Handbook of Decision Support Systems for Neurological Disorders

Handbook of Decision Support Systems for Neurological Disorders provides readers with complete coverage of advanced computer-aided diagnosis systems for neurological disorders. While computer-aided decision support systems for different medical imaging modalities are available, this is the first book to solely concentrate on decision support systems for neurological disorders. Due to the increase in the prevalence of diseases such as Alzheimer, Parkinson’s and Dementia, this book will have significant importance in the medical field. Topics discussed include recent computational approaches, different types of neurological disorders, deep convolution neural networks, generative adversarial networks, auto encoders, recurrent neural networks, and modified/hybrid artificial neural networks.

• Includes applications of computer intelligence and decision support systems for the diagnosis and analysis of a variety of neurological disorders
• Presents in-depth, technical coverage of computer-aided systems for tumor image classification, Alzheimer’s disease detection, dementia detection using deep belief neural networks, and morphological approaches for stroke detection
• Covers disease diagnosis for cerebral palsy using auto-encoder approaches, contrast enhancement for performance enhanced diagnosis systems, autism detection using fuzzy logic systems, and autism detection using generative adversarial networks
• Written by engineers to help engineers, computer scientists, researchers and clinicians understand the technology and applications of decision support systems for neurological disorders

• Language: English
• Publisher: Academic Press
• Release date: Mar 30, 2021
• ISBN: 9780128222720

Book preview

Handbook of Decision Support Systems for Neurological Disorders

Editor

Hemanth D. Jude

Department of Electronics and Communication Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India

Table of Contents

Cover image

Title page

Copyright

Contributors

Preface

Chapter 1. A review of deep learning-based disease detection in Alzheimer's patients

1.1. Introduction

1.2. Literature review

1.3. Methods of Alzheimer's detection using neuroimaging data

1.4. Comparison of detection methods

1.5. Conclusion

Chapter 2. Brain tissue segmentation to detect schizophrenia in gray matter using MR images

2.1. Introduction

2.2. Data collection

2.3. Methods

2.4. Results and discussion

2.5. Conclusion

Chapter 3. Detection of small tumors of the brain using medical imaging

3.1. Introduction

3.2. Types of medical images used in the application

3.3. Theoretical background of the application

3.4. Description of the application

3.5. Testing, installing, and using the application

3.6. Conclusions

Chapter 4. Fuzzy logic-based hybrid knowledge systems for the detection and diagnosis of childhood autism

4.1. Introduction

4.2. The advent of machine learning: a new horizon for autism

4.3. Fuzzy-based systems for autism

4.4. Moving ahead: virtual reality for autism

4.5. Current scenario and scope of improvement

4.6. Discussion and conclusion

List of abbreviations

Chapter 5. Artificial intelligence for risk prediction of Alzheimer's disease: a new promise for community health screening in the older aged

5.1. Introduction

5.2. Etiology and risk factors

5.3. Screening and early detection

5.4. Current methods of early detection

5.5. The rise of AI—a new promise for community health screening

5.6. Common algorithms used in ML for dementia and AD detection

5.7. Artificial intelligence methodologies for screening AD: concept examples

5.8. Promises and challenges of AI applications for predicting AD

5.9. Conclusions and future direction

Chapter 6. Cost-effective assistive device for motor neuron disease

6.1. Introduction

6.2. Motor neuron diseases

6.3. System configuration

6.4. Experimental results

6.5. Conclusion

Chapter 7. EEG signal-based human emotion detection using an artificial neural network

7.1. Introduction

7.2. EEG signal data acquisition

7.3. Statistical features extracted from EEG

7.4. Various ANN methods to classify EEG data

7.5. Classification of EEG-based emotion using ANN

7.6. Experimental analysis

7.7. Conclusion

Chapter 8. Multiview decision tree-based segmentation of tumors in MR brain medical images

8.1. Introduction

8.2. Multiview decision tree-based segmentation

8.3. Results and discussion

8.4. Conclusion

Chapter 9. Multiclass SVM coupled with optimization techniques for segmentation and classification of medical images

