臺灣博碩士論文加值系統

在臨床應用上，白血球細胞除了常見的5大種類，還會進一步依據細胞的分化程度細分為13種類別。
本篇論文將針對即時串流的白血球細胞顯微影像進行分類，我們提出一種階層式深度模型，將13種白血球類別透過層級化分類使其更加精準捕捉各個類別之間的特徵差異。白血球細胞影像經常會因為不同染劑的差異以及人員操作顯微鏡的習慣而影響呈現效果，為了排除影像中染色情形和光線強弱等干擾因素，我們運用物件偵測、細胞分割和色彩正規化等技術對白血球細胞影像進行前處理，且為了避免特定白血球類別損失明顯的色彩特徵，我們調整色彩正規化的順序以達到最好的效果。
我們使用 ResNet和DenseNet 兩種深度學習網路模型做為分類結果的效能比較，結果顯示，我們提出的方法結合ResNet18，在第一階段對於常見5大類別的平均準確率達98.01%，且針對13種類別的整體平均準確率為84.89%。

White blood cells (WBCs) are not only classified into the common five main types but are further subdivided into thirteen categories based on their degree of differentiation in clinical application.
This thesis focuses on classifying real-time streaming microscope images of WBCs. We propose hierarchical deep models using layered classification to more accurately differentiate among thirteen categories. WBC image presentations are often influenced by variations in stains from different components and microscope operation habits. To eliminate interfering factors such as variations in lighting intensity and staining in the images, preprocessing techniques such as object detection, cell segmentation, and color normalization are applied to WBC images. Moreover, we optimize color normalization sequencing to retain distinct color characteristics for specific WBC categories and minimize loss.
We compare the performance of the classification results using two deep learning network models, ResNet and DenseNet. Results indicate that the proposed method with ResNet18 achieves an average accuracy of 98.01% for the five main types in the first stage, and an overall average accuracy of 84.89% across all thirteen categories.

摘要
ABSTRACT
致謝
目錄
圖目錄
表目錄
第一章 緒論
1.1 研究動機
1.2 論文架構
第二章 文獻探討
2.1 血液檢驗之白血球細胞鑑識
2.1.1 白血球細胞分類
2.1.2 血液抹片製作及細胞染色
2.2 白血球影像辨識相關研究
2.3 物件偵測
2.4 影像分割
2.4.1 U-Net
2.4.2 UNeXt
2.5 影像色彩正規化
2.6 物件辨識之深度網路模型
2.6.1 ResNet
2.6.2 DenseNet
第三章 研究方法
3.1 系統流程
3.1.1 系統流程介紹
3.1.2 訓練資料集
3.2 影像前處理
3.2.1 物件偵測(Object Detection)
3.2.2 影像分割(Image Segmentation)
3.2.3 影像色彩正規化(Color-Normalization)
3.3 階層式深度模型網路架構
3.4 系統介紹
第四章 實驗結果
4.1 實驗環境
4.2 評估方法
4.3 實驗設定
4.4 實驗結果
第五章 結論探討
參考文獻

