深度學習及加速強健特徵之CT影像跟骨骨折辨識及偵測

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：7

、訪客IP：18.216.121.55

姓名

优嘉逸(Yoga Dwi Pranata) 查詢紙本館藏

畢業系所

資訊工程學系在職專班

論文名稱

深度學習及加速強健特徵之CT影像跟骨骨折辨識及偵測
(Deep Learning and Speeded Up Robust Features (SURF) for Calcaneus Fracture Classification and Detection on CT Images)

相關論文

★ Single and Multi-Label Environmental Sound Recognition with Gaussian Process	★ 波束形成與音訊前處理之嵌入式系統實現
★ 語音合成及語者轉換之應用與設計	★ 基於語意之輿情分析系統
★ 高品質口述系統之設計與應用	★ 基於風格向量空間之個性化協同過濾服裝推薦系統
★ RetinaNet應用於人臉偵測	★ 金融商品走勢預測
★ 整合深度學習方法預測年齡以及衰老基因之研究	★ 漢語之端到端語音合成研究
★ 基於 ARM 架構上的 ORB-SLAM2 的應用與改進	★ 基於深度學習之指數股票型基金趨勢預測
★ 探討財經新聞與金融趨勢的相關性	★ 基於卷積神經網路的情緒語音分析
★ 運用深度學習方法預測阿茲海默症惡化與腦中風手術存活	★ 運用LLM自動生成食譜方法與系統

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 ( 永不開放)

摘要(中)

跟骨骨也稱為足跟骨，是形成足部後部的最大的t骨。支柱骨與其前後兩側與距骨骨骼相連。跟骨被認為是最易發生骨折的t骨。跟骨骨折僅佔所有骨折的2％，但60％的t骨骨折[1]。根據距下關節受累，跟骨骨折可分為關節間骨折和關節外骨折兩種。關節後骨折是較常見的後路關節面涉及跟骨。
患者數據可以以幾種成像格式存儲，例如，計算機斷層掃描（CT）數據。 CT圖像是最近用於確定患者疾病的醫學圖像的演變。它是構建3-D圖像的2-D圖像序列。 CT圖像包含大量信息，例如可能無法通過目視檢查徹底和準確地分析的每個2D圖像切片信息。
本研究提出了一種對跟骨骨CT圖像進行分類和檢測骨折的新方法。在這個實驗中，我們在兩個數據集中進行實驗。第一個只有一個Dicom案例，第二個在所有數據集（兩個Dicom案例）中。我們結合形態操作和邊緣檢測方法，在分類和檢測過程中實現更好的輸入。在預處理步驟中，我們調整尺寸相同（224×224）的圖像以適應CNN方法。之後，我們轉換灰度圖像中的輸入圖像。之後，我們用圖像減去圖像的意思。卷積神經網絡也應用於分類過程，將骨分為幾類。我們從冠狀，橫向和矢狀視角的三個視角分為兩類（骨折和非骨折）。分類後，我們從骨折類別進行檢測，並確定跟骨的哪一部分被破壞。檢測也是三個觀點，也是橫向，矢狀和冠狀視角

摘要(英)

Calcaneus, also called as heel bone, is the largest tarsal bone that forms the rear part of the foot. Cuboid bone articulates with its anterior and superior sides together with talus bone. Calcaneus is known to be the most fracture prone tarsal bone. Calcaneal fractures represent only about 2% of all fractures but 60% of tarsal bones fractures[1]. Based on subtalar joint involvement, calcaneal fractures can be categorized into two types: intraarticular fracture and extraarticular fracture. Intraarticular fractures are more common where posterior talar articular facet involves calcaneus.
Patient data can be stored in several kinds of imaging format, e.g. Computer Tomography (CT) data. The CT images is the evolution of the medical images that recently used for determine the disease from the patient. It is a sequence of 2-D images that construct 3-D images. CT images contain a significant amount of information, such as fracture information in each slice of 2-D images that may not be thoroughly and accurately analyzed via visual inspection.
This study proposed a new method to classify and detect the fracture in calcaneus bone CT images. In this experiment, we do the experiment in two dataset. The first one in just one Dicom case and the second one in the all dataset (two Dicom case). Both morphological operation and edge detection methods were combined in order to achieve better input in classification and detection processes. In the pre-processing step, the images were resized into the same size (224 x 224) to fit in the CNN method. After that the input images were converted in the grayscale images. After that, the images were subtracted with the images mean. Convolutional Neural Network was also applied in the classification process in order to classify the bone into several classes. Two classes were classified (fracture and non-fracture) from three views that are coronal, transversal, and sagittal view. After classification, the detection were done from the fracture class and determine in which part of calcaneus bone was broken. The detection is from three views also that is transversal, sagittal, and coronal views.

