成功大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	陳瑾漩 Chen, Jin-Xuan
論文名稱：	基於關鍵點的深度學習方法自動評估全景及根尖X光片上的牙周骨損失 Keypoint-based Deep Learning Methods in Automatic Bone Loss Evaluation for Panoramic and Periapical Radiographs
指導教授：	孫永年 Sun, Yung-Nien
學位類別：	碩士 Master
系所名稱：	電機資訊學院 - 資訊工程學系 Department of Computer Science and Information Engineering
論文出版年：	2023
畢業學年度：	111
語文別：	英文
論文頁數：	58
中文關鍵詞：	卷積神經網路、變換器、全景X光片、根尖X光片、牙周骨損失、牙周病
外文關鍵詞：	Convolutional Neural Network, Transformer, Panoramic Radiographs, Periapical Radiographs, Alveolar Bone Loss, Periodontal
相關次數：	點閱：83 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

牙周病是現代社會常見的一種慢性疾病。其初期症狀往往不明顯，由於許多患者在牙周病急性發作或出現劇烈疼痛時才就醫，因此早期診斷牙周病對於預防其進一步發展至嚴重程度至關重要。
臨床上醫生們可藉由牙周骨損失率來判斷牙齒牙周病的嚴重程度；近年來也有許多的討論及研究提及這個議題的重要性。傳統上，醫生在評估牙周病嚴重程度時，通常根據牙周骨損失率來做判斷。然而手動計算牙周骨損失率既耗時又需要醫生豐富的經驗，不同醫生之間的診斷結果也可能存在差異。
因此，本次研究我們希望能透過深度學習的方法幫助醫生自動計算牙周骨的損失，並針對兩種牙齒疾病診斷上常用的醫學影像，第一部分的目標為偵測出全景X光片內每顆牙齒的三種關鍵點及其牙齒分割圖，利用預測出的關鍵點做牙周骨損失率的計算。第二部分的目標為偵測根尖X光片內每顆牙齒的三種關鍵點以及其牙齒分割圖，除了用使用關鍵點的評估方式以外，本研究會額外使用分割圖的面積比來計算牙周骨損失率。實驗結果表明，我們提出的三種評估方法在F1-Score方面優於醫生的直覺判斷。相對於醫生的人眼觀察，我們的系統能夠更準確且更快速地計算出牙齒的牙周骨損失率，從而在醫生臨床判斷的輔助上起到重要作用。

Periodontal disease is a common chronic condition in modern society. Its early symptoms are often subtle, and many patients only seek medical attention when the disease reaches an acute stage or causes severe pain. Therefore, early diagnosis of periodontal disease is crucial to prevent its progression to a more severe state.
Clinically, dentists assess the severity of periodontal disease based on the extent of alveolar bone loss. In recent years, there has been much discussion and research emphasizing the importance of this parameter. Traditionally, dentists manually calculate the alveolar bone loss rate to evaluate the severity of periodontal disease. However, this method is time-consuming and requires significant expertise, and there may be variations in diagnosis among different dentists.
Therefore, in this study, we aim to assist dentists in automatically calculating the extent of alveolar bone loss using deep learning methods. Specifically, we focus on two commonly used dental imaging modalities for diagnosing dental conditions. In the first part, our goal is to detect the three keypoints for each tooth in panoramic Radiographs and segment the teeth, enabling the calculation of the alveolar bone loss rate based on the predicted landmarks. In the second part, we aim to detect the three keypoints and tooth segmentation in periapical Radiographs. In addition to using the keypoints for evaluation, we incorporate the area ratio from the segmentation maps to calculate the alveolar bone loss rate. Experimental results demonstrate that our proposed evaluation methods outperform the intuitive judgments of dentists in terms of F1-Score. Compared to visual observation by dentists, our system can more accurately and quickly calculate the extent of alveolar bone loss, thus playing a crucial role in assisting clinical judgments.

