PUKYONG

GUIDE OAK 국가리포지터리 부경대학교

검색

PUKYONG Repository 대학원 인공지능융합학과

Metadata Downloads

Abstract: Early identification of COVID-19 may aid in both the planning of a prompt medical response and the slowing of the deadly disease's fast spread. One of the new, innovative, and safe methods to replace RT-PCR in the early screening of COVID-19 is to diagnose COVID-19 using medical imaging modalities. Recently, a lot of research has been done to analyze medical imaging for the early diagnosis of disease due to the improvement of medical imaging technology as well as the success of deep learning applied for vision tasks. The use of deep learning based on chest X-ray and CT scan to identify COVID-19 patients and severity grading is the subject of this thesis. We first offer a thorough explanation of six cutting-edge deep learning techniques before training these models on the specialized Chest X-ray and CT Scan picture dataset. The dataset comprises of COVID-19 cases, Pneumonia patients, and NORMAL instances to analyze how well artificial intelligence systems perform at detecting COVID-19 and determining its severity. The outcomes have demonstrated to us that deep learning may assist radiologists as a thorough, stable, and reliable strategy for early diagnosis of COVID-19.

URI: https://repository.pknu.ac.kr:8443/handle/2021.oak/32909
http://pknu.dcollection.net/common/orgView/200000665485

Table Of Contents: I. Introduction 1
1.1. Coronavirus pandemic (COVID-19) 1
1.2. Deep Learning for Medical Images Analysis 2
1.3. Outline of Thesis 4
II. Related Works 5
2.1. COVID-19 Detection Using Chest X-Ray 5
2.2. COVID-19 Detection Using CT-Scan 6
2.3. COVID-19 Severity Grading Using Chest X-Ray 7
III. The Proposed Methods and Results 10
3.1. Dataset 10
3.1.1. COVID-19 Detection Chest X-Ray Dataset 10
3.1.2. COVID-19 Detection CT Scan Dataset 11
3.1.3. COVID-19 Severity Grading Dataset 14
3.2. Deep Learning Architectures 15
3.2.1. DenseNet121 16
3.2.2. ResNet50 17
3.2.3. InceptionNet 18
3.2.4. Swin Transformer 19
3.2.5. EfficientNet 20
3.2.6. Hybrid EfficientNet and DOLG 22
3.3. Preprocessing and Training Setting 23
3.3.1. Data Augmentation 23
3.3.2. Hardware and Hyperparameter Setting 24
3.4 Evaluation Metrics 25
3.4.1. Precision Metric 26
3.4.2. Recall Metric 26
3.4.3. F1-score Metric 26
3.4.4. Macro-Average Metric 27
3.4.5. Macro-Average Metric 27
3.5. Results and Discussion 28
3.5.1. COVID-19 Chest X-Ray Classification Results 28
3.5.2. COVID-19 CT Scan Classification Results 34
3.5.3. COVID-19 Chest X-Ray Severity Grading Results 37
3.5.4. Limitations of this study 42
IV. Conclusion 45
References 46
Acknowledgement 52

Authorize & License

qrcode