9.1. Introduction

9.2. Classification of support vector machine

9.3. Parameter tuning of multiclass SVM using optimization algorithms

9.4. Results and discussion

9.5. Conclusion

Chapter 10. Brain tissues segmentation in magnetic resonance imaging for the diagnosis of brain disorders using a convolutional neural network

10.1. Introduction

10.2. Materials and methods

10.3. Results and discussion

10.4. Conclusion and future work

Chapter 11. Fine motor skills and cognitive development using virtual reality-based games in children

11.1. Introduction

11.2. Neurological disorders in children and the role of VR games in children's rehabilitation

11.3. Leap Motion Controller and Unity3D

11.4. Game design and working

11.5. Experimental results and discussion

11.6. Future work

11.7. Conclusions

Chapter 12. A CAD software application as a decision support system for ischemic stroke detection in the posterior fossa

12.1. Introduction

12.2. Computer-aided diagnosis system implementation

12.3. Proposed CAD system for ischemic stroke detection

12.4. Experimental results and discussions

12.5. Conclusion

Chapter 13. Optimization-based multilevel threshold image segmentation for identifying ischemic stroke lesion in brain MR images

13.1. Introduction

13.2. Methodology

13.3. Results and discussion

13.4. Conclusion

Chapter 14. A study of machine learning algorithms used for detecting cognitive disorders associated with dyslexia

14.1. Introduction to neurological disorders

14.2. Classification of neurological disorders

14.3. Machine learning algorithms

14.4. Conclusion

Chapter 15. A Critical Analysis and Review of Assistive Technology: Advancements, Laws, and Impact on Improving the Rehabilitation of Dysarthric Patients

15.1. Introduction

15.2. Dysarthria

15.3. Assistive technologies for dysarthria

15.4. Design considerations in the development of Assistive Technology

15.5. Assistive technology laws for improving the rehabilitation of dysarthric patients

15.6. Impact of assistive technology on quality of life of dysarthric individuals

15.7. Inference

15.8. Summary of observations

15.9. Conclusion

Chapter 16. A comparative study on the application of machine learning algorithms for neurodegenerative disease prediction

16.1. Introduction

16.2. Literature survey

16.3. Description of the dataset

16.4. Support vector machine

16.5. Decision tree

16.6. Random forest tree

16.7. Results and discussions

16.8. Conclusion and future enhancements

Index

Copyright

Academic Press is an imprint of Elsevier

125 London Wall, London EC2Y 5AS, United Kingdom

525 B Street, Suite 1650, San Diego, CA 92101, United States

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom

Copyright © 2021 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-822271-3

For information on all Academic Press publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Mara Conner

Acquisitions Editor: Chris Katsaropoulos

Editorial Project Manager: Fernanda Oliveira

Production Project Manager: Prem Kumar Kaliamoorthi

Cover Designer: Victoria Pearson

Typeset by TNQ Technologies

Contributors

Akshay Aggarwal,     Bharati Vidyapeeth's College of Engineering, Delhi, India

R. Amutha,     Department of ECE, Sri Sivasubramaniya Nadar College of Engineering, Chennai, Tamil Nadu, India

K. Ashwini,     Department of ECE, Sri Sivasubramaniya Nadar College of Engineering, Chennai, Tamil Nadu, India

Alan Swee Hock Ch'ng

Clinical Research Center, Seberang Jaya Hospital, Ministry of Health Malaysia, Seberang Jaya, Penang, Malaysia

Department of Medicine, Seberang Jaya Hospital, Seberang Jaya, Penang, Malaysia

Rino Cherian,     KVGCE, CSE, Sullia, Karnataka, India

Nisha Dayana,     Dhanraj Baid Jain College, Chennai, Tamil Nadu, India

A. Lenin Fred,     Mar Ephraem College of Engineering and Technology, Elavuvilai, Tamil Nadu, India

Nik Farhan Nik Fuad,     Department of Radiology, UKM Medical Centre, Kuala Lumpur, Malaysia

Kurubaran Ganasegeran,     Clinical Research Center, Seberang Jaya Hospital, Ministry of Health Malaysia, Seberang Jaya, Penang, Malaysia

