乳癌為全球婦女發生率第1位之癌症，目前可以經由乳房X光攝影檢查偵測出乳房鈣化點或微小腫瘤，發現無症狀的0期乳癌並及早治療，因此應審慎分辨乳房鈣化點是否為因乳癌而呈現的惡性影像。現在已有醫師在著手開發更好的乳癌篩檢方法，並增進乳癌與其他非癌性乳房病灶的鑑別率，例:電腦輔助診斷系統(CADx, computer-aided diagnosis system) 針對影像鈣化點進行判讀並提供醫師第二意見。隨著深度學習網路效能的進步，許多研究提出使用深度學習模型對鈣化區進行分類，本論文提出一個基於深度學習網路的電腦輔助分類系統來更準確的區分乳房X光攝影上的鈣化區域為良性或惡性。
首先，本論文提出了前處理的方法，使用半透明的遮罩疊合 (Maskoverlap)，結合了傳統採用二值化分割以及影像增強的優點，使鈣化點在加強輪廓以及鈣化點形狀的同時能夠保有一定程度的背景資訊供分類做判斷；再者，提出了能夠同時參考鈣化型態及其分佈的作法，稱為自適應多尺度決策融合模組 (AMDF, Adaptively Multi-scale Decision Fusion)，可以有效地運用YOLOv4網路架構中的特徵金字塔架構將微鈣化的局部形態到大範圍鈣化分佈進行綜合性的考量；最終，提出的深度融合模組利用多層感知器 (MLP, Multi-Layer Perceptron) 綜合模型的分類分數以及臨床資訊 (含年齡、醫師對鈣化的描述等)，並且能夠有效提高整體分類表現。在台北榮總提供的微鈣化資料集上進行5次留出法驗證 (Holdout validation) 可以達平均AUC 88.8% 以及準確率84.6%。

Breast cancer has the highest incidence among all diagnosed cancer cases of women worldwide. At present, it is possible to detect breast calcification points or tiny tumors through mammography, finding asymptomatic breast cancer and treating it as soon as possible. Therefore, it is crucial to carefully distinguish whether breast calcification points are malignant images due to breast cancer. Radiologists are now beginning to develop better breast cancer screening methods and improve the discrimination rate between breast cancer and other non-cancerous breast lesions. For example, a computer-aided diagnosis system (CADx) interprets image calcification points and provides radiologists with a second opinion. With advances in deep learning, some methods proposed to further improve classification accuracy. We proposed the deep learning-based computer-aided classification system for better classification performance of microcalcification on mammograms.
We first proposed pre-processing method Maskoverlap, which combines the advantages of using MC segmentation and image enhancement techniques. The proposed Maskoverlap can strengthen the contour and the shape of the MC points while retaining the background information for classification. In addition, we proposed a novel approach named adaptive multi-scale decision fusion module (AMDF-module) based on popular YOLOv4 which can effectively correlate the features of calcifications from local morphology to global distribution with feature pyramid architecture. Finally, the proposed deep ensemble module (DEM) uses a multi-layer perceptron (MLP) to integrate the classification scores of models and additional clinical information, including age, radiologist's descriptions of calcification, etc., and can effectively improve the overall classification performance. In summary, experiments on both public CBIS-DDSM and private VGH-TPE data sets confirm the effectiveness of our method. We achieve an AUC of 88.8 % and an accuracy rate of 84.6% on 5 holdout validations of the VGH-TPE microcalcifications data set.

[1] World Health Organization, "Breast cancer," 26 3 2021. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/breast-cancer.
[2] World Health Organization, “WHO position paper on mammography screening,” WHO Library Cataloguing-in-Publication Data, Switzerland, 2014.
[3] Breastcancer.org, "Understanding Breast Calcifications," 2020. [Online]. Available: https://www.breastcancer.org/symptoms/testing/types/mammograms/mamm_show/calcifications.
[4] Suhail, Zobia and Denton, Erika RE and Zwiggelaar, Reyer, "Tree-based modelling for the classification of mammographic benign and malignant micro-calcification clusters," Multimedia Tools and Applications, vol. 77, no. 5, pp. 6135--6148, 2018.
[5] EA Sickles, CJ d'Orsi, LW Bassett, CM Appleton, WA Berg, ES Burnside, et al., "Acr bi-rads® mammography," in ACR BI-RADS® atlas, breast imaging reporting and data system, 2013, p. 5:2013.
