本校學位論文庫
CITYU Theses & Dissertations
論文詳情
刘传
楊雲
數據科學學院
數據科學碩士學位課程(中文學制)
碩士
2024
基於多尺度金字塔融合機制的癌症分類
Cancer classification based on multi-scale pyramid fusion mechanism
數據科學 ; 人工智能 ; 自監督對比學習 ; 計算機病理學
Data science ; Artificial intelligence ; Self supervised contrastive learning ; Computer pathology
公開日期：1/8/2027
根據當今癌症現狀，乳腺癌已經成為全球第一大惡性腫瘤。傳統的病理診斷通過人工閱片的方式，根據需求在不同倍率圖像下分析，具有一定主觀性和重複性。隨著人工智能的快速發展，如何通過改進模型方法準確地對癌症圖像進行分類已經成為數字病理圖像中十分重要的部分。同時隨著醫學圖像科技的不斷進步，臨床醫學家更加關注如何獲取病理圖像中的各種細節信息。目前方法與模型都是在利用多示例學習的基礎上進行改進，通過對陽性和陰影區間進行識別，決定最終的分類結果，然而這些方法仍然存在如下問題：採用流行的深度卷積神經網絡架構對補丁級別的圖像進行特征提取時，沒有進行注意力機制等特征表達關鍵區域的突出處理，所提取的特征表示缺少癌症圖像的細節信息，進而對後續的分類結果產生一定影響；同時弱監督方法可能會受到標籤不準確或噪聲標籤的干擾，這可能會導致模型學習到錯誤的模式或規律，影響模型的性能和泛化能力。
因此在上述問題的基礎上本文致力於研究並採用局部注意力機制和重建損失的增強自監督對比學習與多尺度金字塔融合機制相結合的方法Attention Pyramid Contrastive Learning Representation(AP-CLR)獲取更多細節信息的特征，以提高癌症分類的準確性。對於特征表示獲取過程中進行了改進，採用局部注意力機制的增強對比學習和重建函數，無監督的學習方式避免了標籤不準確的問題；通過局部注意力機制，獲取圖像局部信息也結合了整體信息；重建函數減少了噪聲的干擾；與多尺度金字塔機制相結合，通過將不同尺度下選取的特徵進行融合，進一步有效捕獲影像中的細微結構和全局信息，從而提高分類器的性能優勢。
本文的研究目的是解決乳腺癌圖像分類問題。整個實驗使用常用的乳腺癌圖像數據集CAMELYON-16和BreakHis。在各種自監督對比學習模型中進行實驗，以ResNet18為基礎，通過引入注意力機制，從圖像的不同層次篩選重要信息，提升了模型分類性能；另外，本文對不同注意力機制方法進行實驗對比，選出最佳的注意力機制方法為空間註意力機制；篩選出最佳的金字塔尺度為5×+20×的雙層金字塔特征融合。
最終實驗結果表明，本文所提出的方法在乳腺癌任務中取得了較為不錯的性能提升，評估指標控製在了70%以上，並且相比於其他的模型方法，準確率和AUC也具有一定的性能提升。本研究在乳腺癌圖像分類中取得了良好的效果，為進一步提升醫學影像分類技術提供了有益的參考和借鑒。
因此在上述問題的基礎上本文致力於研究並採用局部注意力機制和重建損失的增強自監督對比學習與多尺度金字塔融合機制相結合的方法Attention Pyramid Contrastive Learning Representation(AP-CLR)獲取更多細節信息的特征，以提高癌症分類的準確性。對於特征表示獲取過程中進行了改進，採用局部注意力機制的增強對比學習和重建函數，無監督的學習方式避免了標籤不準確的問題；通過局部注意力機制，獲取圖像局部信息也結合了整體信息；重建函數減少了噪聲的干擾；與多尺度金字塔機制相結合，通過將不同尺度下選取的特徵進行融合，進一步有效捕獲影像中的細微結構和全局信息，從而提高分類器的性能優勢。
本文的研究目的是解決乳腺癌圖像分類問題。整個實驗使用常用的乳腺癌圖像數據集CAMELYON-16和BreakHis。在各種自監督對比學習模型中進行實驗，以ResNet18為基礎，通過引入注意力機制，從圖像的不同層次篩選重要信息，提升了模型分類性能；另外，本文對不同注意力機制方法進行實驗對比，選出最佳的注意力機制方法為空間註意力機制；篩選出最佳的金字塔尺度為5×+20×的雙層金字塔特征融合。
最終實驗結果表明，本文所提出的方法在乳腺癌任務中取得了較為不錯的性能提升，評估指標控製在了70%以上，並且相比於其他的模型方法，準確率和AUC也具有一定的性能提升。本研究在乳腺癌圖像分類中取得了良好的效果，為進一步提升醫學影像分類技術提供了有益的參考和借鑒。
According to the current status of cancer, breast cancer has become the world's largest malignant tumor. Traditional pathological diagnosis relies on manual film reading and analysis under different magnification images according to needs, which has a certain degree of subjectivity and repeatability. With the rapid development of artificial intelligence, how to accurately classify cancer images through improved model methods has become a very important part of digital pathological images. With the continuous advancement of medical imaging technology, clinical physicians are paying more attention to how to obtain various detailed information in pathological images. At present, methods and models are all improved based on the use of multi instance learning, by identifying positive and shadow intervals to determine the final classification results. However, these methods still have the following problems: when using the popular deep convolutional neural network architecture to extract features from patch level images, there is no attention mechanism or other feature expression for highlighting key areas, and the extracted features lack the detailed information of cancer images, which can have a certain impact on subsequent classification results; At the same time, weakly supervised methods may be affected by inaccurate labels or noisy labels, which may lead to the model learning incorrect patterns or patterns, affecting the performance and generalization ability of the model.
