請用此 Handle URI 來引用此文件：
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/101417完整後設資料紀錄
| DC 欄位 | 值 | 語言 |
|---|---|---|
| dc.contributor.advisor | 莊曜宇 | zh_TW |
| dc.contributor.advisor | Eric Y. Chuang | en |
| dc.contributor.author | 劉子豪 | zh_TW |
| dc.contributor.author | Tzu-Hao Liu | en |
| dc.date.accessioned | 2026-02-03T16:07:31Z | - |
| dc.date.available | 2026-02-04 | - |
| dc.date.copyright | 2026-02-03 | - |
| dc.date.issued | 2026 | - |
| dc.date.submitted | 2026-01-27 | - |
| dc.identifier.citation | [1] K. A. Glatter and P. Finkelman, "History of the Plague: An Ancient Pandemic for the Age of COVID-19," The American Journal of Medicine, vol. 134, no. 2, 2020, doi: 10.1016/j.amjmed.2020.08.019.
[2] C. E., "WHO Coronavirus Disease (COVID-19) Dashboard," Bangladesh Physiotherapy Journal, vol. 10, no. 1, 2020, doi: 10.46945/bpj.10.1.03.01. [3] J. E. Gagnon, S. B. Kamin, and J. Kearns, "The impact of the COVID-19 pandemic on global GDP growth," Journal of the Japanese and International Economies, vol. 68, 2023, doi: 10.1016/j.jjie.2023.101258. [4] T. N. Palmore. "Infection control measures for prevention of seasonal influenza." UpToDate. https://www.uptodate.com/contents/infection-control-measures-for-prevention-of-seasonal-influenza (accessed Oct. 26, 2025). [5] B. A. Smith. "COVID-19: General approach to infection prevention in the health care setting." https://www.uptodate.com/contents/covid-19-general-approach-to-infection-prevention-in-the-health-care-setting (accessed May 5, 2025). [6] T. Kamalrathne, D. Amaratunga, R. Haigh, and L. Kodituwakku, "Need for effective detection and early warnings for epidemic and pandemic preparedness planning in the context of multi-hazards: Lessons from the COVID-19 pandemic," International Journal of Disaster Risk Reduction, vol. 92, 2023, doi: 10.1016/j.ijdrr.2023.103724. [7] P. J. Rosenthal, "The Importance of Diagnostic Testing during a Viral Pandemic: Early Lessons from Novel Coronavirus Disease (COVID-19)," The American Journal of Tropical Medicine and Hygiene, vol. 102, no. 5, 2020, doi: 10.4269/ajtmh.20-0216. [8] W. WJ and P. HC, "What Is COVID-19?," Journal of the American Medical Association, vol. 324, no. 8, 2020, doi: 10.1001/jama.2020.12984. [9] C. B. Jackson et al., "Mechanisms of SARS-CoV-2 entry into cells," Nature Reviews Molecular Cell Biology, vol. 23, no. 1, 2021, doi: 10.1038/s41580-021-00418-x. [10] G. E. Pfyffer and F. Wittwer, "Incubation Time of Mycobacterial Cultures: How Long Is Long Enough To Issue a Final Negative Report to the Clinician?," Journal of Clinical Microbiology, vol. 50, no. 12, 2012, doi: 10.1128/JCM.02283-12. [11] L. Garibyan and N. Avashia, "Research Techniques Made Simple: Polymerase Chain Reaction (PCR)," The Journal of investigative dermatology, vol. 133, no. 3, 2013, doi: 10.1038/jid.2013.1. [12] M. N. Esbin, O. Whitney, S. Chong, A. C. Maurer, X. Darzacq, and R. Tjian, "Overcoming the bottleneck to widespread testing: a rapid review of nucleic acid testing approaches for COVID-19 detection," RNA: A publication of the RNA Society, vol. 26, no. 7, 2020, doi: 10.1261/rna.076232.120. [13] S. I. Kamel, D. C. Levin, L. Parker, and V. M. Rao, "Utilization trends in noncardiac thoracic imaging, 2002-2014," Journal of the American College of Radiology, vol. 14, no. 3, pp. 337-342, 2017. [14] Y. Zhang, X. Li, W. P. Segars, and E. Samei, "Comparison of patient specific dose metrics between chest radiography, tomosynthesis, and