臺灣博碩士論文加值系統

組織切片檢查是臨床病理診斷上的檢查方式之一，主要是藉由判斷病變的部位、範圍、性質和腫瘤的良惡性，作出準確的病理診斷並確定治療方案。因此，組織切片的完整性影響病理診斷的精確性。
以往，判斷組織切片完整性的方式分為裸視初步觀察及使用顯微鏡兩種，裸視初步觀察每一個組織切片會花費3至5秒的時間，但會造成分辨上不夠精確。使用顯微鏡判斷雖更加精準，但每一片的觀察時間也要10秒左右。因此，目前仍然缺少一個有效且快速為組織切片分辨完整性的方法。
本論文提供以深度學習方式分類組織切片及組織染色瑕疵辨識，以減少人工判別之時間與提高組織切片完整性之準確度。本論文運用YOLOv5演算法進行分類判斷及驗證準確性，經過實驗評估後，得到YOLOv5演算法檢測準確率為83.8%以及召回率為90.7%，比起裸視初步觀察與使用顯微鏡之方式，相較之下YOLOv5演算法更有能力分辨組織切片之完整性。

Biopsy is one of the methods of clinical pathological diagnosis, mainly by judging the diseased region, scope, nature and benign or malignant of the tumor to make an accurate pathological diagnosis and determine the treatment plan. Therefore, the integrity of tissue sections affects the accuracy of pathological diagnosis.
In the past, the methods of judging the integrity of tissue sections were divided into two types: naked-vision preliminary observation and using a microscope. The naked-vision preliminary observation of each tissue section would take 3 to 5 seconds, but the resolution would not be accurate enough. Although it is more accurate to judge using a microscope, the observation time of each piece is about 10 seconds. Therefore, there is still a lack of an efficient and rapid method to distinguish the integrity of tissue sections.
This study provides a deep learning method to classify tissue slice images and Pathological dyeing slide Defects Inspection to reduce manual discrimination time and improve the accuracy of tissue slice integrity.This Study uses Yolo v5 as classification judgment and verification accuracy. After experimental evaluation, the detection accuracy of the YOLOv5 Accuracy is 83.8% and the Recall rate is 90.7%.The YOLOv5 algorithm is more capable of distinguishing the integrity of tissue sections.

摘 要 i
ABSTRACT ii
誌 謝 iii
目錄 iv
表目錄 v
圖目錄 vi
壹、 緒論 1
一、 研究動機 1
二、 研究目標 1
三、 研究範圍與限制 1
貳、 文獻探討 2
一、 病理學及組織染色技術介紹 2
(一) 病理組織切片及組織染色技術 2
二、 機器學習演算法 6
三、 YOLO演算法 6
參、 研究方法 9
一、 YOLO玻片偵測 10
二、 染色片影像的分類及切片瑕疵判定 13
三、 染色切片資料集YOLOv5訓練及檢測 16
肆、 實驗結果及分析 18
一、 模型評估 18
二、 實驗環境 19
三、 實驗資料 19
四、 YOLOv5 模型訓練實驗結果 21
伍、 結論 24
參考文獻 25

1.徐祥恩, 使用臨床資料計算腹主動脈鈣化分數及鈣化總長度, in 資訊工程學系. 2021, 亞洲大學: 台中市. p. 53.
2.施祈安, 應用機器視覺與深度學習於鋼珠表面之瑕疵檢測, in 工業工程與管理系. 2021, 國立臺北科技大學: 台北市. p. 60.
3.劉, 振., 組織病理染色技術與圖譜 : 組織化學染色 = An Atlas and manual of histopathological staining methods-histochemistry / 劉振軒等編著. 初版 ed. Atlas and manual of histopathological staining methods : histochemistry. 1996, 苗栗縣竹南: 台灣養豬科學硏究所.
4.組織病理相關-石蠟切片製作流程及相關儀器. Available from: https://smedpath.kmu.edu.tw/index.php/%E5%84%80%E5%99%A8%E8%A8%AD%E5%82%99/%E7%B5%84%E7%B9%94%E7%97%85%E7%90%86%E7%9B%B8%E9%97%9C.
5.Wang, C.-Y., et al. CSPNet: A new backbone that can enhance learning capability of CNN. in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. 2020.
6.He, K., et al. Deep residual learning for image recognition. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
7.Girshick, R. Fast r-cnn. in Proceedings of the IEEE international conference on computer vision. 2015.
8.Ren, S., et al., Faster r-cnn: Towards real-time object detection with region proposal networks. Advances in neural information processing systems, 2015. 28.
9.Lin, T.-Y., et al. Feature pyramid networks for object detection. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
10.Hinton, G.E. and R.R. Salakhutdinov, Reducing the dimensionality of data with neural networks. science, 2006. 313(5786): p. 504-507.
11.Girshick, R., et al. Rich feature hierarchies for accurate object detection and semantic segmentation. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2014.
12.Redmon, J., et al. You only look once: Unified, real-time object detection. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
13.Yolo：基於深度學習的物件偵測 (含YoloV3). Available from: https://mropengate.blogspot.com/2018/06/yolo-yolov3.html.
14.Redmon, J. and A. Farhadi, Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767, 2018.
15.Bochkovskiy, A., C.-Y. Wang, and H.-Y.M. Liao, Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934, 2020.
16.Ultralytics. Yolov5. Available from: https://github.com/ultralytics/yolov5.