Therefore, based on the above issues, this article is committed to studying and adopting a method combining local attention mechanism and reconstruction loss enhanced self supervised contrastive learning with multi-scale pyramid fusion mechanism, Attention Pyramid Comparative Learning Representation (AP-CLR), to obtain more detailed features and improve the accuracy of cancer classification. Improvements have been made in the process of feature representation acquisition, using enhanced contrastive learning and reconstruction functions with local attention mechanisms, and unsupervised learning to avoid the problem of inaccurate labels; By using local attention mechanism, obtaining local information of the image also combines overall information; The reconstruction function reduces the interference of noise; Combined with the multi-scale pyramid mechanism, by fusing the selected features at different scales, it further effectively captures the subtle structures and global information in the image, thereby improving the performance advantage of the classifier.
The purpose of this paper is to solve the problem of breast cancer image classification. The whole experiment uses the commonly used breast cancer image dataset CAMELYON-16 and BreakHis. Experiments were conducted on various self supervised contrastive learning models, based on ResNet18. By introducing attention mechanisms, important information was filtered from different levels of the image, improving the classification performance of the model; In addition, this article conducts experimental comparisons of different attention mechanism methods and selects the best attention mechanism method as spatial attention mechanism; Select the best double-layer pyramid feature fusion with a pyramid scale of 5×+20×.
The final experimental results show that the method proposed in this paper has achieved a relatively good performance improvement in the breast cancer task, with the evaluation index controlled at more than 70%. Compared with other model methods, the accuracy and AUC also have a certain performance improvement. This study has achieved good results in breast cancer image classification, which provides a useful reference for further improving medical image classification technology.
Therefore, based on the above issues, this article is committed to studying and adopting a method combining local attention mechanism and reconstruction loss enhanced self supervised contrastive learning with multi-scale pyramid fusion mechanism, Attention Pyramid Comparative Learning Representation (AP-CLR), to obtain more detailed features and improve the accuracy of cancer classification. Improvements have been made in the process of feature representation acquisition, using enhanced contrastive learning and reconstruction functions with local attention mechanisms, and unsupervised learning to avoid the problem of inaccurate labels; By using local attention mechanism, obtaining local information of the image also combines overall information; The reconstruction function reduces the interference of noise; Combined with the multi-scale pyramid mechanism, by fusing the selected features at different scales, it further effectively captures the subtle structures and global information in the image, thereby improving the performance advantage of the classifier.
