跳到主要內容

臺灣博碩士論文加值系統

(44.200.194.255) 您好!臺灣時間:2024/07/23 15:02
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:廖本傑
研究生(外文):Pen-Chieh Liao
論文名稱:以 CT 為基礎分割腹主動脈瘤重建三維模型之研究
論文名稱(外文):Study of CT-Based Segmentation of Abdominal Aortic Aneurysm and Reconstruction of 3D Model
指導教授:詹永寬詹永寬引用關係
指導教授(外文):Yung-Kuan Chan
口試委員:王圳木喻石生
口試委員(外文):Chuin-Mu WangShyr-Shen Yu
口試日期:2023-12-28
學位類別:碩士
校院名稱:國立中興大學
系所名稱:資訊管理學系所
學門:電算機學門
學類:電算機一般學類
論文種類:學術論文
論文出版年:2024
畢業學年度:112
語文別:中文
論文頁數:50
中文關鍵詞:電腦斷層掃描腹主動脈瘤影像識別影像分割3D 建模
外文關鍵詞:Computed TomographyAbdominal Aortic AneurysmImage RecognitionImage Segmentation3D Modeling
相關次數:
  • 被引用被引用:0
  • 點閱點閱:25
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
本研究旨在根據電腦斷層掃描(CT)切片開發一套自動影像處理和分析並建立
三維模型的方法,以提升腹主動脈瘤治療的效能。腹主動脈瘤是一種主動脈壁的
異常擴張,若未及時治療可能導致嚴重後果。為了更準確地定位和識別患處,本
研究設計了一套完整的醫學影像處理流程。
首先,在 CT 影像前處理階段,從 DICOM 格式中提取 CT 切片,並進行線性轉
換以生成 BMP 格式的影像。透過對比度調整、Otsu's 二值化的應用,實現了腹部
血管結構的清晰呈現。
為了進一步篩選出與脊椎相關的切片,採用 YOLOv4 模型進行物件識別,從
而有效地定位患處區域,其模型的 Precision 為 93%、Recall 為 95%、F1-score
為 93.9%,為後續的器官定位打下堅實基礎。
其次,本研究引入影像識別步驟進行腎臟定位。通過去除脊椎上下延伸區
域,減少了不必要的計算。使用 YOLOv4 和 YOLOv7 模型,成功識別並篩選出具有
腎臟的切片,用於進一步篩選出主動脈與腎動脈連接處的切片影像。其模型的各
項指標,YOLOv4 於左腎的 Precision 為 97%、Recall 為 96%、F1-score 為
96.9%,右腎的 Precision 為 90%、Recall 為 94%、F1-score 為 91.9%;YOLOv7
於左腎的 Precision 為 91%、Recall 為 99%、F1-score 為 94.8%,右腎的
Precision 為 88%、Recall 為 100%、F1-score 為 93.6%。
接下來,本研究進行了血管輪廓的影像分割,採用 U-net 模型對血管輪廓進
行分割。通過訓練模型,其模型的各項指標,分割主動脈的 Precision 為
96.7%、Recall 為 96.2%、F1-score 為 96.4%、Accuracy 為 95.3%;分割主動脈
與腎動脈連接處的 Precision 為 96.5%、Recall 為 89.8%、F1-score 為 93%、
Accuracy 為 87.2%。
最終,透過前處理和建立體素點,本研究以三角網格建立了 STL 格式的血管
輪廓 3D 模型,供醫療團隊於手術前進行評估。本研究的方法不僅有望提高腹主
動脈瘤治療的精確性和效率,還為醫學影像處理領域的進一步研究提供了有益的
參考。通過綜合應用各種影像處理技術和深度學習模型,為腹主動脈瘤的診斷和
治療提供了一種全新的、更為精準的方法。
This study aims to develop an automated image processing and analysis
method based on computed tomography (CT) slices to enhance the efficiency
of abdominal aortic aneurysm (AAA) treatment. AAA is an abnormal dilation
of the aortic wall, and if not treated promptly, it can lead to severe
consequences. In order to accurately locate and identify the affected
area, this study designs a comprehensive medical image processing
workflow.
Firstly, in the CT image preprocessing stage, CT slices are extracted
from DICOM format and undergo linear transformation to generate BMPformat images. The application of contrast adjustment and Otsu's
binarization achieves a clear presentation of the abdominal vascular
structure.
To further filter out slices related to the spine, the YOLOv4 model is
employed for object recognition, effectively locating the affected area.
The model demonstrates a precision of 93%, recall of 95%, and F1-score
of 93.9%, laying a solid foundation for subsequent organ localization.
