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：	張善能 Zhang, Shan-Neng
論文名稱：	利用預訓練深度學習於陰道鏡子宮頸癌分類之前期研究 A Pilot Study of Cervical Intraepithelial Neoplasia Classification from Colposcopy Images Using an AI Pre-trained Method
指導教授：	杜翌群 Du, Yi-Chun 楊子賢 Yang, Tzu-Hsien
學位類別：	碩士 Master
系所名稱：	工學院 - 生物醫學工程學系 Department of BioMedical Engineering
論文出版年：	2024
畢業學年度：	112
語文別：	英文
論文頁數：	60
中文關鍵詞：	子宮頸癌、人工智慧、深度學習、陰道鏡、子宮頸上皮內瘤
外文關鍵詞：	Cervical Cancer, Artificial Intelligence, Deep Learning, Colposcopy, Cervical Intraepithelial Neoplasia
相關次數：	點閱：50 下載：2
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

考慮到台灣女性子宮頸癌的高發病率，增強預防和篩查方法顯得至關重要。由於細胞學檢查與子宮頸切片之間常存在差異，加上手術後傷口復原可能引起的炎症反應會影響抹片的判讀，導致診斷不足的問題變得更加突出。因此，治療後的抹片追蹤變得非常重要。所以本研究旨在針對抹片異常的患者，減少「異常切片的偽陰性」和「不必要的抹片及切片檢查」。過去的研究顯示，傳統抹片的準確率僅為55-80%（平均約75%），而近期使用VGG和ResNet等AI模型的準確率已超過75%，達到與抹片相當的水平，但這些模型較為笨重很難在資源較為匱乏的地區派上用場，所以本研究希望能在邊緣式裝置下達到相同的效果，因此選用了基於ImageNet權重的輕量級MobileNetV2預訓練模型，並在其後加上CBAM 模塊，開發出一個針對子宮頸轉化區類型和子宮頸上皮內瘤變的分類辨別模型。本研究使用了Intel於Kaggle舉辦的陰道鏡子宮頸轉化區種類分類比賽中的資料集進行訓練，模型在僅有1.79M 參數量的情況下，實現了79.52%的準確率和79.46%的Macro-F1分數，並將其作為預訓練模型應用於國立成功大學醫學院附設醫院的子宮頸CIN分類資料中。結果顯示，使用此預訓練的方式在有限的資料和資源下的效果可以與其他較為繁重的模型訓練結果相媲美，達到78.2%的準確率和79.5%的Macro-F1分數。研究結論強調了AI在改進子宮頸癌篩查準確性方面的潛力，特別是在減少偽陰性和不必要的抹片檢測中有顯著貢獻。模型的預測結果令人鼓舞，證明了輕量型模型在準確分類子宮頸病變和幫助識別治療後變化方面的可行性。未來本研究將進一步添加更多的成大醫院病患資料，以鞏固模型在各種臨床環境中的應用性。

Given the high incidence of cervical cancer among women in Taiwan, enhancing prevention and screening methods was crucial. Discrepancies often existed between cytology and cervical biopsy results, and inflammation from post-surgical wound healing could affect smear interpretations, exacerbating the issue of underdiagnosis. Therefore, post-treatment smear follow-up became increasingly important. This study aimed to reduce false negatives in abnormal biopsies and minimize unnecessary smears and biopsies in patients with abnormal smears. Past studies showed that the accuracy of traditional smears ranged from 55% to 80% (averaging around 75%). Recently, AI models like VGG and ResNet achieved accuracy rates exceeding 75%, comparable to smear tests. However, these models were cumbersome and challenging to deploy in resource-limited areas. Thus, this study aimed to achieve similar results on edge devices by selecting a lightweight MobileNetV2 pre-trained model based on ImageNet weights, enhanced with a CBAM module. This model was developed to classify cervical transformation zone types and cervical intraepithelial neoplasia (CIN). The study utilized the dataset from the Intel & MobileODT Cervical Cancer Screening competition on Kaggle for training. The model, with only 1.79 million parameters, achieved 79.52% accuracy and a Macro-F1 score of 79.46%. It was then applied as a pre-trained model to CIN classification data from National Cheng Kung University Hospital, resulting in 78.2% accuracy and a Macro-F1 score of 79.5%. The study concluded that the lightweight model demonstrated comparable performance to heavier models, especially in reducing false negatives and unnecessary smears. The encouraging results underscored the feasibility of using lightweight models for accurately classifying cervical lesions and aiding in post-treatment identification. In the future, this study will incorporate additional patient data from National Cheng Kung University Hospital to further strengthen the model's applicability in various clinical settings.