The purpose of this paper is to solve the problem of breast cancer image classification. The whole experiment uses the commonly used breast cancer image dataset CAMELYON-16 and BreakHis. Experiments were conducted on various self supervised contrastive learning models, based on ResNet18. By introducing attention mechanisms, important information was filtered from different levels of the image, improving the classification performance of the model; In addition, this article conducts experimental comparisons of different attention mechanism methods and selects the best attention mechanism method as spatial attention mechanism; Select the best double-layer pyramid feature fusion with a pyramid scale of 5×+20×.
The final experimental results show that the method proposed in this paper has achieved a relatively good performance improvement in the breast cancer task, with the evaluation index controlled at more than 70%. Compared with other model methods, the accuracy and AUC also have a certain performance improvement. This study has achieved good results in breast cancer image classification, which provides a useful reference for further improving medical image classification technology.
2024
中文
53
致 謝 I
摘 要 II
Abstract III
圖目錄 VII
表目錄 VIII
第一章 緒論 1
1.1 研究背景及意義 1
1.2 國內外研究發展 3
1.2.1 乳腺癌研究現狀 3
1.2.2 基於深度卷積神經網絡的癌症圖像分類研究 4
1.2.3 基於遷移學習的癌症圖像分類研究 4
1.2.4 基於自監督對比學習的癌症圖像分類研究 5
1.3 本文主要內容及章節安排 7
第二章 SimCLR與注意力機制等方法介紹 8
2.1 多示例學習 8
2.2 自監督對比學習SimCLR 11
2.3 注意力機制 14
2.4 損失函數 15
2.4.1 對比損失函數 15
2.4.2 重建損失函數 16
2.5 特征金字塔 16
2.6 本章小結 18
第三章 局部注意力特征增强对比学习 19
3.1 問題描述 19
3.2 模型構建 19
3.2.1 引入多視角增強 20
3.2.2 融合局部特征和全局特征 21
3.2.3 聯合對比與重建損失 24
3.2.4 多尺度金字塔融合機制 26
3.3 本章小結 28
第四章 實驗設計與結果分析 29
4.1 實驗數據介紹 29
4.2 實驗過程 30
4.3 實驗結果 36
4.3.1 乳腺癌可視化熱圖結果展示 36
4.3.2 局部注意力特征增强对比学习與重建損失實驗結果 38
4.3.2.1 實驗參數batch_size確定 38
4.3.2.2 注意力機制對比實驗 39
4.3.2.3 傳統模型對比實驗 40
4.3.2.4 基於AP-CLR的對比實驗結果 41
4.4 消融實驗 43
4.4.1 自監督對比學習的消融結果 43
4.4.2 多尺度金字塔的消融結果 44
4.5 本章小結 45
第五章 總結與展望 46
5.1 總結 46
5.2 展望 47
參考文獻 48