CT for adult patients of wide ranging body habitus," Medical physics, 2014, doi: 10.1118/1.4859315. [15] N. W. Tsai, C. W. Ngai, K. L. Mok, and J. W. Tsung, "Lung ultrasound imaging in avian influenza A (H7N9) respiratory failure," Critical Ultrasound Journal, 2014, doi: 10.1186/2036-7902-6-6. [16] M. A. Chavez et al., "Lung ultrasound for the diagnosis of pneumonia in adults: a systematic review and meta-analysis," Respiratory Research, vol. 15, no. 1, 2014, doi: 10.1186/1465-9921-15-50. [17] N. F. Ramon, M. O. Bravo, G. T. Cortada, J. S. Culleré, M. S. Cabús, and J. M. P. Peruga, "Clinical and ultrasound characteristics in patients with sars-cov-2 pneumonia, associated with hospitalization prognosis. e-covid project," BMC Pulmonary Medicine, vol. 24, 2024, doi: 10.1186/s12890-024-03439-2. [18] L. DA and M. GA, "Relevance of lung ultrasound in the diagnosis of acute respiratory failure: the BLUE protocol," Chest, vol. 134, no. 1, 2008, doi: 10.1378/chest.07-2800. [19] M. Chen et al., "Impact of human and artificial intelligence collaboration on workload reduction in medical image interpretation," npj Digital Medicine, vol. 7, no. 1, 2024, doi: 10.1038/s41746-024-01328-w. [20] Y. Lee et al., "The application of a deep learning system developed to reduce the time for RT-PCR in COVID-19 detection," Scientific Reports, vol. 12, no. 1, 2022, doi: 10.1038/s41598-022-05069-2. [21] I. El Naqa and M. J. Murphy, "What Are Machine and Deep Learning?," Machine and Deep Learning in Oncology, Medical Physics and Radiology, 2022, doi: 10.1007/978-3-030-83047-2_1. [22] W. Morningstar et al., "Augmentations vs Algorithms: What Works in Self-Supervised Learning," arXiv.org, 2024, doi: 10.48550/arXiv.2403.05726. [23] T. R. Network. The Cancer Genome Atlas (TCGA) Research Network Dataset. [Online]. Available: https://www.cancer.gov/tcga [24] S. H. Lee, I. H. Song, and H. J. Jang, "Feasibility of deep learning-based fully automated classification of microsatellite instability in tissue slides of colorectal cancer," International Journal of Cancer, 2021, doi: 10.1002/ijc.33599. [25] "Research on Urine Sediment Images Recognition Based on Deep Learning," IEEE Access, vol. 7, 2019, doi: 10.1109/ACCESS.2019.2953775. [26] L. S.K., S. N. Mohanty, S. K., A. N., and G. Ramirez, "Optimal deep learning model for classification of lung cancer on CT images," Future Generation Computer Systems, vol. 92, 2019, doi: 10.1016/j.future.2018.10.009. [27] M. Constantinou, T. Exarchos, A. G. Vrahatis, and P. Vlamos, "COVID-19 Classification on Chest X-ray Images Using Deep Learning Methods," International Journal of Environmental Research and Public Health, vol. 20, no. 3, 2023, doi: 10.3390/ijerph20032035. [28] S. E. Ebadi et al., "Automated detection of pneumonia in lung ultrasound using deep video classification for COVID-19," Informatics in Medicine Unlocked, vol. 25, 2021, doi: 10.1016/j.imu.2021.100687. [29] V. Sharma, S. K. Gupta, and K. K. Shukla, "Deep learning models for tuberculosis detection and infected region visualization in chest X-ray images," Intelligent Medicine, vol. 4, no. 2, 2024, doi: 10.1016/j.imed.2023.06.001. [30] G. Fontes, L. M. Matos, A. Matta, A. Pilastri, and P. Cortez, "An Empirical Study on Anomaly Detection Algorithms for Extre," 2022, doi: 10.1007/978-3-031-08333-4_7. [31] M. Z. A. Mayaki and M. Riveill, "AnoRand: A Semi Supervised Deep Learning Anomaly Detection Method by Random Labeling," 2023, doi: 