Secondly, this study introduces an image recognition step for kidney
localization. By removing the upper and lower extension regions of the
spine, unnecessary computations are reduced. Utilizing YOLOv4 and YOLOv7
models, slices with kidneys are successfully identified and filtered for
further selection of slices at the connection between the aorta and
renal arteries. The models exhibit various indicators; YOLOv4 achieves
a precision of 97% for the left kidney, 96% recall, and 96.9% F1-score.
iii
For the right kidney, the precision is 90%, recall is 94%, and F1-score
is 91.9%. YOLOv7 demonstrates a precision of 91% for the left kidney,
99% recall, and 94.8% F1-score. For the right kidney, precision is 88%,
recall is 100%, and F1-score is 93.6%.
Subsequently, this study performs image segmentation of the vascular
contour using the U-net model. Through model training, various indicators
are achieved: segmentation of the aorta with a precision of 96.7%, recall
of 96.2%, F1-score of 96.4%, and accuracy of 95.3%. Segmentation of the
connection between the aorta and renal arteries attains a precision of
96.5%, recall of 89.8%, F1-score of 93%, and accuracy of 87.2%.
Finally, through preprocessing and voxel point establishment, this study
utilizes a triangular mesh to construct a 3D STL format vascular contour
model, facilitating preoperative assessment for medical teams. This
methodology not only holds the promise of enhancing the precision and
efficiency of AAA treatment but also provides valuable insights for
further research in the medical image processing field. Through the
comprehensive application of various image processing techniques and
deep learning models, it presents a novel and more precise approach for
the diagnosis and treatment of AAA.
摘要 i
Abstract ii
目錄 iv
表目錄 vii
圖目錄 viii
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 2
1.3論文架構 2
第二章 相關文獻回顧 3
2.1醫學知識、手術與解剖學 3
2.1.1動脈瘤內腔修補手術 3
2.1.2 支架漏血 4
2.1.3 EVAR手術支架安裝位置 4
2.1.4腎臟對應脊椎椎數出現區域 5
2.2 解剖平面 6
2.2.1 矢狀面 6
2.2.2 冠狀面 7
2.2.3 水平面 7
2.3 影像處理相關技術 8
2.3.1 對比拉開 8
2.3.2 Otsu’s演算法 8
2.3.3 連通分量標記 9
2.3.4 DICOM格式線性轉換至BMP格式 10
2.4 卷積神經網路 11
2.4.1 卷積層 11
2.4.2 池化層 12
2.4.3 全連接層 13
2.5 物件偵測模型 14
2.5.1 YOLOv4 14
2.5.2 YOLOv7 17
2.6 物件分割模型 19
第三章 研究方法 21
3.1 CT影像前處理 22
3.1.1 DICOM輸出BMP 22
3.1.2 對比強化 23
3.1.3 Otsu's演算法(群間最大差異二質化) 24
3.1.4連通分量標記 25
3.2 脊椎辨識定位 26
3.2.1脊椎資料集標記 26
3.2.2辨識結果與胸椎與腰椎基準點 28
3.3 腎臟辨識定位 29
3.3.1 去除脊椎上下延伸之前處理 30
3.3.2 腎臟物件偵測 30
3.4 血管輪廓分割 31
3.4.1 U-net模型訓練與血管輪廓分割 31
3.4.2 U-net模型訓練腹主動脈與腎動脈血管輪廓分割 33
3.5建立血管輪廓模型 34
3.5.1 邊緣前處理 34
3.5.2 建立三維模型 35
第四章 實驗結果與討論 36
4.1 實驗使用資料集 36
4.2 軟體與硬體環境設置 37
4.3 脊椎辨識參數與各項指標 38
4.4 腎臟辨識參數與各項指標 39
4.5血管輪廓分割參數與各項指標 41
4.6 STL格式三維模型 43
第五章 結論與未來展望 46
參考文獻 47
[1] Kent, K. Craig, et al. "Analysis of risk factors for abdominal aortic aneurysm in a cohort of more than 3 million individuals." Journal of vascular surgery 52.3 (2010): 539-548.
[2] https://vascular.luxliving.hk/aaa/
[3] Greenhalgh, R. M., and Charing Cross Campus. "Endovascular aneurysm repair versus open repair in patients with abdominal aortic aneurysm (EVAR trial 1): randomised controlled trial." surgery 5 (2005): 6.
[4] Biasi, Lukla, et al. "Intraoperative DynaCT detection and immediate correction of a type 1a endoleak following endovascular repair of abdominal aortic aneurysm." Cardiovascular and interventional radiology 32.3 (2009): 535-538.
[5] Wanhainen, A., et al. "European Society for Vascular Surgery Guidelines on the management of Aort0-iliac Abdominal Aortic Aneurysms."
[6] Kassem, Tamer W. "Follow up CT angiography post EVAR: Endoleaks detection, classification and management planning." The Egyptian Journal of Radiology and Nuclear Medicine 48.3 (2017): 621-626.
