Abstract
COVID-19 is a worldwide pandemic since the beginning of 2020. It records a high death rate across many countries in the world. Efforts have been put in place to control the spread and the associated deaths. Vaccination, isolation, mass testing, and artificial intelligence models have been used to control the disease. Due to the high number of cases per day, manual monitoring of progression has been difficult and associated with false negatives. Ground-Glass opacities (GGOs) identification has been used in the detection and classification of COVID-19 positive and negative cases. The localization of the GGOs and the volumes in the lungs can be used to identify and monitor the COVID-19 progression in the lungs. This research developed an adoptive computational model to help in COVID-19 monitoring. It identifies the localization of GGOs in the lungs responsible for COVID-19. The research also used automated feature extraction convolutional neural networks (CNN) models to enhance speed and accuracy. Feature extraction and modeling were done with standard CNN and CNN with transfer learning with augmentation models. CNN with the transfer learning model was chosen for the implementation because of the high accuracy of 97.36%. The model was used to identify GGOs given new examples to classify COVID-19 positive and negative cases accurately.
Keywords: Ground-glass opacities, COVID-19, convolutional neural networks, progression
Table of Contents
Declaration i
Abstract ii
Acknowledgments iii
Dedication iv
List of Figures vii
List of Tables viii
List of Abbreviation ix
CHAPTER ONE: INTRODUCTION
1.1. Background 1
1.2. Problem Statement 1
1.3. Objectives 2
1.3.1. Overall Objective 2
1.3.2. Specific Objectives 2
1.4. Justification 2
1.5. Scope 3
CHAPTER TWO: LITERATURE REVIEW
2.1. Introduction 4
2.2. Overview of CAD systems for Lung Infections 4
2.3. AI-Based CAD systems for Identification of GGOs 4
2.4. Convolutional Neural Network 5
2.4.1. Input Layer 6
2.4.2. Convolution Layers 6
2.4.3. Pooling Layers 6
2.4.4. ReLU Layer 6
2.4.5. Fully Connected Layers 6
2.5. Segmentation of COVID-GGOs 6
2.6. Research Gaps 8
2.7. Proposed Model 8
2.8. Process Model 9
CHAPTER THREE: METHODOLOGY
3.1. Introduction 11
3.2. Methodology Overview 11
3.2.1. Business Understanding 12
3.2.2. Data Understanding 13
3.2.3. Data Preparation 13
3.2.4. Modeling 15
3.2.5. Evaluation 16
3.2.6. Deployment 17
CHAPTER FOUR: RESULTS AND DISCUSSION
4.1. Introduction 19
4.2. Results 19
4.3. Discussion 23
4.4. Achievements 24
4.5. Limitations 25
4.6. Further Work 25
References 26
Appendices 29
Appendix 1: Code Listings 29
List of Figures
Figure 1: CNN Architecture 6
Figure 2: Process Model 10
Figure 3: CRISP-DM Methodology 12
Figure 4: Positive and negative images for GGOs 13
Figure 5: sample of gray scale image 14
Figure 6: Scrum development methodology 16
Figure 7: Model Accuracy 20
Figure 8: Model Loss 21
Figure 9: Model Accuracy 22
Figure 10: Model Loss 23
List of Tables
Table 1: Hardware and Software Requirements 17
Table 2: Confusion Matrix 16
Table 3: Standard CNN 19
Table 4: CNN with transfer learning and data augmentation result 21
List of Abbreviation
GGO: Ground Glass Opacities CT: Computed Tomography
SARS-CoV-2: Acute respiratory syndrome coronavirus 2 CNN: Convolutional Neural Network
COVID-19: Coronavirus Disease 2019 CAD: Computer Aided Diagnosis SVM: Support Vector Machine
CRISP-DM: Cross Industry Standard Process for Data Mining
CHAPTER ONE
INTRODUCTION
1.1. Background
Coronavirus disease 2019 (COVID-19) is caused by acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged in December 2019 in China and has rapidly developed into a global outbreak. As at March 25th 2021, there were 125,454,721 cases, 2,757,158 deaths, and 101,312,279 recoveries across the world (worldometers.info, 2021). In Kenya alone, as at March 25th 2021, there were 124,707 cases, 2066 deaths, and 90770 recoveries (worldometers.info, 2021). COVID-19 is a global threat and is treated as an urgent threat.
Countries across the world have placed efforts and resources in managing the virus through diagnosis, prognosis, and treatment. Nucleic acid detection is currently used as the standard procedure to confirm the virus in the lungs; the strategy is associated with a high false-negative rate. Kenya is fighting the spread of COVID-19 by mass testing, reinforcing compulsory wearing of facemasks and curfews within certain hours to reduce social interactions.