10.48550/arXiv.2305.18389. [32] T. Schlegl, P. Seeböck, S. M. Waldstein, U. Schmidt-Erfurth, and G. Langs, "Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery," 2017, doi: 10.48550/arXiv.1703.05921. [33] V. Tiwari et al., "Automatic Generation of Chest X-Ray Medical Imaging Reports using LSTM-CNN," International Conference on Data Science, Machine Learning and Artificial Intelligence, 2021, doi: 10.1145/3484824.3484918. [34] "Adaptive Memory Networks With Self-Supervised Learning for Unsupervised Anomaly Detection," IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 12, 2023, doi: 10.1109/TKDE.2021.3139916. [35] M. E. Tschuchnig and M. Gadermayr, "Anomaly Detection in Medical Imaging - A Mini Review," Data Science – Analytics and Applications, 2022, doi: 10.1007/978-3-658-36295-9_5. [36] N. T et al., "Unsupervised Deep Anomaly Detection in Chest Radiographs," Journal of Digital Imaging, vol. 34, no. 2, 2021, doi: 10.1007/s10278-020-00413-2. [37] M. A et al., "The Clinical Utility of Chest Radiography for Identifying Pneumonia: Accounting for Diagnostic Uncertainty in Radiology Reports," AJR. American journal of roentgenology, vol. 213, no. 6, 2019, doi: 10.2214/AJR.19.21521. [38] D.-E. J et al., "Deep-learning based detection of COVID-19 using lung ultrasound imagery," PLoS One, vol. 16, no. 8, 2021, doi: 10.1371/journal.pone.0255886. [39] M. F. H. Mohamed et al., "Frequency of Abnormalities Detected by Point-of-Care Lung Ultrasound in Symptomatic COVID-19 Patients: Systematic Review and Meta-Analysis," The American Journal of Tropical Medicine and Hygiene, vol. 103, no. 2, 2020, doi: 10.4269/ajtmh.20-0371. [40] J. Born et al., "Accelerating Detection of Lung Pathologies with Explainable Ultrasound Image Analysis," Applied Sciences, vol. 11, no. 2, 2021, doi: 10.3390/app11020672. [41] Z. Yaniv et al., "SimpleITK Image-Analysis Notebooks: a Collaborative Environment for Education and Reproducible Research," Journal of Digital Imaging 2017 31:3, 2017, doi: 10.1007/s10278-017-0037-8. [42] S. Leclerc et al., "Deep Learning for Segmentation Using an Open Large-Scale Dataset in 2D Echocardiography | IEEE Journals & Magazine | IEEE Xplore," IEEE Transactions on Medical Imaging, vol. 38, no. 9, 2019, doi: 10.1109/TMI.2019.2900516. [43] A. Choudhari. "Pcos detection using ultrasound images." https://www.kaggle.com/datasets/anaghachoudhari/pcos-detection-using-ultrasound-images (accessed Mar. 24, 2025). [44] M. H. Yap et al., "Automated Breast Ultrasound Lesions Detection Using Convolutional Neural Networks | IEEE Journals & Magazine | IEEE Xplore," IEEE journal of biomedical and health informatics, 2018, doi: 10.1109/JBHI.2017.2731873. [45] W. Al-Dhabyani, M. Gomaa, H. Khaled, and A. Fahmy, "Dataset of breast ultrasound images," Data in Brief, vol. 28, 2020, doi: 10.1016/j.dib.2019.104863. [46] N. Vallez et al., "BUS-UCLM: Breast ultrasound lesion segmentation dataset," Scientific data, vol. 12, no. 1, 2025, doi: 10.1038/s41597-025-04562-3. [47] A. Turki, A. M. Obaid, H. Bellaaj, M. Ksantini, and A. AlTaee, "UIdataGB: Multi-Class ultrasound images dataset for gallbladder disease detection," Data in Brief, vol. 54, 2024, doi: 10.1016/j.dib.2024.110426. [48] M. Alzubaidi, M. Agus, M. Makhlouf, F. Anver, K. Alyafei, and M. Househ, "Large-scale annotation dataset for fetal head biometry in ultrasound images," Data in Brief, vol. 51, 2023, doi: 10.1016/j.dib.2023.109708. [49] ShengWang1130. "Ultrasound image set of