[1]O'Shea, K., & Nash, R., “An introduction to convolutional neural networks,” arXiv:1511.08458, 2015.
[2]Walker, H. K., Hall, W. D., & Hurst, J. W., “Clinical methods: the history, physical, and laboratory examinations,” 3rd ed. Boston: Butterworths, 1990.
[3]Hoffman, R., Benz Jr, E. J., Silberstein, L. E., Heslop, H., Anastasi, J., & Weitz, J., “Hematology: basic principles and practice,” Elsevier Health Sciences, 2013.
[4]Koepke, J.A., Bentley, S., Pierre, R.V., Richardson-Jones, A., Ross, D., Saunders, A.M., Simson, E., Assendelft, O.W., & Wilding, P., “Reference Leukocyte Differential Count (Proportional) and Evaluation of Instrumental Methods; Approved Standard,” 1992.
[5]Bain, B. J., “Diagnosis from the blood smear,” NEJM, 353(5), 498-507, 2005.
[6]Sullivan, B. M., & Locksley, R. M., “Basophils: a nonredundant contributor to host immunity,” j.immuni, 30(1), 12-20, 2009.
[7]Arinobu, Y., Iwasaki, H., & Akashi, K., “Origin of basophils and mast cells,” allergol int, 58(1), 21-28, 2009.
[8]Fahy, J. V., “Eosinophilic and neutrophilic inflammation in asthma: insights from clinical studies,” PATS, 6(3), 256-259, 2009.
[9]Gauvreau, G. M., Ellis, A. K., & Denburg, J. A., “Haemopoietic processes in allergic disease: eosinophil/basophil development,” Clin Exp Allergy, 39(9), 1297-1306, 2009.
[10]J Corrons, J. L. V., Albarède, S., Flandrin, G., Heller, S., Horvath, K., Houwen, B., et al, “Guidelines for blood smear preparation and staining procedure for setting up an external quality assessment scheme for blood smear interpretation. Part I: control material,” CCLM, 42(8), 922-926, 2004.
[11]SHEN, P. F., & PATTERSON, L. T., “A simplified Wright's stain technique for routine avian blood smear staining,” Poult. Sci., 62(5), 923-924, 1983.
[12]Giemsa, G., “Färbemethoden für malariaparasiten. Zentralbl Bakteriol,” 31, 429-430, 1902.
[13]Das, P. K., & Meher, S., “An efficient deep convolutional neural network based detection and classification of acute lymphoblastic leukemia,” Expert Syst. Appl., 183, 115311, 2021.
[14]Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., et al, “Mobilenets: Efficient convolutional neural networks for mobile vision applications,” arXiv:1704.04861, 2017.
[15]e, K., Zhang, X., Ren, S., & Sun, J., “Deep residual learning for image recognition,” CVPR, pp. 770-778, 2015.
[16]Alharbi, A. H., Aravinda, C. V., Lin, M., Venugopala, P. S., Reddicherla, P., & Shah, M. A., “Segmentation and classification of white blood cells using the UNet,” CMMI, 2022, 2022.
[17]Kaushik, H., Singh, D., Kaur, M., Alshazly, H., Zaguia, A., & Hamam, H., “Diabetic retinopathy diagnosis from fundus images using stacked generalization of deep models,” IEEE Access, 9, 108276-108292, 2021.
[18]Hahn, T. V., & Mechefske, C. K., “Self-supervised learning for tool wear monitoring with a disentangled-variational-autoencoder,” IJHM, 4(1), 69-98, 2021.
[19]Bozorgpour, A., Azad, R., Showkatian, E., & Sulaiman, A., “Multi-scale regional attention deeplab3+: Multiple myeloma plasma cells segmentation in microscopic images,” arXiv:2105.06238, 2021.
[20]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A., “You only look once: Unified, real-time object detection,” CVPR, pp. 779-788, 2016.
[21]Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., et al, “Going Deeper with Convolutions,” CVPR, pp. 1-9, 2015.
[22]Wang, C. Y., Bochkovskiy, A., & Liao, H. Y. M., “YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,” CVPR, pp. 7464-7475, 2023.
[23]Ronneberger, O., Fischer, P., & Brox, T., “U-net: Convolutional networks for biomedical image segmentation,” arXiv:1505.04597, 2015.
[24]Valanarasu, J. M. J., & Patel, V. M., “UNeXt: MLP-based Rapid Medical Image Segmentation Network,” arXiv:2203.04967, 2022.
[25]Vahadane, A., Peng, T., Sethi, A., Albarqouni, S., Wang, L., Baust, M., et al, “Structure-preserving color normalization and sparse stain separation for histological images,” IEEE, 35(8), 1962-1971, 2016.
[26]LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P., “Gradient-based learning applied to document recognition,” IEEE, 86(11), 2278-2324, 1998.
[27]Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q., “Densely connected convolutional networks,” CVPR, pp. 4700-4708, 2017.
[28]https://www.differencebetween.com/difference-between-giemsa-stain-and-vs-wright-stain
[29]https://www.kaggle.com/datasets/bzhbzh35/peripheral-blood-cell
[30]https://www.kaggle.com/datasets/livepriyanka09/aml-cytology
[31]https://github.com/heartexlabs/labelImg
[32]https://github.com/wkentaro/labelme