摘要	I
Abstract	II
誌謝	IV
Contents	V
List of Figures	VII
List of Tables	IX
Chapter 1.	Introduction	1
1-1	Background	1
1-1-1	Cervical Intraepithelial Neoplasia (CIN)	3
1-1-2	Transformation Zone (TZ)	4
1-2	Research Motivation	5
1-3	Research Objective	7
Chapter 2.	Literature Review	8
2-1	The Methods for Colposcopic Image Classification	8
2-1-1	Used Fine-Tune MobileNetV2 in Colposcopic Images	8
2-1-2	Relevant Paper on Data Augmentation	9
2-2	Artificial Intelligence Models and Algorithms	11
2-2-1	Convolutional Block Attention Module (CBAM)	11
2-2-2	MobileNetV2	12
2-2-2-1	Depthwise Separable Convolutions	12
2-2-2-2	Linear Bottlenecks	13
2-2-2-3	Inverted Residuals	14
2-2-3	Developed Pre-Trained Model with Domain-Specific Data	15
Chapter 3.	Materials and Methods	17
3-1	Overview of Datasets Utilized in the Study	17
3-1-1	Introduction to Cervical TZ Types Data set	17
3-1-2	Introduction to CIN Data Set	17
3-2	Ten-Fold Cross-Validation	18
3-3	StratifiedKFold	18
3-4	Data Augmentation	18
3-5	Introduction to the Experimental Procedure	20
3-5-1	Model Architecture Used in This Study	21
Chapter 4.	Experimental Design and Results	25
4-1	Introduction to Model Performance Metrics	25
4-2	Pre-training Method Based on Open-Source Colposcopic Images	28
4-2-1	The Learning Curve of Pre-trained Model Based on Open-Source Colposcopic Images	30
4-2-2	Experimental Results of Pre-trained Model Based on Open-Source Colposcopic Images	31
4-2-2-1	TZ Type 1 Experimental Results	32
4-2-2-2	TZ Type 2 Experimental Results	33
4-2-2-3	TZ Type 3 Experimental Results	34
4-2-2-4	Average Performance and Comparison with Related Studies	36
4-3	Fine-Tuning a Pre-Trained Model Based on Open-Source Colposcopic Images Using NCKUH Colposcopic Images	37
4-3-1	The Learning Curve of Fine-Tuning with NCKUH Images	39
4-3-2	Experimental Results of Fine-Tuning with NCKUH Colposcopic Images	40
4-4	Deployed the Model on Mobile Devices	42
4-5	Comparison of the Model with Other ImageNet Weighted Models	44
Chapter 5.	Discussion and Conclusion	45
References	47
                                    