hemangioma3." https://www.kaggle.com/datasets/shengwang1130/ultrasound-image-set-of-hemangioma3 (accessed Mar. 25, 2025). [50] G. Chen et al., "PSFHS: Intrapartum ultrasound image dataset for AI-based segmentation of pubic symphysis and fetal head," Scientific Data 2024 11:1, vol. 11, no. 1, 2024, doi: 10.1038/s41597-024-03266-4. [51] A. Kumar et al., "Creation of an Open-Access Lung Ultrasound Image Database For Deep Learning and Neural Network Applications," medRxiv, 2025, doi: 10.1101/2025.05.09.25327337. [52] J. Yang, C. Li, X. Dai, L. Yuan, and J. Gao, "Focal Modulation Networks," 2022, doi: 10.48550/arXiv.2203.11926. [53] S. Woo et al., "ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders," 2023, doi: 10.48550/arXiv.2301.00808. [54] I. Radosavovic, R. P. Kosaraju, R. Girshick, K. He, and P. Dollár, "Designing Network Design Spaces," 2020/03/30, doi: 10.48550/arXiv.2003.13678. [55] Z. Dai, H. Liu, Q. V. Le, and M. Tan, "CoAtNet: Marrying Convolution and Attention for All Data Sizes," 2021/06/09, doi: 10.48550/arXiv.2106.04803. [56] S. Mehta and M. Rastegari, "MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer," 2021, doi: 10.48550/arXiv.2110.02178. [57] A. F. Agarap, "Deep Learning using Rectified Linear Units (ReLU)," 2018, doi: 10.48550/arXiv.1803.08375. [58] S. Elfwing, E. Uchibe, and K. Doya, "Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learning," 2017, doi: 10.48550/arXiv.1702.03118. [59] K. He, X. Chen, S. Xie, Y. Li, P. Dollár, and R. Girshick, "Masked Autoencoders Are Scalable Vision Learners," 2021, doi: 10.48550/arXiv.2111.06377. [60] "Multiscale structural similarity for image quality assessment," doi: 10.1109/ACSSC.2003.1292216. [61] W. J. Youden, "Index for rating diagnostic tests," Cancer, 1950, doi: 10.1002/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3. [62] "Majority Systems and the Condorcet Jury Theorem," The Statistician, vol. 38, no. 3, 1989, doi: 10.2307/2348873. [63] "New Bounds on the Accuracy of Majority Voting for Multiclass Classification," IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 4, 2025, doi: 10.1109/TNNLS.2024.3387544. [64] K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," 2015/12/10, doi: 10.48550/arXiv.1512.03385. [65] K. Simonyan and A. Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition," 2014, doi: 10.48550/arXiv.1409.1556. [66] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, "Rethinking the Inception Architecture for Computer Vision," 2015, doi: 10.48550/arXiv.1512.00567. [67] N. Awasthi, A. Dayal, L. R. Cenkeramaddi, and P. K. Yalavarthy, "Mini-COVIDNet: Efficient Lightweight Deep Neural Network for Ultrasound Based Point-of-Care Detection of COVID-19," Ieee Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 68, no. 6, 2021, doi: 10.1109/TUFFC.2021.3068190. [68] I. Loshchilov and F. Hutter, "Decoupled Weight Decay Regularization," 2017, doi: 10.48550/arXiv.1711.05101. [69] "PyTorch | Proceedings of the 33rd International Conference on Neural Information Processing Systems," doi: 10.5555/3454287.3455008. [70] C. R. Harris et al., "Array programming with NumPy," Nature 2020 585:7825, vol. 585, no. 7825, 2020, doi: 10.1038/s41586-020-2649-2. [71] G. Bradski, "The OpenCV Library," Dr. Dobb's Journal of Software Tools, 2000. [72] W. McKinney, "Data Structures for Statistical Computing in Python - SciPy Proceedings," scipy, 2010, doi: 10.25080/Majora-92bf1922-00a. [73] PedregosaFabian et al., "Scikit-learn: Machine Learning in Python," The Journal