R. Geetha,     Department of Electronics and Communication Engineering, Sri Sairam Engineering College, Chennai, Tamilnadu, India

M.S. Geetha Devasena,     Department of Computer Science and Engineering, Coimbatore, Tamil Nadu, India

Varun P. Gopi,     Department of Electronics and Communication Engineering, National Institute of Technology, Tiruchirappalli, Tamilnadu, India

G. Gopu,     Department of Electronics and Communication Engineering, Sri Ramakrishna Engineering College, Coimbatore, Tamil Nadu, India

Balazs Gulyas,     Nanyang Technological University, Singapore

Ajay Kumar Haridhas,     Mar Ephraem College of Engineering and Technology, Elavuvilai, Tamil Nadu, India

G. Indumathi,     Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India

Rachna Jain,     Bharati Vidyapeeth's College of Engineering, Delhi, India

Jothimani,     KVGCE, CSE, Sullia, Karnataka, India

S.N. Kumar,     Amal Jyothi College of Engineering, Kottayam, Kerala, India

Vaibhav Kumar,     Bharati Vidyapeeth's College of Engineering, Delhi, India

Irene Looi

Clinical Research Center, Seberang Jaya Hospital, Ministry of Health Malaysia, Seberang Jaya, Penang, Malaysia

Department of Medicine, Seberang Jaya Hospital, Seberang Jaya, Penang, Malaysia

Ramalatha Marimuthu,     Department of Electronics and Communication Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India

Antoanela Naaji,     Vasile Goldis Western University of Arad, Arad, Romania

Ain Nazari,     Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor, Malaysia

J. Neelaveni,     Department of Computer Science and Engineering, Coimbatore, Tamil Nadu, India

Parasuraman Padmanabhan,     Nanyang Technological University, Singapore

R. Ponuma,     Department of ECE, Sri Sivasubramaniya Nadar College of Engineering, Chennai, Tamil Nadu, India

Marius Popescu,     Vasile Goldis Western University of Arad, Arad, Romania

E. Priya,     Department of Electronics and Communication Engineering, Sri Sairam Engineering College, Chennai, Tamilnadu, India

Sahar Qazi,     Department of Computer Science, Jamia Millia Islamia, New Delhi, India

Fakhrul Razan Rahmad,     Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor, Malaysia

Khalid Raza,     Department of Computer Science, Jamia Millia Islamia, New Delhi, India

S.R. Reeja,     School of Engineering, CSE, Dayananda Sagar University, Bangalore, Karnataka, India

Manicka Nagarajan Saroja,     Department of Electronics and Communication Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India

V. Sathananthavathi,     Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India

D. Selvathi,     Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamilnadu, India

Arun Kumar Shanmugam,     Department of Electronics and Communication Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India

Sathyamangalam Natarajan Shivappriya,     Department of Electronics and Communication Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India

Anis Azwani Muhd Suberi

Faculty of Engineering and Information Technology, Southern University College, Skudai, Johor, Malaysia

Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor, Malaysia

Razali Tomari,     Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor, Malaysia

Kiran Waghmare,     School of Engineering, CSE, Dayananda Sagar University, Bangalore, Karnataka, India

Wan Nurshazwani Wan Zakaria,     Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor, Malaysia

Preface

The human brain is one of the most significant organs of the human body. This significance is mainly because the human brain has not yet been completely explored. Ironically, disorders that are related to the human brain pose a major challenge for physicians in the context of finding a cure for many neurological disorders. Hence, medical specialists need the help of engineers to explore the human brain, to diagnose the abnormalities, and to make decisions for treatment planning. Currently, much research on computer-aided systems is being carried out at a significant pace. However, with the sprouting of new neurological disorders, the need for improvement in computer-aided systems for disease diagnosis is increasing every day. This book has been developed with this background in mind. Several novel computer-aided diagnosis systems are explored for different neurological applications, which are challenging to human society.

