(3.236.175.108) 您好!臺灣時間:2021/02/27 06:10
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:簡祥秦
研究生(外文):CHIEN, HSIANG-CHIN
論文名稱:以深度學習架構應用於冠狀動脈心血管影像分割及標 註平台開發
論文名稱(外文):Application of Deep Learning Architecture to the Development of Coronary Artery Cardiovascular Image Segmentation and Annotation Platform
指導教授:李佳燕李佳燕引用關係
指導教授(外文):LEE, CHIA-YEN
口試委員:王宗道李文正
口試委員(外文):WANG, TZUNG-DAULEE, WEN-JENG
口試日期:2020-07-24
學位類別:碩士
校院名稱:國立聯合大學
系所名稱:電機工程學系碩士班
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2020
畢業學年度:108
語文別:中文
論文頁數:62
中文關鍵詞:冠心病臟冠狀動脈電腦斷層攝影multi-channel U-Net自動冠狀動脈分 割標註平台深度學習
外文關鍵詞:Coronary heart diseasecoronary computed tomographic angiographymultichannel U-Netautomatic coronary segmentationannotation platformdeep learning
相關次數:
  • 被引用被引用:0
  • 點閱點閱:42
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
心臟為人類重要器官之一,功能為運輸血液以及養分,而提供心肌養分的血管稱為冠狀動脈;若冠狀動脈產生斑塊進而導致狹窄便會使血液供給不足,將造成心肌缺氧即便形成冠心病(Coronary Artery Disease, CAD)。CCTA 是臨床上常見的初步診斷方法,CCTA 可以提供基本的心臟、冠狀動脈及斑塊三維解剖構造外,CCTA 冠狀動脈分割更可以用於冠心病診斷例如:CT-FFR、狹窄分析、術前規劃以及手術引導,因此 CCTA 冠狀動脈血管分割是非常基礎的前置作業,但是冠狀動脈分割上仍存在許多挑戰性,而前現行的市售軟體無法達到全自分割仍然需要臨床人員介入,故本論文提出基於深度學習之自動冠狀動脈分割演算法並開發標註平台。
本論文提出 3D multi-channel U-Net 冠狀動脈血管分割演算法,使用 Frangi filter初步管狀特徵提取作為multi-channel管狀特徵的輸入,並使用VOI(Volume-of-Interest)以類管狀特徵做為引導以減少冠狀動脈血管與背景的不平衡問題,並使用 Dice loss與 Focal loss 作為損失函數,並為了更精準的評估分割演算法於臨床的可用性,本論文除了用常用於影像分割評估的 Dice,另外提出了三個性能評估指標(Dice_GT、Dice_GT+NCA、血管長度與分支數分佈),以評估冠狀動脈的邊界準確度與管狀動脈的完整度,與 3D FCN 相比本論文提出的 Proposed Dice Loss 與Proposed Focal Loss法可以更準確 且豐富 地 分 出 冠 狀 動 脈FCN(Dice:0.71,Dice_GT+NCA:0.75,Dice_GT:0.79, CA 血管長度:107.45mm, CA 血管支數:6.6 條) Proposed Dice Loss(Dice:0.72, Dice_GT+NCA:0.82, Dice_GT:0.82, CA 血管長度:276.81mm, CA 血管支數:17.1 條) Proposed Focal Loss(Dice:0.78,Dice_GT+NCA:0.84, Dice_GT:0.84, CA 血管長度:185.55mm, CA 血管支數:10.5 條);最後利用 python 3.6 配合 QT5 與 VTK 8.1.2將演算法包裝成使用者介面方便臨床人員使用。
The heart is one of the important organs of human beings. Its function is to transport nutrients and blood. The blood vessels that provide nutrients for the myocardium are called coronary arteries. If the coronary arteries occurred plaque and stenosis, the blood supply will be insufficient and caused Heart disease (Coronary Artery Disease, CAD). CCTA is a common preliminary diagnosis tool. CCTA can provide basic three-dimensional anatomical structures of the heart, coronary arteries and plaques, and CCTA coronary artery segmentation can also be used for coronary heart disease diagnosis such as CT-FFR, stenosis analysis, preoperative planning and surgical guidance. Thus CCTA coronary artery segmentation is a very important pre-work, but there are still many challenges in coronary segmentation. Current commercial software cannot achieve auto segmentation still requires clinical intervention. Therefore, this paper proposes a deep learning based auto coronary artery segmentation algorithm and the develop a coronary artery segmentation Graphical User Interface.
This paper proposes a Volume-of-interest(VOI) based 3D multi-channel U-Net coronary artery vessel segmentation algorithm, using Frangi filter to generated preliminary tubular feature map as the input of multi-channel. Than used preliminary tubular feature map to guided VOI cropping reduce the sample imbalance problem between coronary arteries and background. This paper used Dice loss and Focal loss for deep learning loss function. To more accurately evaluate the clinical usability of the segmentation algorithm, this paper used Dice, which is commonly used for image segmentation evaluation, and also proposes three performance evaluation metrics(Dice_GT, Dice_GT+NCA, coronary artery length and branch number distribution) to evaluate the quality of coronary artery segmentation boundary and the integrity of the coronary artery. Compared with 3D FCN, the Proposed Dice Loss and Proposed Focal Loss can more accurately and richly segmented coronary artery FCN (Dice: 0.71, Dice_GT+NCA: 0.75, Dice_GT: 0.79, CA vessel length: 107.45mm, CA vessel number: 6.6 numbers) Proposed Dice Loss (Dice: 0.72, Dice_GT+NCA: 0.82, Dice_GT :0.82, CA vessel length: 276.81mm, CA vessel branch number: 17.1 numbers) Proposed Focal Loss (Dice: 0.78, Dice_GT+NCA: 0.84, Dice_GT: 0.84, CA vessel length: 185.55mm, CA vessel branch number: 10.5 numbers). Finally, this paper use python 3.6 with QT5 and VTK 8.1.2 to package the algorithm into a user interface for clinical use.