of Machine Learning Research, 2011, doi: 10.5555/1953048.2078195. [74] P. Virtanen et al., "SciPy 1.0: fundamental algorithms for scientific computing in Python," Nature Methods 2020 17:3, vol. 17, no. 3, 2020, doi: 10.1038/s41592-019-0686-2. [75] F. Sadik, A. G. Dastider, S. A. Fattah, F. Sadik, A. G. Dastider, and S. A. Fattah, "SpecMEn-DL: spectral mask enhancement with deep learning models to predict COVID-19 from lung ultrasound videos," Health Information Science and Systems 2021 9:1, vol. 9, no. 1, 2021, doi: 10.1007/s13755-021-00154-8. [76] J. Born et al., "Accelerating Detection of Lung Pathologies with Explainable Ultrasound Image Analysis," Applied Sciences, 2021, doi: 10.3390/app11020672. [77] M. Karar, M. Shouman, C. Chalopin, M. Karar, M. Shouman, and C. Chalopin, "Adversarial Neural Network Classifiers for COVID-19 Diagnosis in Ultrasound Images," Computers, Materials & Continua, vol. 70, no. 1, 2021, doi: 10.32604/cmc.2022.018564. [78] G. Madhu et al., "XCovNet: An optimized xception convolutional neural network for classification of COVID-19 from point-of-care lung ultrasound images," Multimedia Tools and Applications 2023 83:11, vol. 83, no. 11, 2023, doi: 10.1007/s11042-023-16944-z. [79] "Lung Ultrasound Image Classification Using Deep Learning and Histogram of Oriented Gradients Features for COVID-19 Detection," 2023 IEEE Sensors Applications Symposium (SAS), 2023, doi: 10.1109/SAS58821.2023.10254002. [80] D. Chen, D. Wang, T. Darrell, and S. Ebrahimi, "Contrastive Test-Time Adaptation," 2022, doi: 10.48550/arXiv.2204.10377. [81] X. Huang and S. Belongie, "Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization," 2017, doi: 10.48550/arXiv.1703.06868. | - |
| dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/101417 | - |
| dc.description.abstract | 在新興呼吸道傳染病爆發的關鍵初期，臨床應對常受限於確診病例稀缺與影像標註成本高昂，致使傳統深度學習模型陷入數據不足的冷啟動(Cold-start)困境。肺部超音波(Lung ultrasound)憑藉其高可攜性、無游離輻射及即時床邊檢測之優勢，成為理想的第一線篩檢工具，惟其高度依賴操作者經驗及顯著的影像變異性，為自動化分析帶來巨大挑戰。為解決上述限制，本研究提出一套創新的雙路徑分析架構，旨在有限數據條件下提供穩定的分析工具。核心模型採用FocalNet自動編碼器(Autoencoder)作為骨幹網路，並引入遮罩自動編碼器(Mask Autoencoder)訓練策略與混合損失函數，於超過30,000張無標註醫療超音波影像上進行預訓練，以學習具備高度結構判別力的穩健特徵表示。在此架構下，本研究整合了兩大技術流程，首先是無監督異常偵測流程，該流程利用滑動視窗計算區域重建誤差，並結合統計閾值機制，在完全無須病理標註的情況下即可識別異常影像，於COVID-19數據集上達到了0.770的平衡準確率。其次，針對臨床隨後逐漸取得少量確診病例的情境，本研究設計了少樣本學習(few-shot learning)流程，利用預訓練編碼器作為強效特徵提取器，僅需微調單層線性分類器即可快速適應特定疾病。實驗數據顯示，僅需3%的極少量標註數據，模型即可達成高達0.923的平衡準確率，展現優異的數據效益與臨床實用性。總結而言，本研究建立了一套具可擴展性的工程解決方案，實現了從無監督偵測到少樣本監督學習的漸進式過渡，有效克服了新興傳染病初期的數據稀缺難題，為未來面對大流行病時的快速反應與輔助篩檢，提供了一項具備高度臨床價值的技術架構。 | zh_TW |
| dc.description.abstract | During the critical early stages of emerging respiratory infectious disease outbreaks, the clinical response is often hampered by a scarcity of confirmed cases and the high costs associated with annotation. This makes conventional supervised deep learning models less effective due to the "cold-start" dilemma. Lung ultrasound (LUS) is an ideal first-line screening tool given its portability, absence of ionizing radiation, and real-time bedside capability. However, its effective deployment is hindered by significant operator dependency and considerable image variability. To address these challenges, this study introduces a novel Dual-Pipeline Framework based on a FocalNet Autoencoder (FocalNet-AE) that integrates unsupervised anomaly detection with few-shot supervised classification capabilities. The core model was pre-trained on over 30,000 unlabeled medical ultrasound images using masked autoencoding combined