關鍵字(中)

★ Convolutional Neural Network
★ Calcaneus
★ Segmentation
★ Classification
★ Detection

關鍵字(英)

★ Convolutional Neural Network
★ Calcaneus
★ Segmentation
★ Classification
★ Detection

論文目次

中文摘要 i
ABSTRACT iii
ACKNOWLEDGEMENT v
TABLE OF CONTENTS vii
LIST OF FIGURES ix
LIST OF TABLES xv
CHAPTER 1 INTRODUCTION 1
CHAPTER 2 LITERATURE STUDY 4
2.1 Medical Images Evolution 4
2.1.1 X-Ray Images 4
2.1.2 Magnetic Resonance Imaging (MRI) 8
2.2 Calcaneus Bone 14
2.2.1 Calcaneus Bone Fracture 17
2.3 Deep Learning 21
2.3.1 Convolutional Neural Network 25
2.4 SURF (Speeded Up Robust Feature) 34
CHAPTER 3 PROPOSED SYSTEM 41
3.1 Data Collection 41
3.1.1 Patient Data Collection 41
3.2 Classification Pre-Processing 46
3.2.1 Resize the Image 47
3.2.2 Change the Image to Vector 48
3.2.3 Normalize the Images 48
3.3 Classification Step (Convolutional Neural Network) 49
3.3.1 VGG Network Architecture 50
3.3.2 Residual Network Architecture 53
3.4 Detection Pre-Processing 55
3.4.1 Create the Fracture Region 55
3.5 Detection Step (Speeded Up Robust Feature) 57
3.5.1 Detect the Feature 57
3.5.2 Extract the Feature 58
3.5.3 Match the Feature 59
3.5.4 Find the Homograph 60
3.5.5 Input the Reference 61
3.5.6 Find the Red Area 62
3.5.7 Crop the Images 62
3.5.8 Find the Edges 62
3.5.9 Find the Contour 63
3.5.10 Blend the Images 63
CHAPTER 4 EXPERIMENT AND DISCUSSION 64
4.1 Classification Pre-processing 64
4.2 Classification Step 67
4.2.1 VGG Network Architecture Experiment Setup 67
4.2.2 VGG Network Architecture Classification Result 68
4.2.3 Residual Network Experiment Setup 76
4.2.4 Residual Network Classification Result 78
4.3 Detection Step 85
CHAPTER 5 CONCLUSION AND FUTURE WORK 102
5.1 Conclusion 102
5.2 Future Work 103
REFERENCES 104