長庚紀念醫院婦產部, "婦女健康照護網: 長庚婦產通訊第35期," https://www1.cgmh.org.tw/intr/intr2/c4710/contents/m/35-5.htm.
Cleveland Clinic, "Colposcopy," My.clevelandclinic.org, 2024. https://my.clevelandclinic.org/health/diagnostics/4044-colposcopy.
吳香達, "陰道鏡檢查," 台北榮民總醫院, https://wd.vghtpe.gov.tw/obgy/Fpage.action?fid=3793.
International Agency for Research on Cancer, "Colposcopy digital atlas - Cancer screening at IARC," https://screening.iarc.fr/index.php.
D. G. Ferris, Modern Colposcopy: Textbook and Atlas. Dubuque, IA: Kendall Hunt, 2004.
Brookside Associates, "Cervical disease and neoplasia," https://obgynmorningrounds.com/blog6/lectures/gynecology/cervical-disease-and-neoplasia.
亞洲大學附屬醫院, "衛教資訊: 子宮頸上皮內腫瘤", https://www.auh.org.tw/NewsInfo/HealthEducationInfo?docid=1633.
Prendiville, W., & Sankaranarayanan, R. (2017). Colposcopy and treatment of cervical precancer. International Agency for Research on Cancer, World Health Organization.
Intel and MobileODT, "Intel & MobileODT Cervical Cancer Screening," https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/data.
Arbyn, M., Weiderpass, E., Bruni, L., de Sanjosé, S., Saraiya, M., Ferlay, J., & Bray, F. (2020). Estimates of incidence and mortality of cervical cancer in 2018: a worldwide analysis. The Lancet Global Health, 8(2), e191-e203.
Mullapally, S. K., Digumarti, L., & Digumarti, R. (2022). Cervical Cancer in low-and Middle-Income countries: a Multidimensional Approach to closing the gaps. JCO Oncology Practice, 18(6), 423-425.
Singh, G. K., Azuine, R. E., & Siahpush, M. (2012). Global inequalities in cervical cancer incidence and mortality are linked to deprivation, low socioeconomic status, and human development. International Journal of MCH and AIDS, 1(1), 17.
Uho企劃部, "十大癌症發生率最高的竟是乳癌！乳癌病友協會呼籲政府應重視", http://www.tbca-npo.org.tw/List-1.aspx?id=59&sn=1167&nowPage=4&kw=.
辛仲宏, 雷軍強, 郭城, 張亞萍, 陳梓嫻. (2022). 深度學習用於宮頸癌診療研究進展. 中國醫學影像技術, 38(5), 779-782.
Xue, P., Ng, M. T. A., & Qiao, Y. (2020). The challenges of colposcopy for cervical cancer screening in LMICs and solutions by artificial intelligence. BMC medicine, 18(1), 169.
Cho, B. J., Choi, Y. J., Lee, M. J., Kim, J. H., Son, G. H., Park, S. H., ... & Park, S. T. (2020). Classification of cervical neoplasms on colposcopic photography using deep learning. Scientific reports, 10(1), 13652.
Aina, O. E., Adeshina, S. A., Adedigba, A. P., & Aibinu, A. M. (2021). Classification of cervical intraepithelial neoplasia (cin) using fine-tuned convolutional neural networks. Intelligence-Based Medicine, 5, 100031.
Xu, T., Li, P., & Wang, X. (2021, October). Cervical Lesions Classification Based on Pre-trained MobileNet Model. In 2021 IEEE 15th International Conference on Anti-counterfeiting, Security, and Identification (ASID) (pp. 93-96). IEEE.
Shorten, C., & Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of big data, 6(1), 1-48.
Woo, S., Park, J., Lee, J. Y., & Kweon, I. S. (2018). Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision (ECCV) (pp. 3-19).
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4510-4520).
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., ... & Adam, H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
Chitnis, S. R., Liu, S., Dash, T., Verlekar, T. T., Di Ieva, A., Berkovsky, S., ... & Srinivasan, A. (2023, July). Domain-specific pre-training improves confidence in whole slide image classification. In 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) (pp. 1-4). IEEE.
Yuan, C., Yao, Y., Cheng, B., Cheng, Y., Li, Y., Li, Y., ... & Lu, W. (2020). The application of deep learning based diagnostic system to cervical squamous intraepithelial lesions recognition in colposcopy images. Scientific reports, 10(1), 11639.
Peng, G., Dong, H., Liang, T., Li, L., & Liu, J. (2021). Diagnosis of cervical precancerous lesions based on multimodal feature changes. Computers in Biology and Medicine, 130, 104209.
Liu, L., Wang, Y., Liu, X., Han, S., Jia, L., Meng, L., ... & Qiao, X. (2021). Computer-aided diagnostic system based on deep learning for classifying colposcopy images. Annals of Translational Medicine, 9(13).
yeyupiaoling, "yeyupiaoling/ClassificationForAndroid", GitHub, https://github.com/yeyupiaoling/ClassificationForAndroid.
Li, W., Venkataraman, S., Gustafsson, U., Oyama, J. C., Ferris, D. G., & Lieberman, R. W. (2009). Using acetowhite opacity index for detecting cervical intraepithelial neoplasia. Journal of Biomedical Optics, 14(1), 014020-014020.
Sankaranarayanan, R., Rajkumar, R., Theresa, R., Esmy, P. O., Mahe, C., Bagyalakshmi, K. R., ... & Cherian, J. (2004). Initial results from a randomized trial of cervical visual screening in rural south India. International journal of cancer, 109(3), 461-467.
亞洲大學附屬醫院, "衛教資訊: 子宮頸上皮內腫瘤," https://www.auh.org.tw/NewsInfo/HealthEducationInfo?docid=1633.
衛生福利部. "衛生福利部新聞-妳抹了嗎？六分鐘護一生子宮頸癌篩檢有效降低子宮頸癌發生率與死亡率," https://www.mohw.gov.tw/cp-6566-73949-1.html.
Hong, Z., Xiong, J., Yang, H., & Mo, Y. K. (2024). Lightweight Low-Rank Adaptation Vision Transformer Framework for Cervical Cancer Detection and Cervix Type Classification. Bioengineering, 11(5), 468.
Muktiar Singh, J. S. (2019). Automatic Classification of Cervix Type. IRC.
Chen, P., Liu, F., Zhang, J., & Wang, B. (2024). MFEM-CIN: A Lightweight Architecture Combining CNN and Transformer for the Classification of Pre-Cancerous Lesions of the Cervix. IEEE Open Journal of Engineering in Medicine and Biology, 5, 216-225.
Kim, J., Park, C. M., Kim, S. Y., & Cho, A. (2022). Convolutional neural network-based classification of cervical intraepithelial neoplasias using colposcopic image segmentation for acetowhite epithelium. Scientific Reports, 12(1), 17228.
Gwirtz, K., Morzfeld, M., Fournier, A., & Hulot, G. (2021). Can one use Earth’s magnetic axial dipole field intensity to predict reversals?. Geophysical Journal International, 225(1), 277-297.
Loopik, D. L., Bentley, H. A., Eijgenraam, M. N., IntHout, J., Bekkers, R. L., & Bentley, J. R. (2021). The natural history of cervical intraepithelial neoplasia grades 1, 2, and 3: a systematic review and meta-analysis. Journal of Lower Genital Tract Disease, 25(3), 221-231.

校內：立即公開
校外：立即公開

簡易檢索 / 詳目顯示

相關論文