with a hybrid loss function to learn robust structural representations. Within this framework, the unsupervised pipeline employs sliding-window reconstruction error analysis to identify anomalies without prior knowledge of pathology, achieving a balanced accuracy of 0.770 in zero-shot scenarios. Additionally, the supervised pipeline utilizes the pre-trained encoder to swiftly adapt to specific diseases with minimal labels, achieving a balanced accuracy of 0.923 using only 3% of the training data. This study presents a data-efficient and scalable engineering solution that facilitates a smooth transition from unsupervised anomaly detection to few-shot classification, providing a clinically applicable tool for rapid response during future pandemics. | en |
| dc.description.provenance | Submitted by admin ntu (admin@lib.ntu.edu.tw) on 2026-02-03T16:07:31Z No. of bitstreams: 0 | en |
| dc.description.provenance | Made available in DSpace on 2026-02-03T16:07:31Z (GMT). No. of bitstreams: 0 | en |
| dc.description.tableofcontents | 口試委員審定書 i
致謝 ii 摘要 iii Abstract iv Contents v List of Abbreviations ix List of Figures xi List of Tables xii Chapter 1 Introduction 1 1.1 Emerging Respiratory Infectious Diseases and Global Health Security 1 1.2 Transmission Dynamics and the Imperative for Early Screening 1 1.3 Pathophysiology and Clinical Manifestations of COVID-19 2 1.4 Clinical Screening Tools for Respiratory Diseases 3 1.4.1 Culture and identification 3 1.4.2 NAT-based Method 4 1.4.3 Antigen Rapid Test 5 1.4.4 Chest X-ray and X-ray Computed Tomography 6 1.4.5 Ultrasonography 7 1.5 Integrating Deep Learning into Screening Tools 8 1.5.1 Introduction of Deep Learning 9 1.5.2 Self-Supervised Learning 10 1.5.3 Deep Learning in Medical Imaging 11 1.6 Anomaly Detection 12 1.6.1 Overview of Anomaly Detection 12 1.6.2 Unsupervised Learning for Anomaly Detection 14 1.6.3 Anomaly Detection in Medical Imaging: Current Status and Limitations 15 1.7 Motivation 16 1.8 Specific Aim 16 Chapter 2 Materials and Methods 18 2.1 Overview of Proposed Workflow 18 2.2 Data Preprocessing 20 2.2.1 Datasets for Unsupervised Learning Training 20 2.2.2 Datasets for Classification Tasks with Supervised and Unsupervised Models 21 2.2.3 Data Cleaning and Preprocessing 23 2.2.4 Image Augmentation 25 2.2.5 Dataset Allocation Strategy for Training 25 2.3 Unsupervised Training Pipeline for Autoencoder 26 2.3.1 Encoder Architectures for Unsupervised and Supervised Learning 26 2.3.2 Model Architecture of Autoencoders 28 2.3.3 Proposed Model with Advanced Encoder Backbone 29 2.3.4 Masked-Autoencoder Training Strategy 30 2.3.5 Hybrid Loss Function 31 2.3.6 Sliding Window Loss Evaluation for Validation and Testing 33 2.3.7 Weighted Training Loss 34 2.3.8 Early Stopping 35 2.3.9 Threshold Determination for Unsupervised Anomaly Detection 36 2.3.10 Aggregating Frame-Level Predictions to Patient Level 38 2.3.11 Baseline Models Used in Evaluation 39 2.4 Supervised Learning Pipeline 39 2.4.1 Fine-tune Supervised Classification Head with Minimized Abnormal Data. 39 2.4.2 Evaluation Strategy under Limited Data Scenarios 40 2.5 Evaluation Metrics 40 2.6 Implementation Details 42 Chapter 3 Results 44 3.1 Dataset Characteristics and Reconstruction Results 45 3.1.1 Comparison Between Internal and Holdout Datasets under Cross-Domain Ultrasound Variability 45 3.1.2 Reconstruction Results of the Proposed Model 46 3.2 Autoencoder Fine-Tuning Strategies 48 3.2.1 Comparison of FocalNet-LRF