考試委員審定書 I
致謝 II
摘要 III
Abstract IV
目錄 VI
圖目錄 VIII
表目錄 X
第一章 緒論 1
1.1研究背景與目的 1
1.2 文獻探討 3
1.3 論文架構 8
第二章 基礎理論 9
2.1 Frangi Filter 9
2.2 卷積神經網路(Convolutional Neural Network , CNN) 10
2.2.1 卷積層(Convolution Layer) 11
2.2.2 池化層(Pooling Layer) 12
2.2.3 全連接層(Fully Connected Layer) 12
2.3 全卷積神經網路(Fully Convolutional Networks, FCN) 13
2.4 Focal Loss 15
2.5 Dice Loss 16
2.6 Performance Assessment 16
2.7 體繪製(Volume Rendering) 17
第三章 研究材料及方法 18
3.1 研究材料 18
3.2 研究流程 19
3.2.1 資料前處理(Data Pre-processing) 20
3.2.2 multi-channel U-Net 24
3.2.3 後處理(Post-Processing) 25
3.2.4 使用者介面(Graphical User Interface, GUI) 26
3.2.5 Performance Assessment 26
第四章 結果與討論 30
4.1 基於Dice的評估指標 30
4.2 血管長度支數評估分析 32
4.3 使用者介面 34
4.4 討論 36
4.5 演算法限制 41
第五章 結論 42
參考資料 43
附錄 48


[1]N. Townsend, et al., "Cardiovascular disease in Europe: epidemiological update 2016," European heart journal, vol. 37, pp. 3232-3245, 2016.
[2]G. Mowatt, et al., "64-Slice computed tomography angiography in the diagnosis and assessment of coronary artery disease: systematic review and meta-analysis," Heart, vol. 94, pp. 1386-1393, 2008.
[3]J. K. Min, et al., "Prognostic value of multidetector coronary computed tomographic angiography for prediction of all-cause mortality," Journal of the American College of Cardiology, vol. 50, pp. 1161-1170, 2007.
[4]M. F. Wilson, et al., "FFR As A Gold Standard For CT-FFR Validation. Does The FFR Measurement Device Alter The Flow?," Journal of Cardiovascular Computed Tomography, vol. 13, p. S27, 2019.
[5]A. F. Frangi, et al., "Multiscale vessel enhancement filtering," in International conference on medical image computing and computer-assisted intervention, 1998, pp. 130-137.
[6]R. Moreno and Ö. Smedby, "Gradient-based enhancement of tubular structures in medical images," Medical image analysis, vol. 26, pp. 19-29, 2015.
[7]S.-L. Papadopoulou, et al., "Assessment of atherosclerotic plaques at coronary bifurcations with multidetector computed tomography angiography and intravascular ultrasound-virtual histology," European Heart Journal–Cardiovascular Imaging, vol. 13, pp. 635-642, 2012.
[8]G. Yang, et al., "Quantitative measurement of coronary artery stenosis in CCTA images using a 2D parametric intensity model," in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2014, pp. 1071-1074.
[9]Z. Gao, et al., "Automatic segmentation of coronary tree in CT angiography images," International Journal of Adaptive Control and Signal Processing, vol. 33, pp. 1239-1247, 2019.
[10]C. P. Cheng, et al., "Optimization of three-dimensional modeling for geometric precision and efficiency for healthy and diseased aortas," Computer methods in biomechanics and biomedical engineering, vol. 21, pp. 65-74, 2018.
[11]Y. Wang and P. Liatsis, "A fully automated framework for segmentation and stenosis quantification of coronary arteries in 3D CTA imaging," in 2009 Second International Conference on Developments in eSystems Engineering, 2009, pp. 136-140.