[7] Endovascular treatment of abdominal aortic aneurysms (EVAR) -CIRSE https://www.cirse.org/patients/general-information/ir-procedures/endovascular-treatment-of-abdominal-aortic-aneurysms-evar/
[8] Gray, H. (2013). Grays Anatomy. Arcturus Publishing.
[9] Gross Anatomy of the Kindy | Biology of Aging https://courses.lumenlearning.com/atd-herkimer-biologyofaging/chapter/gross-anatomy-of-the-kidney/
[10] Xu, Chenyang, and Jerry L. Prince. "Snakes, shapes, and gradient vector flow." IEEE Transactions on image processing 7.3 (1998): 359-369.
[11] Guan, Xu, et al. "An image enhancement method based on gamma correction." 2009 Second international symposium on computational intelligence and design. Vol. 1. IEEE, 2009.
[12] Otsu, Nobuyuki. "A threshold selection method from gray-level histograms." IEEE transactions on systems, man, and cybernetics 9.1 (1979): 62-66.
[13] Suzuki, Satoshi. "Topological structural analysis of digitized binary images by border following." Computer vision, graphics, and image processing 30.1 (1985): 32-46.

[14] He, Yixuan, Tianyi Hu, and Delu Zeng. "Scan-flood fill (SCAFF): An efficient automatic precise region filling algorithm for complicated regions." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. 2019.
[15] Williamson, Jonathan P., et al. "Quantifying tracheobronchial tree dimensions: methods, limitations and emerging techniques." European Respiratory Journal 34.1 (2009): 42-55.
[16] 解剖學方位 – Wikipedia
https://zh.wikipedia.org/zh-tw/%E8%A7%A3%E5%89%96%E5%AD%A6%E6%96%B9%E4%BD%8D
[17] Teilo, K. "An Introduction to Convolutional Neural Networks." arXiv preprint arXiv:1511.08458 (2015).
[18] R. E. Turner, “Lecture 14 : Convolutional neural networks for computer vision,”2014.
[19] Girshick, Ross, et al. "Rich feature hierarchies for accurate object detection and semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2014.
[20] Ren, Shaoqing, et al. "Faster r-cnn: Towards real-time object detection with region proposal networks." Advances in neural information processing systems 28 (2015).
[21] Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014).
[22] Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. "Yolov4: Optimal speed and accuracy of object detection." arXiv preprint arXiv:2004.10934 (2020).
[23] Wang, Chien-Yao, Alexey Bochkovskiy, and Hong-Yuan Mark Liao. "YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023.
[24] Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement." arXiv preprint arXiv:1804.02767 (2018).
[25] Wang, Chien-Yao, et al. "CSPNet: A new backbone that can enhance learning capability of CNN." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. 2020.

[26] He, Kaiming, et al. "Spatial pyramid pooling in deep convolutional networks for visual recognition." IEEE transactions on pattern analysis and machine intelligence 37.9 (2015): 1904-1916.
[27] Liu, Shu, et al. "Path aggregation network for instance segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
[28] Yun, Sangdoo, et al. "Cutmix: Regularization strategy to train strong classifiers with localizable features." Proceedings of the IEEE/CVF international conference on computer vision. 2019.
[29] Zheng, Zhaohui, et al. "Distance-IoU loss: Faster and better learning for bounding box regression." Proceedings of the AAAI conference on artificial intelligence. Vol. 34. No. 07. 2020.
[30] Hosang, Jan, Rodrigo Benenson, and Bernt Schiele. "Learning non-maximum suppression." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[31] Wang, Chien-Yao, Hong-Yuan Mark Liao, and I-Hau Yeh. "Designing network design strategies through gradient path analysis." arXiv preprint arXiv:2211.04800 (2022).
[32] Lin, Tsung-Yi, et al. "Feature pyramid networks for object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[33] Liu, Shu, et al. "Path aggregation network for instance segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
[34] Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. "U-net: Convolutional networks for biomedical image segmentation." Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18. Springer International Publishing, 2015.
[35] Long, Jonathan, Evan Shelhamer, and Trevor Darrell. "Fully convolutional networks for semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2015.
[36] Tzutalin. LabelImg. Git code (2015).
https://github.com/tzutalin/labelImg.
[37] Wada, K. (November). Labelme: Image Polygonal Annotation with Python. Retrieved November 7, 2018
https://github.com/wkentaro/labelme
[38] STL(file format) - Wikipedia
https://en.wikipedia.org/wiki/STL_(file_format)
電子全文 電子全文(網際網路公開日期:20290122)
連結至畢業學校之論文網頁點我開啟連結
註: 此連結為研究生畢業學校所提供,不一定有電子全文可供下載,若連結有誤,請點選上方之〝勘誤回報〞功能,我們會盡快修正,謝謝!
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top