and FocalNet-SRF Backbones 49 3.2.2 Impact of MAE Training and Loss Calculations on Model Performance 49 3.2.3 Impact of Different Loss Functions on Model Performance 51 3.2.4 Impact of Target Dataset Reweighting on Model Performance 52 3.3 Results of Downstream Unsupervised Tasks 53 3.3.1 Impact of Masking in Downstream Unsupervised Tasks 54 3.3.2 Impact of Sliding-Window-Based Loss in Downstream Unsupervised Tasks 54 3.3.3 Impact of Dynamic Thresholding in Downstream Unsupervised Tasks 55 3.3.4 Impact of Sliding Window Size in Downstream Unsupervised Tasks 56 3.3.5 Impact of Gamma Adjustment in Downstream Unsupervised Tasks 57 3.3.6 Comparison of the Proposed Autoencoder with Other Backbone Architectures 59 3.3.7 Holdout Test Results on Unsupervised Tasks 60 3.4 Results of Downstream Supervised Tasks 61 3.4.1 Impact of MAE Training and Loss Calculations on Supervised Tasks 61 3.4.2 Impact of Different Loss Functions on Supervised Tasks 62 3.4.3 Impact of Target Dataset Reweighting on Supervised Tasks 63 3.4.4 Comparison of FocalNet-LRF and FocalNet-SRF Backbones on Supervised Tasks 64 3.4.5 Comparison with Other Common Supervised Models 65 3.4.6 Independent Test Results on Supervised Tasks 65 3.4.7 Ablation Study of Fine-Tuned and ImageNet Pretrained Weights 66 Chapter 4 Discussion 68 4.1 Characteristics of the Proposed FocalNet-AE and pipelines 68 4.2 Design Considerations for Autoencoder Fine-Tuning 70 4.2.1 Variants of FocalNet 70 4.2.2 Importance of MAE in the Proposed Pipeline 71 4.2.3 Risk of Catastrophic Forgetting During Fine-Tuning 72 4.2.4 Determination of the Loss Functions 73 4.3 Design Choices and Technical Considerations in the Unsupervised Pipeline 74 4.4 Comparison Between the Existing Models 75 4.5 Clinical Applicability and Potential of the Proposed Model 78 4.6 Limitations 80 4.7 Future Work 81 Chapter 5 Conclusions 82 Chapter 6 References 84 | - |
| dc.language.iso | en | - |
| dc.subject | 肺部超音波 | - |
| dc.subject | 無監督異常偵測 | - |
| dc.subject | 少樣本學習 | - |
| dc.subject | FocalNet | - |
| dc.subject | 遮罩自動編碼器 | - |
| dc.subject | Lung Ultrasound | - |
| dc.subject | Unsupervised Anomaly Detection | - |
| dc.subject | Few-Shot Learning | - |
| dc.subject | FocalNet | - |
| dc.subject | Masked Autoencoder | - |
| dc.title | 針對新興流行病的雙路徑肺部超音波分析：基於FocalNet-AE的無監督異常偵測與少樣本分類 | zh_TW |
| dc.title | Dual-Pipeline Lung Ultrasound Analysis for Emerging Pandemics: Unsupervised Anomaly Detection and Few-Shot Classification via FocalNet-Autoencoder | en |
| dc.type | Thesis | - |
| dc.date.schoolyear | 114-1 | - |
| dc.description.degree | 碩士 | - |
| dc.contributor.coadvisor | 陳翔瀚 | zh_TW |
| dc.contributor.coadvisor | Hsiang-Han Chen | en |
| dc.contributor.oralexamcommittee | 賴亮全;陳佩君 | zh_TW |
| dc.contributor.oralexamcommittee | Liang-Chuan Lai;Pei-Chun Chen | en |
| dc.subject.keyword | 肺部超音波,無監督異常偵測少樣本學習FocalNet遮罩自動編碼器 | zh_TW |
| dc.subject.keyword | Lung Ultrasound,Unsupervised Anomaly DetectionFew-Shot LearningFocalNetMasked Autoencoder | en |
| dc.relation.page | 93 | - |
| dc.identifier.doi | 10.6342/NTU202600268 | - |
| dc.rights.note | 同意授權(限校園內公開) | - |
| dc.date.accepted | 2026-01-28 | - |
| dc.contributor.author-college | 電機資訊學院 | - |
| dc.contributor.author-dept | 生醫電子與資訊學研究所 | - |
| dc.date.embargo-lift | 2031-01-27 | - |
| 顯示於系所單位： | 生醫電子與資訊學研究所 | |
文件中的檔案：
| 檔案 | 大小 | 格式 | |
|---|---|---|---|
| ntu-114-1.pdf 未授權公開取用 | 2.77 MB | Adobe PDF | 檢視/開啟 |
系統中的文件，除了特別指名其著作權條款之外，均受到著作權保護，並且保留所有的權利。