Alzheimer’s disease has become one of the common disorders among the aged community. A complete review of the practical feasibility of computer-aided disease diagnosis methods is discussed in Chapter 1. The inclusion of deep learning for this analysis adds weight to this chapter. Detection of schizophrenia using magnetic resonance (MR) brain images is the focal point of Chapter 2. Markov random field-based brain tissue segmentation is carried out in this chapter for the detection of this abnormality. Chapter 3 is focused on brain tumor detection in MR images. Specially framed software for the segmentation of brain tumors is discussed in this chapter. The hardware set-up necessary for this application is also discussed in detail. Autism has become a challenging disorder among children, and in Chapter 4, a combination of intelligent agents and fuzzy logic is used for the diagnosis of this specific disorder. Another work on Alzheimer’s disease is discussed in Chapter 5. This chapter talks about the possibility of AI-based community screening rather than the technologies behind the diagnosis process.

Assistive devices for motor neuron diseases is the focus of Chapter 6. The design set-up for assistive devices along with the hardware and software details are discussed in this chapter. Cost-effective set-up, which is the need of the hour, is the main objective of this work. Electroencephalogram signal-based analysis for the detection of human mental behavior is discussed in Chapter 7. Artificial neural networks are used in this work for experimental analysis. Chapter 8 deals with MR brain image segmentation using decision tree approaches. This method is used to segment the different tissue types of the human brain, which can be used for the diagnosis of any disorder. Support vector machine-based classification and segmentation of medical images are discussed in Chapter 9. Optimization techniques are also used in this work, which adds an additional dimension to the conventional medical image analysis system. Deep learning-based brain image analysis is the focal point of Chapter 10. Convolutional neural networks are used in this work for the detection of different brain disorders using MR brain images.

Cognitive development in children is always a challenging task, especially for children with neurological disorders. Chapter 11 provides a possible solution to this problem by developing game-based therapy for such children. The inclusion of virtual reality for game development is an attractive feature of this work. Detection of ischemic stroke in computed tomography (CT) images is the focal point of Chapter 12. Deep learning-based methodologies are used in this work to detect the abnormalities in the posterior region of the brain. Lesion detection in MR images is discussed in detail in Chapter 13. Detection is carried out with novel hybrid methods, which involve the concepts of optimization and thresholding techniques. Chapter 14 deals with dyslexia, which is another challenging disease in this neurological category. A detailed literature review on the various types of computer-aided systems and machine-learning techniques used for diagnosis is available in this chapter. Assistive technology is the main theme of Chapter 15. A detailed literature review on various assistive devices available for patients with neurological disorders is discussed in this chapter. Much emphasis is given to the rehabilitation process of dysarthric patients. Disease prediction is the future of the medical diagnostic process. An extensive survey on the possible prediction of neurodegenerative disorders using machine-learning techniques is available in Chapter 16.

On the whole, the contents of this book are an excellent mixture of artificial intelligence techniques and the neurological disorders of human beings. We are grateful to the authors and reviewers for their excellent contributions, which make this book possible. Our special thanks go to Elsevier, especially to Mr. Chris Katsaropoulos (Senior Acquisitions Editor) for his excellent collaboration. Finally, we would like to thank Ms. Oliveira Fernanda who coordinated the entire proceedings. This edited book covers the fundamental concepts and application areas in detail, which is one of the main advantages of this book. Being an interdisciplinary book, we hope that it will be useful to a wide variety of readers and will provide useful information to professors, researchers, and students.

Hemanth D. Jude

Chapter 1: A review of deep learning-based disease detection in Alzheimer's patients

Rachna Jain, Akshay Aggarwal, and Vaibhav Kumar     Bharati Vidyapeeth's College of Engineering, Delhi, India

Abstract

Early and accurate diagnosis is a significant step in lowering the risk of Alzheimer's progression. Brain-imaging techniques can help pinpoint the abnormalities related to Alzheimer's. Using artificial intelligence-driven systems, brain images are classified into healthy and abnormal based on the extracted features. These features are capable of precisely detecting the Alzheimer's-related symptoms by identifying changes in the brain segments such as ventricle size, hippocampus shape, cortical thickness, and brain volume. This chapter discusses how Alzheimer's affects different parts of the brain, the stages of Alzheimer's, neuroimaging techniques that are being used to visualize the effects, resources of relevant datasets, and how different AI-based methods developed over the years are able to identify changes in the brain and then formulate the defining features. Finally, we compare all the discussed methods with a relevant performance metric.

