牙醫師在幫病患進行牙齦內洗牙時，由於無法用肉眼確認牙齦內部的牙結石，使得以往是根據牙醫師的經驗來清除。單純根據牙醫經驗來清除牙齦內牙結石無法有效對於正確的牙齦內牙結石進行清除，導致遺漏未被清除的牙結石以及多餘的檢查，清除的過程在未使用麻醉的情況下，病人承受不必要的疼痛。光學同調斷層掃描(Optical Coherence Tomography, OCT)可穿透人體牙齦遮蔽進行牙齦內牙結石偵測，但在偵測牙齦內是否有牙結石這項技術還尚未成熟，掃描產生的雜訊以及過長的掃描時間造成人體實驗上的困難。在人體上進行偵測前，由於人體實驗不確定變因較多，需要在能控制變因的離體實驗上驗證其可行性，因此現今利用光學同調斷層掃描(OCT)來進行偵測被遮蔽之牙結石大多為離體實驗，其多半是利用人眼觀察的方式進行偵測，且由於光學同調斷層掃描(OCT)的結果為能量訊號，因此需要調整至合適的能量強度範圍使光學同調斷層掃描(OCT)的輸出影像中的牙結石明顯，才能進行偵測，使得偵測牙結石過程繁雜，且無法自動進行。

因此，本論文提出一套可以自動偵測體外生物組織遮蔽下之牙結石系統，利用圖像分割卷積神經網路(Segnet)可自動將影像進行分割分類之特性，分割光學同調斷層掃描(OCT)結果，自動將牙結石區域分割並偵測。由於光學同調斷層掃描(OCT)之結果並無法直接使圖像分割卷積神經網路(Segnet)對牙結石進行理想的分割，因此本論文先將光學同調斷層掃描(OCT)結果調整能量強度範圍，產生牙結石明顯之二維影像，並且進行高斯模糊去除雜訊，並在結果影像中選取表面資訊清晰的影像，作為圖像分割卷積神經網路(Segnet)之輸入，產生將被遮蔽之牙結石從影像中進行自動分割的模型，並利用分割結果來完成牙結石區域的偵測。最後，本論文將這套系統進行診斷工具的評測，透過此評測計算系統的偵測準確率。

When the dentist is making a teeth cleaning, since the position of the calculus in gingiva cannot be directly confirmed, all slit between the teeth and between the gingiva will be directly clean. Regardless of whether there is Calculus or not, make such an dental cleaning resulting in discomfort to patient. Optical Coherence Tomography (OCT) can penetrate the gingiva of the human body to detect calculus under the gingiva. However, the technique of detecting calculus in the gingiva is not yet a mature technique. Before the detection on the human body, since the human body experiment has many uncertainties, it is necessary to verify the feasibility of the in vitro experiment that can control the cause. Therefore, the OCT is used to detect the covered calculus in vitro often detected by human eye. Since the result of OCT is an energy signal, it needs to be adjusted to a suitable energy intensity range to make the result of optical coherence tomography to be readable. Therefore, the detection of the calculus is complicated and cannot be performed automatically. This paper proposes a set of detecting system that can automatically detect dental calculus obscured by biological tissue in vitro. The image segmentation convolutional neural network (SegNet) can automatically segment and classify images. Because of the results of OCT can not directly divided by the dental calculus the image SegNet. This paper firstly adjusts the energy intensity range of OCT results, produce a two-dimensional image of the calculus and perform Gaussian blur to remove the noise, and select the image with clear surface information in the resulting image as the
input of SegNet. The calculus is automatically segmented from the image and the segmentation results are used to detect the calculus area. Finally, this paper will evaluate the diagnostic tools of this system, and use this evaluation to calculate the detection accuracy of the system.