[6] Alam, Nashid and RE Denton, Erika and Zwiggelaar, Reyer, "Classification of microcalcification clusters in digital mammograms using a stack generalization based classifier," Journal of Imaging, vol. 5, no. 9, p. 76, 2019.
[7] Birmohan Singh and Manpreet Kaur, "An approach for classification of malignant and benign microcalcification clusters," Sar{a}dhanar{a}, vol. 43, no. 3, pp. 1-18, 2018.
[8] Krizhevsky, Alex and Sutskever, Ilya and Hinton, Geoffrey E, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems, vol. 25, pp. 1097--1105, 2012.
[9] Simonyan, Karen and Zisserman, Andrew, "Very deep convolutional networks for large-scale image recognition," arXiv preprint arXiv:1409.1556, 2014.
[10] Szegedy, Christian and Liu, Wei and Jia, Yangqing and Sermanet, Pierre and Reed, Scott and Anguelov, Dragomir and Erhan, Dumitru and Vanhoucke, Vincent and Rabinovich, Andrew, "Going deeper with convolutions," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015.
[11] Jinhua Wang, Xi Yang, Hongmin Cai, Wanchang Tan, Cangzheng Jin, and Li Li, "Discrimination of breast cancer with microcalcifications on mammography by deep learning," Scientific reports, vol. 6, no. 1, pp. 1-9, 2016.
[12] Egemen Sert, Seyda Ertekin, and Ugur Halici, "Ensemble of convolutional neural networks for classification of breast microcalcification from mammograms," in In 2017 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society(EMBC), 2017.
[13] Alexey Bochkovskiy, Chien-Yao Wang, and Hong-Yuan Mark Liao, "Yolov4: Optimal speed and accuracy of object detection," arXiv preprint arXiv:2004.10934, 2020.
[14] Hipperson, Victoria L, "Magnification and Compression Views," in Digital Mammography, Springer, 2015, pp. 211-218.
[15] The American Cancer Society medical and editorial content team, "Understanding Your Mammogram Report," American Cancer Society, 2019. [Online]. Available: https://www.cancer.org/cancer/breast-cancer/screening-tests-and-early-detection/mammograms/understanding-your-mammogram-report.html.
[16] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, "Spatial pyramid pooling in deep convolutional networks for visual recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 37, no. 9, pp. 1904-1916, 2015.
[17] Shu Liu, Lu Qi, Haifang Qin, Jianping Shi, and Jiaya Jia, "Path aggregation network for instance segmentation," in In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.
[18] Joseph Redmon and Ali Farhadi, YOLOv3: An incremental improvement, arXiv preprint arXiv:1804.02767, 2018.
[19] Chien-Yao Wang, Hong-Yuan Mark Liao, Yueh-Hua Wu, Ping-Yang Chen, Jun-Wei Hsieh, and I-Hau Yeh, "CSPNet: A new backbone that can enhance learning capability of CNN.," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshop(CVPR Workshop), 2020.
[20] Liu, Songtao and Huang, Di and Wang, Yunhong, "Learning spatial fusion for single-shot object detection," arXiv preprint arXiv:1911.09516, 2019.
[21] Wang, Guangrun and Wang, Keze and Lin, Liang, "Adaptively connected neural networks," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019.
[22] Raschka, Sebastian, "Model evaluation, model selection, and algorithm selection in machine learning," arXiv preprint arXiv:1811.12808, 2018.
[23] Rebecca Sawyer Lee, Francisco Gimenez, Assaf Hoogi, Kanae Kawai Miyake, Mia Gorovoy, and Daniel L Rubin, "A curated mammography data set for use in computer-aided detection and diagnosis research," Scientific data, vol. 4, no. 1, pp. 1-9, 2017.
[24] Michael Heath, Kevin Bowyer, Daniel Kopans, Richard Moore and W. Philip Kegelmeyer, "The Digital Database for Screening Mammography," in Proceedings of the Fifth International Workshop on Digital Mammography, 2001.
[25] Andrik Rampun, Hui Wang, Bryan Scotney, Philip Morrow, and Reyer Zwiggelaar, "Classification of mammographic microcalcification clusters with machine learning confidence levels," in In 14Th International Workshop on Breast Imaging (IWBI 2018), International Society for Optics and Photonics, 2018.
[26] Kim, Tae Kyun, "T test as a parametric statistic," Korean journal of anesthesiology, vol. 68, no. 6, p. 540, 2015.
[27] Unal, Ilker, "Defining an optimal cut-point value in ROC analysis: an alternative approach," Computational and mathematical methods in medicine, vol. 2017, 2017.