Keywords

Alzheimer's disease; Artificial intelligence; Convolutional neural networks; Deep learning; Feature-based methods; Neuroimaging

1.1. Introduction

In today's technology-driven world, artificial intelligence (AI) can be used to correctly identify probable symptoms of certain diseases and assist doctors in determining whether a person with positive symptoms of a specific disease has that disease. Such advanced technologies are a great advantage in the medical field as they help in automating the process of diagnosis of chronical diseases. In this chapter, we discuss one such chronic disease—Alzheimer's. Alzheimer's disease (AD) is an incurable, chronically progressive cause of dementia that results in deprivation of cognitive abilities. Hence, to allow patients to take preventive measures and slow down the worsening of symptoms before the brain is damaged irrevocably, it is of great importance to utilize AI-based computer-aided systems for accurate and effective diagnosis at the prodromal stage. Popular neuroimaging or brain-imaging techniques such as magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET) can provide potential multimodal information to analyze anatomical neural changes and help develop such systems. In earlier systems, information from the modalities was extracted by computing and hand engineering features such as hippocampus volumes, amygdala volumes, cortical thickness, gray matter (GM) densities, etc. However, with the recent advent of deep learning technologies in numerous fields, researchers are inclining more toward layered networks that have the advantage of providing multilevel and shared feature representation for the input data. Convolutional neural networks (CNN), a deep learning method, have predominantly excelled for various computer vision domains, including the medical imaging field. Various CNN architectures have been exploited for the task of accurate prediction of AD within the past few years. These architectures usually take as input one or a combination of modalities for neuroimaging data and output a class score. This chapter discusses how deep learning is being leveraged to build robust, end-to-end systems to assist neurologists for the appropriate diagnosis of Alzheimer's patients.

1.1.1. Alzheimer's disease

AD is named after Dr. Alois Alzheimer who studied the damage in the neural tissues of a deceased patient who expired due to a rarely encountered mental sickness. As discussed earlier, AD progressively deteriorates brain cells and slowly affects the memory and thinking capability of the affected person, gradually taking away a sense of oneself while worsening the ability to do simple daily life chores. The disease ranks among the top causes of death globally, which shows the severity of the disease and why it must be diagnosed and treated as early as possible.

Dementia by definition is the loss of memory and ability to do everyday activities, and Alzheimer's is the top reason for dementia among elderly people. To be precise, dementia is the later stage of AD, as in the early stages, the affected person does not have to depend on others for their personal and everyday activities. Generally, Alzheimer's-led dementia occurs in the mid-60s; the symptoms, according to the Alzheimer's Association, arise around or after 65  years of age [1].

Different changes in brain tissues lead to varying forms of dementia, and some common dementias include frontotemporal disorders and vascular dementia. It can be noted that people may suffer from a mixed form of dementia. For instance, some people have both later forms of AD and frontotemporal disorders [3]. Since the brain tissues are affected therefore, the side-by-side comparison in Fig. 1.1 shows how severely AD can affect a human brain.

1.1.1.1. Stages of Alzheimer's

AD is a progressive, incurable, neurodegenerating disease that can affect several parts of the brain resulting in a decline in performing everyday life activities. Thus the symptoms of the disease worsen gradually over time in different stages, and each stage has more severe effects than the previous one. As per the Alzheimer's Association, there are three stages in which this disease progresses: early, middle, and severe [4]. Also, it should be noted that each individual will experience the symptoms of these stages differently in different time periods.

Figure 1.1  Healthy brain versus Alzheimer's-affected brain [ 2].