[1] Retrived from the World Wide Web: http://www.toothclub.gov.hk/chi/index.html.
[2] L. C. e. a. Huang D and Swanson EA. Optical coherence tomography. Science(New York, NY), 1991.
[3] Masaki Tsubokawa, Akira Aoki, Sho Kakizaki, Yoichi Taniguchi, Kenichiro Ejiri, Koji Mizutani, Greena Koshy, Tatsuya Akizuki, Shigeru Oda, Yasunori Sumi, and Yuichi Izumi. In vitro and clinical evaluation of optical coherence to-mography for the detection of subgingival calculus and root cementum. Journal of Oral Science, 2018.
[4] Vijay Badrinarayanan, Ankur Handa, and Roberto Cipolla. Segnet: A deep convolutional encoder-decoder architecture for robust semantic pixel-wise la-beling. 2015.
[5] Grit Meissner and Thomas Kocher. Calculus-detection technologies and their
clinical application. Periodontology 2000, pages 189-204, 2011.
[6] Avradopoulos V, Wilder RS, Chichester S, and Oenbacher S. Clinical and in ammatory evaluation of perioscopy on patients with chronic periodontitis. J Dent Hyg, pages 30-38, 2004.
[7] Geisinger ML, Mealey BL, Schooleld J, and Mellonig JT. The eectiveness of
subgingival scaling and root planing: an evaluation of therapy with and without the use of the periodontal endoscope. J Periodontol, pages 22-28, 2007.
[8] Michaud RM, Schooleld J, Mellonig JT, and Mealey BL. The ecacy of
subgingival calculus removal with endoscopyaided scaling and root planing: a
study on multirooted teeth. J Periodontol, pages 2238-2245, 2007.
[9] Wilson TG Jr, Carnio J, Schenk R, and Myers G. Absence of histologic signs
of chronic in ammation following closed subgingival scaling and root planing
using the dental endoscope: human biopsies - a pilot study. J Periodontol, pages 36-41, 2008.
[10] Kasaj A, Moschos I, Rohrig B, and Willershausen B. The eectiveness of a novel optical probe in subgingival calculus detection. Int J Dent Hyg, pages 143-147, 2008.
[11] Kasaj A, Sculean A, and Willershausen B. Studie zur uberpru-fung der wirk-
samkeit des detectar-systems hinsichtlich einer verbesserung der compliance under mundhygiene. Parodontologie, pages 343-353, 2004.
[12] Krause F, Braun A, Jepsen S, and Frentzen M. Detection of subgingival calculus with a novel led-based optical probe. J Periodontol, pages 1202-1206, 2005.
[13] Meissner G, Oehme B, Strackeljan J, and Kocher T. Influence of handling-
relevant factors on the behaviour of a novel calculus-detection device. J Clin Periodontol, pages 323-328, 2005.
[14] Meissner G, Oehme B, Strackeljan J, and Kocher T. In vitro calculus detection with a moved smart ultrasonic device. J Clin Periodontol, pages 130-134, 2006.
[15] Meissner G, Oehme B, Strackeljan J, and Kocher T. A new system to detect residual subgingival calculus: in vitro detection limits. J Clin Periodontol,
pages 195-199, 2006.
[16] Meissner G, Oehme B, Strackeljan J, and Kocher T. Clinical subgingival calculus detection with a smart ultrasonic device: a pilot study. J Clin Periodontol, pages 126-132, 2008.
[17] Alfredo Garcia-Layana, Gianfranco Ciuo, Javier Zarranz-Ventura, and Aurora
Alvarez-Vidal. Optical coherence tomography in age-related macular degeneration. Retrived from the World Wide Web: http://amdbook.org/content/optical-coherence-tomography-age-related-macular-degeneration, 2017.
[18] Chih-Cheng Chen. Interactive oct-based tooth scan and reconstruction. 2018.
[19] Yao-Sheng Hsieh, Yi-Ching Ho, Shyh-Yuan Lee, Chih-Wei Lu, Cho-Pei, Cho-Pei Jiang, Ching-Cheng Chuang, Chun-Yang Wang, and Chia-Wei Sun. Sub-gingival calculus imaging based on swept-source optical coherence tomography.2011.
[20] D. Grangier, L. Bottou, and R. Collobert. Deep convolutional networks for scene parsing. In ICML Workshop on Deep Learning, 2009.
[21] Magnus Erik Hvass Pedersen. Convolutional neural network. Retrived from the World Wide Web: https://github.com/Hvass-Labs/TensorFlow-Tutorials/blob/master/02_Convolutional_Neural_Network.ipynb.
[22] Yohei Takada, Tokiko Shiina, Hiroyasu Usami, and Yuji Iwahori M. K. Bhuyan.
Defect detection and classication of electronic circuit boards using keypoint extraction and cnn features. PATTERNS 2017 , The Ninth International Conferences on Pervasive Patterns and Applications, pages 113-116, 2017.
[23] Automated optical inspection. Retrived from the World Wide Web: https://en.wikipedia.org/wiki/Automated_optical_inspection.
[24] D. Soukup and R. Huber-Mork. Convolution neural networks for steel surface defect detection from photometric stereo images. ISCV 2014:Advance in Visual Computing, 2014.
[25] Alex Krizhevsky, Ilya Sutskever, and Georey E. Hinton. Imagenet classication with deep convolutional neural networks. NIPS 2012, 2012.
[26] J.M. Schmitt. Optical coherence tomography (oct): a review. pages 1205-1215, 1999.
[27] A.F.Fercher, C.K.Hitzenberger, G.Kamp, and S.Y.El-Zaiat. Measurement of intraocular distances by backscattering spectral interferometry. In Optics Communications, pages 43-48, 1995.
[28] John Canny. A computational approach to edge detection. IEEE Transactions
on Pattern Analysis and Machine Intelligence, pages 679-698, 1986.
[29] Nobuyuki Otsu. A threshold selection method from gray-level histograms. IEEE
Transactions on Systems, Man, and Cybernetics, pages 62-66, 1979.
[30] Bayes' theorem. Retrived from theWorld WideWeb: https://en.wikipedia.org/wiki/Bayes\%27_theorem.
[31] S. Kakizaki, A. Aoki, M. Tsubokawa, T. Lin, K. Mizutani, G. Koshy, A. Sadr, S. Oda, Y. Sumi, and Y. Izumi. Observation and determination of periodontal tissue prole using optical coherence tomography. Journal of Periodontal
Research, 2017.