[12]D. Lesage, et al., "A review of 3D vessel lumen segmentation techniques: Models, features and extraction schemes," Medical image analysis, vol. 13, pp. 819-845, 2009.
[13]M. T. Dehkordi, et al., "A review of coronary vessel segmentation algorithms," Journal of medical signals and sensors, vol. 1, p. 49, 2011.
[14]A. Hennemuth, et al., "One-click coronary tree segmentation in CT angiographic images," in International Congress Series, 2005, pp. 317-321.
[15]C. Wang, et al., "Vessel segmentation using implicit model-guided level sets," in MICCAI Workshop" 3D Cardiovascular Imaging: a MICCAI segmentation Challenge", Nice France, 1st of October 2012., 2012.
[16]F. Lugauer, et al., "Precise lumen segmentation in coronary computed tomography angiography," in International MICCAI Workshop on Medical Computer Vision, 2014, pp. 137-147.
[17]U. Sivalingam, et al., "Inner and outer coronary vessel wall segmentation from CCTA using an active contour model with machine learning-based 3D voxel context-aware image force," in Medical Imaging 2016: Computer-Aided Diagnosis, 2016, p. 978502.
[18]K. Fukushima, et al., "Neocognitron: A neural network model for a mechanism of visual pattern recognition," IEEE transactions on systems, man, and cybernetics, pp. 826-834, 1983.
[19]A. Krizhevsky, et al., "Imagenet classification with deep convolutional neural networks," in Advances in neural information processing systems, 2012, pp. 1097-1105.
[20]A. Graves, et al., "Speech recognition with deep recurrent neural networks," in 2013 IEEE international conference on acoustics, speech and signal processing, 2013, pp. 6645-6649.
[21]A. Graves, et al., "Hybrid speech recognition with deep bidirectional LSTM," in 2013 IEEE workshop on automatic speech recognition and understanding, 2013, pp. 273-278.
[22]Y. LeCun, et al., "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, pp. 2278-2324, 1998.
[23]K. He, et al., "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[24]J. Long, et al., "Fully convolutional networks for semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3431-3440.