At the early stage, the person is capable of functioning independently but may often find it difficult to remember names, routes, etc. As Alzheimer's is more common with individuals over 65  years old, this early stage may go unnoticed as it can be considered as a normal decline of functionality with age and can thus eventually lead to a more severe stage over time. Therefore neurologists regularly recommend accounting for the symptoms for early diagnosis. Sometimes, the normal process of aging, which is referred to as cognitively normal (CN), is also postulated as a preclinical stage of dementia due to the presence of beta-amyloid [5].

The second stage, popularly known as mild cognitive impairment (MCI), is the phase that causes the decline in cognitive impairment, which, as it becomes more severe, results in dementia [6]. Symptoms of MCI may or may not lead to Alzheimer's. The study of brain imagery at this stage shows several changes in the physical properties of the brain such as shrinkage of the hippocampus, which is an intricate structure in the temporal lobe and plays a vital role in learning and memory [7], increase in the brain's fluid-filled spaces and ventricle space, and reduction in the usage of glucose in parts of the brain.

The final stage, which is the severe stage and commonly referred to as AD, is the total impairment of cognitive abilities. At this stage, the individual loses the ability to interact with the environment and requires around-the-clock assistance for routine tasks. The causes discussed at the MCI stage become more severe during the final stage. Subsequently, more and more nerve cells die or become degenerated, and disease spreads through the whole cerebral cortex in the brain. An Alzheimer's patient, on average, may live up to 8–10  years after he/she has been diagnosed.

1.1.1.2. Cause of Alzheimer's

There is ongoing research into this disease as scientists have not yet fully understood or come to an agreement over what causes Alzheimer's in most humans. Some form of genetic mutation is mostly credited for the occurrence of early Alzheimer's but pinpointing the cause of late onset is still difficult. AD in later years could arise from a series of uncommon brain changes over many years. The brains of different persons react differently to some of the probable causes of AD, which include environmental, genetic, and lifestyle choices.

1.1.1.3. Signs and symptoms

The inception of Alzheimer's is marked with memory problems, which are among the first few signs of cognitive impairment led by AD. MCI occurs in patients with AD, which relates to increased memory difficulties as compared to people of the same age; surprisingly, the symptoms do not hugely affect everyday lives during the initial stages of the disease. MCI corresponds to difficulty in movement and impairment of the sense of smell. MCI is an important factor contributing to AD in older people.

1.1.1.4. Anatomical changes in the brain

The extent of complex changes in the brain can be directly correlated to the duration of progression of the disease. The longer the person suffers, the greater the extent to which brain tissues are affected; this progression may begin a few decades before MCI. These changes are the result of neuron changes. Unusual doses of protein from the amyloid affect the connectivity of the neurons; as healthy neurons lose connection they lose all of their motivity as cells inside die. The hippocampus (the inner brain structure responsible for emotional responses) takes the severest hit from the deteriorating functioning of the brain. The hippocampus and entorhinal cortex are essential parts of the brain for memory and retentive power. Consequently, dead neurons lead to shrunken tissues and other abnormalities. As time passes, the different levels of Alzheimer's can clearly be associated with the amount of damaged brain tissue. Alzheimer's can then be classified as mild, moderate, and severe, and as the name suggests these describe how much cerebral (the largest part of the brain) damage has taken place.

1.1.2. Deep learning

Deep learning is the term often used with AI, but not many people understand what it actually stands for, how it is employed, and what its relationship is with AI. Deep learning is a fascinating topic and is used for a multitude of applications from understanding the patterns in galaxies to self-operating robots.

The process of deep learning requires the intake of input data points and then development of hidden layers that can be used for predicting outcome. Without going into deep mathematics, the steps involved can be generalized as the selection of data points as inputs, multiplying the inputs with some predecided or arbitrary weights, the formation of hidden layers, and the obtaining of output. Plural hidden layers can be made by multiplying the hidden layer with other weights. Fig. 1.2 shows the basic model of neural networks and Fig. 1.3 shows intuition of the activation function in neural networks.