iii
Contents
摘要 ................................................................................................................................ i
Contents ........................................................................................................................ iii
致謝 ............................................................................................................................... v
List of Tables ................................................................................................................ vi
List of Figures ............................................................................................................. viii
CHAPTER 1 INTRODUCTION ................................................................................... 1
1 Research Background and Motivation .................................................................... 1
2 Panoramic and Periapical Radiography................................................................... 2
3 Related Work ........................................................................................................... 3
3.1 Periodontal Bone Loss Evaluation on panoramic radiographs ............................. 3
3.2 Periodontal Bone Loss Evaluation on periapical radiographs .............................. 3
3.3 Object Detection Based on CNN .......................................................................... 4
3.4 Object Segmentation Based CNN ........................................................................ 5
3.5 Keypoints Detection Based on CNN .................................................................... 6
4 Overview ................................................................................................................. 6
CHAPTER 2 METHODOLOGY .................................................................................. 8
1 Method Overview .................................................................................................... 8
2 Method of Panoramic Radiographs System ............................................................ 8
3 Method of Periapical Radiographs System ............................................................. 9
4 Object Detection Branch ....................................................................................... 10
5 Object Segmentation Branch ................................................................................. 11
5.1 Object Segmentation Architecture ...................................................................... 11
5.2 Neighborhood Attention Transformer ................................................................ 12
5.3 Neighborhood Self Attention .............................................................................. 13
5.4 CBAM ................................................................................................................ 14
6 Keypoint Detection Branch ................................................................................... 16
6.1 Ghost Module ..................................................................................................... 17
6.2 ASPP ................................................................................................................... 18
CHAPTER 3 EXPERIMENTAL RESULTS AND DISSCUSSION........................... 20
1 Data Collection ...................................................................................................... 20
1.1 Annotation Process of Panoramic Radiographs ................................................. 20
1.2 Annotation Process of Periapical Radiographs................................................... 22
1.3 Five-Fold Cross Validation ................................................................................. 24
1.3.1 Five-Fold Cross Validation for Object Detection ............................................ 25
1.3.2 Five-Fold Cross Validation for Object Segmentation...................................... 25
iv
1.3.3 Five-Fold Cross Validation for Keypoint Detection ........................................ 26
2 Experimental setup ................................................................................................ 26
3 Evaluation Metrics and Experimental Results....................................................... 27
3.1 Metrics and Experimental Results for Objection Detection ............................... 27
3.2 Metrics and Experimental Results for Objection Segmentation ........................ 30
3.3 Metrics and Experimental Results for Keypoint Detection ................................ 31
4 Severity Evaluation of Periodontal Bone Loss ...................................................... 32
4.1 Methods for Assessing the Severity of Periodontal Bone Loss in Panoramic
Radiographs ................................................................................................................. 33
4.2 Methods for Assessing the Severity of Periodontal Bone Loss in Periapical
Radiographs ................................................................................................................. 35
4.3 Results of Severity Evaluation of Periodontal Bone Loss .................................. 36
4.3.1 Periodontal Bone Loss on Panoramic Radiographs ........................................ 36
4.3.2 Periodontal Bone Loss on Periapical Radiograph ........................................... 42
4.3.3 Comparison with Reference ............................................................................ 44
4.4 Visualization of Severity Evaluation Results of Periodontal Bone Loss............ 46
5 Ablation Study ....................................................................................................... 48
5.1 Ablation Study for Object Segmentation ............................................................ 48
5.2 Ablation Study for Keypoint Detection .............................................................. 49
6 Dataset Consistency Assessment ........................................................................... 49
6.1 Cohen's Kappa Coefficient ................................................................................. 50
6.2 Inter-examiner consistency ................................................................................. 51
6.3 Intra-examiner consistency ................................................................................. 51
7 The time for annotating a single radiographs ........................................................ 53
CHAPTER 4 CONCLUSION AND FUTURE WORK .............................................. 54
1 Conclusion ............................................................................................................. 54
2 Future Work ........................................................................................................... 55
References ................................................................................................................... 56
                                    

