CNN) [2]..................................................................................6
7. Data Gathering Approach...................................................................................................................7
8. FYP Timeline........................................................................................................................................7
9. References..........................................................................................................................................7
1. Introduction
Alzheimer is a brain disorder that slowly destroys memory and thinking and eventually the
ability to perform the simplest tasks. People with Alzheimer's disease also experience changes in
behavior and personality. It is commonly found in the aged people, typically the people having
the age-range of 65-75 years. However, this can also be found in the younger people but the
chances to existence is commonly lesser.
The symptoms of Alzheimer’s can vary from one person to another. Memory problems are
typically one of the first signs of the disease. Decline in non-memory aspects of cognition, such
as finding the right word, trouble understanding, and impaired reasoning or judgment, may also
signal the early stages of Alzheimer’s. As the disease progresses, symptoms become more severe
and include increased confusion and behavior changes. Many of the studies suggest that the most
familiar symptoms of Alzheimer’s include:
 Poor judgment, leading to bad decisions.
 Loss of spontaneity and sense of initiative.
 Losing track of dates or knowing current location.
 Taking longer to complete normal daily tasks.
 Repeating questions or forgetting recently learned information
Practically, no effective cure to treat Alzheimer's disease exist to date. However, the early
detection can temporarily slow down the process of Alzheimer’s symptoms and improve the
quality of the life of the patient. It is suggested that ML and DL algorithms, which have proven
their significance in various fields, can help solve the problem of early detection of Alzheimer's
disease

2. Literature Review
Technologies can benefit many people in healthcare, such as machine learning, reducing the
workload of doctors, psychologists, radiologists, pathologists and other providers by using
algorithms to work with insights. Automated tools designed for healthcare are making the real
world better than ever before.
One application of machine learning in the healthcare context is digital diagnostics. ML can
detect patterns of certain diseases in a patient's electronic health records and notify doctors of
any anomalies. In this sense, the machine learning technique can be a second pair of eyes. Thus,
ML opens up the potential to the diagnose systems that are previously needs human efforts like
NORA, PSYCOM etc.

2.1. Nora
Nora is a web-based framework for medical image analysis. It has been developed to bridge the
gap between research and clinic, [1] and to boost medical imaging research to the next level. It
provides a high-level web-interface accessible from any web-browser to visualize, organize,
process and share data in a very customizable way. Depending on your needs, your Nora
instance can run as a web-service in the cloud or as a local installation at your institution.

2.2. PSYCOM
Psycom covers a variety of mental health [2] conditions including bipolar disorder, ADHD,
Alzheimer, schizophrenia, depression, and generalized anxiety disorder. Seeks to inspire patients and
their caregivers to better understand their mental health conditions and to take an active role in their
own care.

3. Problem Statement
Alzheimer is traditionally diagnosed through a lengthy assessment and requires the presence of a
specialist. These assessments can be expensive and difficult for some people to approach. With
the advancement of technology and the availability of smart devices, many opportunities have
been created, one of which is medical applications to provide information that were not available
to everyone in the past. The technology is useful in diagnosing mental disorders such as
Alzheimer, a neural disability in which people slowly and gradually lost their memory due to
shrinking of brain and eventually died.

4. Project objective
Following are the main aims and objectives of our project:
 To make an application that will help in initial diagnosis of Alzheimer.
 To provide platform that will help patient to identify the severity of this disorder.
 To evaluate the developed model with the tested dataset.
 To test and check the working of CNN model’s algorithm of TensorFlow-lite and Google
teachable machine

5. Significance of the Project


The primary goal of our project is to facilitate users get a quick and early diagnosis for
Alzheimer, as an early identification of these disorders is beneficial. Our proposed diagnosis
system will effectively diagnose Alzheimer, especially focusing on elders. For this approach we
will be using the Google-teachable-machine and TensorFlow. Teachable Machine is a web-based
tool that makes creating machine learning models fast, easy, and accessible to everyone.[4] The
model creates on this framework perform accurate with a test accuracy rate of up to 100%. On
the order hand JavaScript library which is TensorFlow is used for the implementation of the
algorithm.