[25]F. Milletari, et al., "V-net: Fully convolutional neural networks for volumetric medical image segmentation," in 2016 fourth international conference on 3D vision (3DV), 2016, pp. 565-571.
[26]W. W. Hsieh and B. Tang, "Applying neural network models to prediction and data analysis in meteorology and oceanography," Bulletin of the American Meteorological Society, vol. 79, pp. 1855-1870, 1998.
[27]G. Litjens, et al., "A survey on deep learning in medical image analysis," Medical image analysis, vol. 42, pp. 60-88, 2017.
[28]J. Fan, et al., "Multichannel fully convolutional network for coronary artery segmentation in x-ray angiograms," Ieee Access, vol. 6, pp. 44635-44643, 2018.
[29]P. Teikari, et al., "Deep learning convolutional networks for multiphoton microscopy vasculature segmentation," arXiv preprint arXiv:1606.02382, 2016.
[30]G. Tetteh, et al., "Deepvesselnet: Vessel segmentation, centerline prediction, and bifurcation detection in 3-d angiographic volumes," arXiv preprint arXiv:1803.09340, 2018.
[31]J. M. Wolterink, et al., "Coronary artery centerline extraction in cardiac CT angiography using a CNN-based orientation classifier," Medical image analysis, vol. 51, pp. 46-60, 2019.
[32]M. A. Gülsün, et al., "Coronary centerline extraction via optimal flow paths and CNN path pruning," in International Conference on Medical Image Computing and Computer-Assisted Intervention, 2016, pp. 317-325.
[33]P. Moeskops, et al., "Deep learning for multi-task medical image segmentation in multiple modalities," in International Conference on Medical Image Computing and Computer-Assisted Intervention, 2016, pp. 478-486.
[34]F. Chen, et al., "Automatic coronary artery lumen segmentation in computed tomography angiography using paired multi-scale 3D CNN," in Medical Imaging 2018: Biomedical Applications in Molecular, Structural, and Functional Imaging, 2018, p. 105782R.
[35]Y. Shen, et al., "Coronary arteries segmentation based on 3D FCN with attention gate and level set function," Ieee Access, vol. 7, pp. 42826-42835, 2019
[36]L. Yu, et al., "3D FractalNet: dense volumetric segmentation for cardiovascular MRI volumes," in Reconstruction, segmentation, and analysis of medical images, ed: Springer, 2016, pp. 103-110.
[37]O. Ronneberger, et al., "U-net: Convolutional networks for biomedical image segmentation," in International Conference on Medical image computing and computer-assisted intervention, 2015, pp. 234-241.


[38]W. Huang, et al., "Coronary artery segmentation by deep learning neural networks on computed tomographic coronary angiographic images," in 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2018, pp. 608-611
[39]P. Mirunalini, et al., "Segmentation of Coronary Arteries from CTA axial slices using Deep Learning techniques," in TENCON 2019-2019 IEEE Region 10 Conference (TENCON), 2019, pp. 2074-2080.
[40]L. Bargsten, et al., "Data Representations for Segmentation of Vascular Structures Using Convolutional Neural Networks with U-Net Architecture," in 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2019, pp. 989-992.
[41]B. Kong, et al., "Learning tree-structured representation for 3D coronary artery segmentation," Computerized Medical Imaging and Graphics, vol. 80, p. 101688, 2020.
[42]C. H. Sudre, et al., "Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations," in Deep learning in medical image analysis and multimodal learning for clinical decision support, ed: Springer, 2017, pp. 240-248.
[43]S. Yang, et al., "Major Vessel Segmentation on X-ray Coronary Angiography using Deep Networks with a Novel Penalty Loss Function," in International Conference on Medical Imaging with Deep Learning--Extended Abstract Track, 2019.
[44]H. Kervadec, et al., "Boundary loss for highly unbalanced segmentation," in International conference on medical imaging with deep learning, 2019, pp. 285-296.
[45]S. S. M. Salehi, et al., "Tversky loss function for image segmentation using 3D fully convolutional deep networks," in International Workshop on Machine Learning in Medical Imaging, 2017, pp. 379-387.
[46]I. Good, "Some terminology and notation in information theory," Proceedings of the IEE-Part C: Monographs, vol. 103, pp. 200-204, 1956.
[47]T.-Y. Lin, et al., "Focal loss for dense object detection," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2980-2988.
[48]N. Abraham and N. M. Khan, "A novel focal tversky loss function with improved attention u-net for lesion segmentation," in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), 2019, pp. 683-687.
[49]L. R. Dice, "Measures of the amount of ecologic association between species," Ecology, vol. 26, pp. 297-302, 1945.

[50]G. Van Rossum and F. L. Drake Jr, Python tutorial vol. 620: Centrum voor Wiskunde en Informatica Amsterdam, 1995.
[51]F. Chollet. (2015), Keras. Available: https://github.com/fchollet/keras
[52]W. Schroeder, et al., "The visualization toolkit, 4th edn. Kitware," New York, 2006.
[53]S. Hochreiter, et al., "Gradient flow in recurrent nets: the difficulty of learning long-term dependencies," ed: A field guide to dynamical recurrent neural networks. IEEE Press, 2001.
[54]R. Pascanu, et al., "Understanding the exploding gradient problem," CoRR, abs/1211.5063, vol. 2, p. 417, 2012.

電子全文 電子全文(網際網路公開日期:20250824)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