A loss function is basically a mathematical equation that quantifies the difference between the output received at the end of propagation through a neural network and the true output. The part where the learning takes place is the hidden layers or network layers that are being refined in every iteration where weights w and the bias b change to minimize the loss function. The weights are generally a multiplying factor and bias is an additive factor, and the changes in w and b signify learning in a neural network, where w and b show how different attributes are intercorrelated. The process of changing w and b to obtain the correct output by going back into the networks is known as backpropagation. Deep is just the term that means multiple layers of networks. These layers are useful in finding patterns in different variables that doctors have to deal with while treating AD. The input is the image frame that is obtained by the MRI, and these images are treated to make it feasible for our deep learning model to properly find the role of different factors in the appearance of tissue damage [9].

Figure 1.2  Basic model of neural networks [ 8].

Figure 1.3  Intuition of activation function in neural networks [ 8 ].

1.2. Literature review

1.2.1. Neuroimaging techniques used in Alzheimer's detection

Neuroimaging or brain-imaging techniques allow doctors to study the brain in living subjects or individuals without any invasive surgery. The techniques are primarily used to gain information about the structure, function, and pharmacology of the brain. In the past, neurophysiological structure and function associated with psychological processes could be studied to a substantial extent. This was because those studies could only be conducted on animal models, postmortem examinations, and on certain individuals who are already suffering from disorders of the brain. Development in neuroimaging techniques has helped to eliminate such methods of studying the brain.

In recent years, a combination of neuroimaging and deep learning have demonstrated tremendous and progressive improvements for early diagnosis of Alzheimer's. Thus it becomes essential to discuss some of the most popular neuroimage modeling techniques that are accepted unanimously throughout the world by researchers.

1.2.1.1. Magnetic resonance imaging

MRI is a harmless technique that makes use of a magnetic field and radio waves for generating a comprehensive representation of soft tissues of the structure being scanned. Specifically, MRI measures the fluid content of the tissues, which, when processed using software, creates a three-dimensional (3D) black and white image. There are three planes in which these images can then be viewed: axial (bottom), coronal (front), and sagittal (sideways). Fig. 1.4 depicts the sagittal view of a subject's brain MRI slice. This type of MRI for 3D structural representation is often referred to as sMRI. There is another variant of MRI that can be used for diagnosis called functional MRI (fMRI). In fMRI, both structural and functional activity of the brain are captured. fMRI works on the same principle as MRI and measures signal changes in the brain with a change in the activity performed. A patient is asked to perform specific actions so that the area of detection can be detected through blood flow by taking multiple pictures one after the other.

Using an MRI scan, doctors can view the subject's brain slice by slice, as if it were sliced layer by layer and each slice was pictured separately. MRI can distinguish brain abnormalities that are caused due to MCI and can then be further used to diagnose AD [11]. In the early stages, an MRI scan of the brain may be similar to that of a normal person's brain scan, but subsequent MRI scans can show relative reduction in the sizes of different parts of the brain such as temporal (side part of the brain) and parietal (middle part of the brain) lobes.

1.2.1.2. Computer tomography

A CT scan uses special X-ray measurements to generate a cross-sectional picture of the brain. As compared to standard X-rays of the head, CT scans are capable of providing detailed information about the structure and tissues of the brain. Though risks associated with CT scans are close to minimal, cumulative exposure to X-ray radiation over time can result in risks of diseases such as cancer.

During a CT scan, the subject is asked to lie on a scan table that slides into a large hollow, cylindrical structure. An X-ray source rotates on a loop around the subject inside the tube, with its beams aimed at the head. Images generated using X-rays are dependent on absorption by the body's tissues. While performing a CT scan is cheaper compared to MRI, for obtaining detailed images of the brain, MRI is prioritized over a CT scan.

Figure 1.4  Sagittal view of magnetic resonance imaging of the brain [ 10].

1.2.1.3. Positron emission tomography

PET is another imaging technique used by doctors to analyze the functioning of the brain. PET uses trace amounts of short-lived radioactive material that have been attached to compounds such as glucose and absorbed into the bloodstream to capture functional activities in the brain. Emission of positrons due to radioactive decay of this material is then picked up by a detector to formulate the information. The difference in the rate of utilization of glucose by dissimilar areas of the brain allows us to see the working of the brain and helps