Computed Tomography (CT) images are used in the identification of Ground Glass Opacities (GGOs) that can help in the classification of COVID-19 cases. The Ground Glass Opacities are abnormalities in the lungs that are seen as lighter-colored or gray patches on CT scan, which shows that the lung is sick. Ground-Glass Opacities identification for detection and classification of COVID-19 has been achieved in different ways. Ouyang et al. (2020) used CT scan images to classify COVID-19 and non-COVID-19 cases. However, the research registered false negatives in most cases. Additionally, the study by Wang et al. (2020) used 3D images without feature reduction that faced complexity, low processing speed, and delayed results. Dong et al. (2020) also researched CT imaging in the detection of COVID-19. X-rays have been used in GGOs identification; the research lacked feature identification and could not monitor the progression (Makris, Kontopoulos & Tserpes, 2020).
1.2. Problem Statement
Ground-Glass Opacities come in different shapes, locations, sizes, and quantities and indicate various pathologies such as viral infections, fibrosis, cancers, and chronic lung diseases. Identification of COVID-19 GGOs is difficult due to the numerous pathologies. According to Caruso et al., (2020), the distribution of GGOs for COVID-19 patients is mainly in the lower lobes and periphery; they also appear bilaterally and multifocal, they have round shapes that seem unusual. The numerous characteristics and pathologies associated with the GGOs make it difficult to identify and classify COVID-19 cases accurately. Moreover, various features are considered, such as differences in distribution and their capacities in the lungs; this makes it quite challenging to differentiate COVID-19 cases from other pathogens.
On the other hand, computational problems need improvement to identify GGOs responsible for COVID-19 accurately. The current methods are highly complex and require more computational resources, especially those that consider 3D in identifying the features. It also becomes difficult to extract features that can be used to classify in the lungs that interfere with the results' accuracy. GGOs less than 15% are categorized as cancer, and those more than 15% is classified as mild COVID-19 case and the severity increase with the percentage (Ng et al., 2020). Therefore, it is problematic to accurately identify the percentages of the GGOs to categorize the progression of COVID-19 in the lungs through physical methods.
1.3. Objectives
1.3.1. Overall Objective
To develop a model that classifies Ground-Glass Opacities in the lungs from the CT images samples to monitor COVID-19 progression.
1.3.2. Specific Objectives
i. To extract specific feature characteristics of GGOs representing COVID-19 infection in the lungs.
ii. To classify GGOs representing COVID-19 in the lungs.
iii. To evaluate GGOs representing COVID-19 infection in the lungs for monitoring progression.
1.3.3. Research Questions
i. What are features that can be extracted and used to identify characteristics of GGOs representing COVID-19 in the lungs?
ii. What are the ways available to classify GGOs representing COVID-19 for monitoring progression?
iii. What are the most effective ways available to identify affected lungs by GGOs representing COVID-19 for monitoring progression?
1.4. Justification
Currently, the interpretation of GGOs in CT images is difficult and intensive, given various characteristics and possible pathologies. It is the work of the radiologists and healthcare practitioners to analyze the images for the possible pathologies physically. Since the progression of COVID-19 is rapid, it cost resources in terms of time to investigate several cases.
Additionally, a considerable percentage of error may arise from human analysis, making it necessary to automate the entire process. On the other hand, the current models under tests experience complexities that interfere with processing speed, as most false-negative cases have been reported. A long time is taken to give room for severe cases before they could be treated, thus leading to more deaths. The research proposes a faster and less costly, adaptive, accurate, efficient, and learning-based model which can identify COVID-19 Ground-Glass Opacities in a CT scan images samples. The model automates the result analysis process and thus, limiting human expertise in the analysis. Most importantly, the model identified the GGOs specific stages to determine the severity of the case and propose monitoring approaches.
1.5. Scope
COVID-19 monitoring and progression were based on Ground-Glass Opacity characteristics. The images were obtained from CT scans from different COVID-19 and non- COVID-19 patients with respiratory complications admitted to various hospitals worldwide. The work only focused on those patients admitted in particular treatment options for COVID-19, cancer, pneumonia, and those experiencing COVID-19 like symptoms.
Buyers has the right to create
dispute within seven (7) days of purchase for 100% refund request when
you experience issue with the file received.
Dispute can only be created when
you receive a corrupt file, a wrong file or irregularities in the table of
contents and content of the file you received.
ProjectShelve.com shall either
provide the appropriate file within 48hrs or
send refund excluding your bank transaction charges. Term and
Conditions are applied.
Buyers are expected to confirm
that the material you are paying for is available on our website
ProjectShelve.com and you have selected the right material, you have also gone
through the preliminary pages and it interests you before payment. DO NOT MAKE
BANK PAYMENT IF YOUR TOPIC IS NOT ON THE WEBSITE.
In case of payment for a
material not available on ProjectShelve.com, the management of
ProjectShelve.com has the right to keep your money until you send a topic that
is available on our website within 48 hours.
You cannot change topic after
receiving material of the topic you ordered and paid for.
Login To Comment