作者簡歷 53
摘 要 II
Abstract III
圖目錄 VII
表目錄 VIII
第一章 緒論 1
1.1 研究背景及意義 1
1.2 國內外研究發展 3
1.2.1 乳腺癌研究現狀 3
1.2.2 基於深度卷積神經網絡的癌症圖像分類研究 4
1.2.3 基於遷移學習的癌症圖像分類研究 4
1.2.4 基於自監督對比學習的癌症圖像分類研究 5
1.3 本文主要內容及章節安排 7
第二章 SimCLR與注意力機制等方法介紹 8
2.1 多示例學習 8
2.2 自監督對比學習SimCLR 11
2.3 注意力機制 14
2.4 損失函數 15
2.4.1 對比損失函數 15
2.4.2 重建損失函數 16
2.5 特征金字塔 16
2.6 本章小結 18
第三章 局部注意力特征增强对比学习 19
3.1 問題描述 19
3.2 模型構建 19
3.2.1 引入多視角增強 20
3.2.2 融合局部特征和全局特征 21
3.2.3 聯合對比與重建損失 24
3.2.4 多尺度金字塔融合機制 26
3.3 本章小結 28
第四章 實驗設計與結果分析 29
4.1 實驗數據介紹 29
4.2 實驗過程 30
4.3 實驗結果 36
4.3.1 乳腺癌可視化熱圖結果展示 36
4.3.2 局部注意力特征增强对比学习與重建損失實驗結果 38
4.3.2.1 實驗參數batch_size確定 38
4.3.2.2 注意力機制對比實驗 39
4.3.2.3 傳統模型對比實驗 40
4.3.2.4 基於AP-CLR的對比實驗結果 41
4.4 消融實驗 43
4.4.1 自監督對比學習的消融結果 43
4.4.2 多尺度金字塔的消融結果 44
4.5 本章小結 45
第五章 總結與展望 46
5.1 總結 46
5.2 展望 47
參考文獻 48
作者簡歷 53
[1]盧峰,張開炯,吳立春.機器學習對乳腺癌診斷應用現狀研究進展[J].社區醫學雜誌,2023,21 (24): 1315-1322. DOI:10.19790/j.cnki.JCM.2023.24.11.
[2]尚怡研,譚紅娜.基於MRI影像組學方法預測乳腺癌新輔助治療療效的現狀及進展[J].磁共振成像, 2023, 14 (07): 181-185+191.
[3]楊鑫.基於深度學習的人工智慧在數字病理學中的進展[J].中國癌症雜誌,2021,31(2):151-155.DOI:10.19401/j.cnki.1007-3639.2021.02.010.
[4]武明虎,馬旭茹,劉敏.基於改進的AlexNet的乳腺腫瘤智慧診斷方法[J].湖北工業大學學報, 2020, 35 (04): 1-5.
[5]嶽洋,張維,苗耀鋒.基於自適應調整VGG16模型的乳腺癌風險預測[J].信息技術, 2024, (05): 138-143. DOI:10.13274/j.cnki.hdzj.2024.05.022.
[6]王恒,李霞,劉曉芳,等.基於ResNet50網絡的乳腺癌病理圖像分類研究[J].中國計量大學學報,2019,30(1):72-77. DOI:10.3969/j.issn.2096-2835.2019.01.012.
[7]張會敏.基於深度學習的乳腺癌組織病理學影像分類方法研究[D].太原理工大學, 2023. DOI:10.27352/d.cnki.gylgu.2023.000659.
[8]王一凡,劉靜,馬金剛等.深度學習在乳腺癌影像學檢查中的應用進展[J].計算機科學與探索, 2024, 18 (02): 301-319.
[9]徐斌.基於病理圖像的癌症分析與診斷科技研究[D].電子科技大學, 2023. DOI:10.27005/d.cnki.gdzku.2023.001668.
[10]王茹.基於深度學習的病理學圖像可解釋性分類[D].山西大學, 2023. DOI:10.27284/d.cnki.gsxiu.2023.002147.
[11]焦一平.基於深度學習的數字病理圖像分析[D].東南大學,2021. DOI:10.27014/d.cnki.gdnau.2021.003855.
[12]張冰琰,徐立君,史晋宇等.全球乳腺癌BRCA1/2突變患者地域分佈研究進展[J/OL].基因組學與應用生物學,1-13[2024-06-14].
[13]何思怡,李賀,曹毛毛等.全球及我國女性乳腺癌疾病負擔年齡分佈及變化趨勢[J].中國腫瘤, 2023, 32 (01): 1-7.
[14]杨娜.自檢+篩查，讓乳腺癌無處遁形[N].中國婦女報, 2022-02-23 (008). DOI:10.28067/n.cnki.ncfnb.2022.000485.