參考文獻

1. K. Badillo, J. A. Pacheco, et al, “Multidetector CT Evaluation of Calcaneal Fractures,” radiographics, January-Febuary, 2011.
2. J. Wu, P. Davuluri et al, “Fracture Detection in Traumatic Pelvic CT Images,” International Journal of Biomedical Imaging, Vol 2012, 2012
3. A. T. C. Mallikarjunaswamy and R. Raman, “Detection of Bone Fracture using Image Processing Methods,” International Journal of Computer Applications, 2015
4. S. F. Kurniawan, I K. G. D. Putra, A.A. K. O. Sudana, “Bone Fracture Detection Using OPENCV,” Journal of Theoretical and Applied Information Technology, 2014
5. A. Sharma, P. Sharma, Rashmi, et al, “Edge Detection of Medical Images Using Morphological Operations,” International Journal for Science and Emerging Technologies with Latest Trend, 2012
6. Holger R. Roth, Y. Wang, et al, "Deep Convolutional Networks for automated detection of posterior-element fractures on spine CT," ArXiv, 2016
7. William G. Bradley, "History of Medical Imaging," Proceedings of American Philosophical Society, Vol 152, No. 3, 2008
8. R. Cierniak, "X-ray Computed Tomography in Biomedical Engineering," ISBN: 978-0-85729-026-7, 2011
9. Hongmei Zhu, "Medical Image Processing Overview," 2004
10. Gerald R. Aben,"CT Scanning," College of Osteopathic Medicine,2012
11. Thomas Flohr, B. Ohnesorge, "Multi-slice CT Technology," 2007
12. M. Schuenke, E. Schulte, U. Schumacher ,"Thieme Atlas of Anatomy: General Anatomy and Musculoskeletal System," ISBN 1-58890-419-9, (2006).
13. http://medical-dictionary.thefreedictionary.com/Calcaneus+bone
14. http://www.healthline.com/human-body-maps/calcaneus-bone
15. https://en.wikipedia.org/wiki/Calcaneus
16. JD. Richman, , PS. Barre , "The plantar ecchymosis sign in fractures of the calcaneus," Clinical Orthopaedics and Related Research (207): 122–5, 1986
17. Kenneth Badillo, Jose A. Pacheco, Samuel O. Padua, Angel A. Gomez, Edgar Colon, Jorge A. Vidal, "Multidetector CT Evaluation of Calcaneal Fractures," RadioGraphics. 31 (1): 81–92, 2011
18. https://radiopaedia.org/articles/sanders-ct-classification-of-calcaneal-fracture-2
19. M. Imran Razzak, Saeeda Naz, A. Zaib, "Deep Learning for Medical Image Processing: Overview, Challenges, and Future," arxiv, 2017
20. K. Fukushima ,"Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position," Biological Cybernetics, 36(4):193-202, 1980
21. G. Litjens, T. Kooi, B.E. Bejnordi, A.A.A. Setio, F. Ciompi, et. al,"A Survey on Deep Learning in Medical Image Analysis," Arxiv:1702.05747v1, 2017
22. David Stutz,"Understanding Convolutional Neural Networks," 2014
23. S. Kevin Zhou, H. Greenspan, D. Shen,"Deep Learning for Medical Image Analysis : 1st edition," Academic Press, 2017
24. H. Bay, T. Tuytelaars, L. V. Gool,"SURF: Speeded Up Robust Features," ECCV, (2006)
25. David G. Lowe,"Distinctive Image Features from Scale-Invariant Keypoints," International Journal of Computer Vision, (2004)
26. Krizhevsky, A., Sutskever, I., and Hinton, G. E., "ImageNet classification with deep convolutional neural networks," In NIPS, pp. 1106–1114, 2012.
27. J. Yim, J. Ju, H. Jung, J. Kim, "Image Classification Using Convolutional Neural Networks With Multi-stage Feature," Robot Intelligence Technology and Applications 3. Advances in Intelligent Systems and Computing, vol 345. Springer, Cham, 2015.
28. L. Kang, J. Kumar, P. Ye, Y. Liy, D. Doermann, "Convolutional Neural Networks for Document Image Classification," International Conference on Pattern Recognition (ICPR) 22nd , 2014
29. K. Simonyan and A. Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition," conference paper at ICLR, 2015.
30. S. Ioffe and C. Szegedy," Batch normalization: Accelerating deep network training by reducing internal covariate shift," ICML, 2015.
31. K. He, X. Zhang, S. Ren, and J. Sun ,"Delving deep into rectifiers: Surpassing human-level performance on imagenet classification," ICCV, 2015.
32. https://www.mathworks.com/help/vision/ref/extractfeatures.html

指導教授

王家慶(Dr. Jia-Ching Wang)

審核日期

2017-7-26

推文