[1] H.-J. Chang et al., "Deep Learning Hybrid Method to Automatically Diagnose Periodontal Bone Loss and Stage Periodontitis," Scientific Reports, vol. 10, no. 1, p. 7531, 2020/05/05 2020, doi: 10.1038/s41598-020-64509-z.
[2] H. S. Shon et al., "Deep Learning Model for Classifying Periodontitis Stages on Dental Panoramic Radiography," Applied Sciences, vol. 12, no. 17, p. 8500, 2022. [Online]. Available: https://www.mdpi.com/2076-3417/12/17/8500.
[3] Z. Kong et al., "Automated periodontitis bone loss diagnosis in panoramic radiographs using a bespoke two-stage detector," Computers in Biology and Medicine, vol. 152, p. 106374, 2023/01/01/ 2023, doi: https://doi.org/10.1016/j.compbiomed.2022.106374.
[4] B. Thanathornwong and S. Suebnukarn, "Automatic detection of periodontal compromised teeth in digital panoramic radiographs using faster regional convolutional neural networks," (in eng), Imaging Sci Dent, vol. 50, no. 2, pp. 169-174, Jun 2020, doi: 10.5624/isd.2020.50.2.169.
[5] S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks," p. arXiv:1506.01497doi: 10.48550/arXiv.1506.01497.
[6] L. Jiang, D. Chen, Z. Cao, F. Wu, H. Zhu, and F. Zhu, "A two-stage deep learning architecture for radiographic staging of periodontal bone loss," (in eng), BMC Oral Health, vol. 22, no. 1, p. 106, Apr 1 2022, doi: 10.1186/s12903-022-02119-z.
[7] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," in International Conference on Medical image computing and computer-assisted intervention, 2015: Springer, pp. 234-241.
[8] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, "YOLOv4: Optimal Speed and Accuracy of Object Detection," p. arXiv:2004.10934doi: 10.48550/arXiv.2004.10934.
[9] R. P. Danks et al., "Automating Periodontal bone loss measurement via dental landmark localisation," International Journal of Computer Assisted Radiology and Surgery, vol. 16, no. 7, pp. 1189-1199, 2021/07/01 2021, doi: 10.1007/s11548-021-02431-z.
[10] A. Newell, K. Yang, and J. Deng, "Stacked Hourglass Networks for Human Pose Estimation," p. arXiv:1603.06937doi: 10.48550/arXiv.1603.06937.
[11] C.-C. Chen et al., "Automatic recognition of teeth and periodontal bone loss measurement in digital radiographs using deep-learning artificial intelligence," Journal of Dental Sciences, vol. 18, no. 3, pp. 1301-1309, 2023/07/01/ 2023, doi: https://doi.org/10.1016/j.jds.2023.03.020.
[12] J. Y. Cha, H. I. Yoon, I. S. Yeo, K. H. Huh, and J. S. Han, "Peri-Implant Bone Loss Measurement Using a Region-Based Convolutional Neural Network on Dental Periapical Radiographs," (in eng), J Clin Med, vol. 10, no. 5, Mar 2 2021, doi: 10.3390/jcm10051009.
[13] K. He, G. Gkioxari, P. Dollár, and R. Girshick, "Mask r-cnn," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2961-2969.
[14] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once: Unified, real-time object detection," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779-788.
[15] W. Liu et al., "SSD: Single Shot MultiBox Detector," p. arXiv:1512.02325doi: 10.48550/arXiv.1512.02325.
[16] R. Girshick, "Fast R-CNN," p. arXiv:1504.08083doi: 10.48550/arXiv.1504.08083.
[17] J. Uijlings, K. Sande, T. Gevers, and A. W. M. Smeulders, "Selective Search for Object Recognition," International Journal of Computer Vision, vol. 104, pp. 154-171, 09/01 2013, doi: 10.1007/s11263-013-0620-5.
[18] Z. Zhou, M. Mahfuzur Rahman Siddiquee, N. Tajbakhsh, and J. Liang, "UNet++: A Nested U-Net Architecture for Medical Image Segmentation," p. arXiv:1807.10165doi: 10.48550/arXiv.1807.10165.
[19] X. Qin, Z. Zhang, C. Huang, M. Dehghan, O. R. Zaiane, and M. Jagersand, "U2-Net: Going deeper with nested U-structure for salient object detection," Pattern recognition, vol. 106, p. 107404, 2020.
[20] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, "YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors," p. arXiv:2207.02696doi: 10.48550/arXiv.2207.02696.
[21] M. Tan and Q. V. Le, "EfficientNetV2: Smaller Models and Faster Training," p. arXiv:2104.00298doi: 10.48550/arXiv.2104.00298.
[22] S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, "CBAM: Convolutional Block Attention Module," p. arXiv:1807.06521doi: 10.48550/arXiv.1807.06521.
[23] A. Hassani, S. Walton, J. Li, S. Li, and H. Shi, "Neighborhood Attention Transformer," p. arXiv:2204.07143doi: 10.48550/arXiv.2204.07143.
[24] M. Tan and Q. Le, "Efficientnet: Rethinking model scaling for convolutional neural networks," in International conference on machine learning, 2019: PMLR, pp. 6105-6114.
[25] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "MobileNetV2: Inverted Residuals and Linear Bottlenecks," p. arXiv:1801.04381doi: 10.48550/arXiv.1801.04381.
[26] J. Hu, L. Shen, S. Albanie, G. Sun, and E. Wu, "Squeeze-and-Excitation Networks," p. arXiv:1709.01507doi: 10.48550/arXiv.1709.01507.
[27] K. Han, Y. Wang, Q. Tian, J. Guo, C. Xu, and C. Xu, "GhostNet: More Features from Cheap Operations," p. arXiv:1911.11907doi: 10.48550/arXiv.1911.11907.
[28] L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille, "DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs," p. arXiv:1606.00915doi: 10.48550/arXiv.1606.00915.
[29] D. P. Kingma and J. Ba, "Adam: A Method for Stochastic Optimization," p. arXiv:1412.6980doi: 10.48550/arXiv.1412.6980.
[30] K.-F. Chen. "Automatic Evaluation of Radiographic Periodontal Bone Loss on Panoramic and Periapical Radiographs Using Deep Learning Method." https://hdl.handle.net/11296/8jg3w5 (accessed.
[31] M. L. McHugh, "Interrater reliability: the kappa statistic," (in eng), Biochem Med (Zagreb), vol. 22, no. 3, pp. 276-82, 2012.

校內：2028-08-29公開
校外：2028-08-29公開電子論文尚未授權公開，紙本請查館藏目錄

簡易檢索 / 詳目顯示

相關論文