[15]陳立.基於改進U-Net的病理圖像分割方法研究[D].陝西科技大學, 2023. DOI:10.27290/d.cnki.gxbqc.2023.000348.
[16]明濤,王丹,郭繼昌等.基於多尺度通道重校準的乳腺癌病理圖像分類[J].浙江大學學報(工學版), 2020, 54 (07): 1289-1297.
[17]程年.基於深度學習的乳腺癌病理圖像分類和分割[D].東南大學, 2019. DOI:10.27014/d.cnki.gdnau.2019.002121.
[18]胡卉.基於遷移學習的女性癌症醫療圖像識別應用研究[D].天津工業大學, 2019.
[19]汪文濤.基於深度學習和遷移學習的乳腺癌組織病理圖像分類方法研究[D].淮北師範大學,2022. DOI:10.27699/d.cnki.ghbmt.2022.000392.
[20]黎英,宋佩華.遷移學習在醫學圖像分類中的研究進展[J].中國圖像圖形學報,2022,27(3):672-686.
[21]劉世博.基於深度學習方法的乳腺病理圖像分類研究[D].杭州電子科技大學, 2023. DOI:10.27075/d.cnki.ghzdc.2023.000413.
[22]李佳盈,蔣文婷,楊林等.基於ViT的細粒度圖像分類[J].計算機工程與設計,2023,44(3):916-921. DOI:10.16208/j.issn1000-7024.2023.03.038.
[23]王國超.醫學影像的深度對比自監督分類算灋研究與應用[D]. 山西:太原科技大學,2023.
[24]周鑫鵬.基於深度學習的問題分類的研究[D].哈爾濱工業大學,2016.
[25]餘凱,賈磊,陳雨强等.深度學習的昨天、今天和明天[J].計算機研究與發展,2013,50(09):1799-1804.
[26]劉一.基於多示例弱監督學習的目標檢測方法研究.西安電子科技大學,2021.
[27]趙璐,袁立明,郝琨.多示例學習算灋綜述.計算機科學(S1),2022:93-99.
[28]劉世峰,王欣.基於對比學習的醫學圖像分類改進方法[J/OL].吉林大學學報(信息科學版), 1-8[2024-06-18].
[29]丁同賀.基於特徵金字塔融合與注意力機制的目標檢測算灋研究[D].山東師範大學, 2023. DOI:10.27280/d.cnki.gsdsu.2023.001823.
[30]李凱,林宇舜,吴晓琳.基於多尺度融合與注意力機制的小目標車輛檢測[J].浙江大學學報,2022, 56 (11): 2241-2250.
[31]吳勝垚,陳星.圖像任務中空間和通道注意力機制研究綜述[J/OL].微電子學與計算機,1-13[2024-05-23].http://kns.cnki.net/kcms/detail/61.1123.tn.20240507.1734.014.html.
[32]李浩鵬,周琬婷,陳玉等.基於域無關迴圈生成對抗網絡的跨模態醫學影像生成[J].數據與計算發展前沿(中英文),2024, 6 (02):80-88.
[33]吳宇杭.基於注意力機制的多尺度融合肝臟器官分割[J].现代计算机,2021,(13):90-96.
[34]石晶晶,周緒川,蔣鳳霞.基於孿生網絡的自監督細粒度分類度量[J].西南民族大學學報（自然科學版）,2023,49(3):297-302. DOI:10.11920/xnmdzk.2023.03.009.
[35]朱雪冬.基於深度卷積網絡的乳腺癌數字化病理圖像的自動診斷研究[D].江蘇:東南大學,2022.
[36]於營,楊婷婷,楊博雄.混淆矩陣分類性能評估及Python實現[J].現代計算機,2021(20):70-73,79. DOI:10.3969/j.issn.1007-1423.2021.20.014.