6. Project Methodology
In recent years, Deep Learning (DL) techniques based on Artificial Neural Network (ANN) have
been successfully applied for diagnosis and detection of diseases. The basic operation in the
networks established in these techniques is designed by inspiring the work of the human brain.
Nonlinear operations are applied using many signals entering the system and output signals are
created. These techniques reflect representative learning. [3] Because these models do not use an
explicit method when extracting features from the data. The feature extraction process is
performed by using hidden data layers from the data. For the feature extraction process in deep
learning models, it is not necessary to extract features separately as in the machine learning
technique. The techniques such as support vector machine, k-tools and neural network were used
in the diagnosis of neurological diseases [2]. In addition, the techniques used in the diagnosis of
some disorders that cause disease in humans have been inspired by nature and have provided
important achievements. Early diagnosis in Alzheimer's disease (AD) is very important in
preventing the progression of the disease and ensuring its treatment. For AD detection, features
extracted from brain MR images are classified. Features should accurately represent AD-related
regions of structures such as ventricular size, hippocampus shape, cortical thickness, and brain
volume. The MRI neuroimaging tool has been explored to develop such a system. Features
derived from existing MRI images are divided into three groups of Alzheimer's disease [5]:

 Mild-Demented
 Moderate-Demented
 Very-Mild-Demented
 Non-Demented

group and classification are established. This system examines the deep learning methods used to
diagnose Alzheimer's and dementia based on MRI scans Convolutional neural networks (CNN)
are important and widely preferred deep learning models. [5] As a result of the research, it has
been seen that the methods are also applied in different diseases.

6.1. Convolutional Neural Networks (CNN) [2]


CNN is one of the most widely used deep learning techniques. CNN is a model in which the
image information is taken from the input in the method, transferred to the output in one
direction and obtained the output according to the desired classification. CNN consists of input,
output and multiple hidden layers. CNN topology uses relationships between data to reduce the
number of parameters required in detection. At the same time, to these relationships, forward
feeding improves. An exemplary model is shown in Figure. This proposed model is called
adaptable CNN. While the lower layers of a CNN developed for a specific purpose separate and
detect the general properties, the upper layers make it easier to classify the problem by using
these properties.
7. Data Gathering Approach
For the data gathering approach, we will survey hospitals specializing in the diagnosis of this
disorders and conduct interviews with several psychiatrists and psychologist to get an
understanding of the methods and procedures they use for diagnosis.
Currently dataset available online for our targeted project, we will first develop a learning
activity to validate the data and then implements it in our project.

8. FYP Timeline

Level 0 Level 1 Level 2


1.1.1 Proposal Document
1.1 Planning
1.1.2 SRS Document

Module 1: Login
Module 2: Web-pages development.
Alzheimer’s disorder
Detection using Deep Module 4: Algorithm Development.
Learning 1.2 Implementation
Module 4: Training the Algorithm.
Module 5: Integration
Module 6: Reports and suggestions

1.3 Testing 1.3.1 Project Testing

1.4 Deployment 1.4.1 Project Improvement.

9. References
1. Sabuncu, M.R. et al., (2015) Clinical Prediction from Structural Brain MRI Scans: A
Large-Scale Empirical Study. Neuroinformatics, vol. 13, no. 1, pp. 31–46.
2. Deng, L., (2014) A tutorial survey of architectures, algorithms, and applications for deep
learning. APSIPA Transactions on Signal and Information Processing 3:1–29.
3. Gulhare, K. K., (2017) Shukla, S. P., and Sharma, L. K., Deep Neural Network
Classification method to Alzheimer’s Disease Detection. International Journals of
Advanced Research in Computer Science and Software Engineering 7(6):1–4
4. Long, M. and Wang, J. (2015) Learning transferable features with deep adaptation
networks. arXiv:1502.02791 [cs.LG].
5. Ravì, D., Wong, C., Deligianni, F., Berthelot, M., Andreu-Perez, J., Lo, B., and Yang,
G., (2017) Deep learning for health informatics. IEEE Journal Of Biomedical and Health
Informatics 21(1):4–21.
6. Pouyanfar, S., Sadiq, S., Yan, Y., Tian, H., Tao, Y., Reyes, M. P., and Iyengar, S. S.,
(2018) A survey on deep learning: Algorithms, techniques, and applications. ACM
Computing Surveys (CSUR) 51(5): 92.

You might also like