[37]許月.基於多尺度特徵融合的深度學習目標檢測科技研究[D].安徽工程大學, 2023. DOI:10.27763/d.cnki.gahgc.2023.000274.
[38]Smith, R. A., et al. (2017).Cancer screening in the United States, 2017: A review of current American Cancer Society guidelines and current issues in cancer screening. CA: A Cancer Journal for Clinicians, 67(2), 100-121.
[39]Lu, M. Y., Williamson, D. F. K., Chen, T. Y., Chen, R. J., Barbieri, M., & Mahmood, F. (2021).Data-efficient and weakly supervised computational pathology on whole-slide images. Nature Biomedical Engineering. doi:10.1038/s41551-020-00682-w.
[40]Maximilian Ilse, Jakub Tomczak, Max Welling. Attention-based Deep Multiple Instance Learning[C].In International Conference on Machine Learning,pages 21272136.PMLR,July 2018.ISSN:2640-3498.
[41]Mahmood T ,Arsalan M ,Owais M , et al. Artificial Intelligence-Based Mitosis Detection in Breast Cancer Histopathology Images Using Faster R-CNN and Deep CNNs [J]. Journal of Clinical Medicine, 2020, 9 (3): 749-749.
[42]Xu G ,Ren M ,Li G . Efficient online surface defect detection using multiple instance learning [J]. Expert Systems With Applications, 2024, 252 (PB): 124244-.
[43]Bolei X ,Jingxin L ,Xianxu H , et al. Attention by Selection: A Deep Selective Attention Approach to Breast Cancer Classification [J]. IEEE Transactions on Medical Imaging, 2019, 39 (6): 1-1.
[44]Pawer M M, Pujari S D, Pawar S P, et al. MuSCF-Net: Multi-scale, Multi-Channel Feature Network Using Resnet-based Attention Mechanism for Breast Histopathological Image Classification[A].//Machine Learning and Deep Learning Techniques for Medical Science[M].CRC Press, 2022: 243-261.
[45]K. H ,H. F ,Y. W , et al. Momentum Contrast for Unsupervised Visual Representation Learning [J]. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2020, 9726-9735.
[46]Xinlei Chen, Saining Xie, Kaiming He. An Empirical Study of Training Self-Supervised Vision Transformers.[C], IEEE International Conference on Computer Vision, 2021, abs/2104.02057: 9620-9629.
[47]TING CHEN, SIMON KORNBLITH, MOHAMMAD NOROUZI, et al. A Simple Framework for Contrastive Learning of Visual Representations[C]. //37th International Conference on Machine Learning: ICML 2020, Online, 13-18 July 2020, Part 3 of 15.:Curran Associates, Inc., 2021:1597-1607.
[48]]Li Bin,Li Yin & Eliceiri Kevin W.(2021).Dual-stream Multiple Instance Learning Network for Whole Slide Image Classification with Self-supervised Contrastive Learning.. Conference on Computer Vision and Pattern Recognition Workshops. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Workshops. doi:10.1109/CVPR46437.2021.01409.
[49]Łukasz Struski,Szymon Janusz,Jacek Tabor. Multiple instance learning for medical image classification based on instance importance [J]. Biomedical Signal Processing and Control, 2024, 91 105874-.
[50]Songpol B ,Natdanai T ,Krisna P , et al. Self-supervised learning advanced plant disease image classification with SimCLR [J]. Advances in Computational Intelligence, 2023, 3 (5).
[51]Kanavati Fahdi,Toyokawa Gouji,Momosaki Seiya.Tsuneki Masayuki.(2020).Weakly-supervised learning for lung carcinoma classification using deep learning.. Scientific reports(1). doi:10.1038/s41598-020-66333-x.
[52]Zhang Y ,Liu C ,Liu M , et al. Attention is all you need: utilizing attention in AI-enabled drug discovery. [J]. Briefings in bioinformatics, 2023, 25 (1):
[53]Liu S ,Lin W ,Wang Y , et al. Convolutional Neural Network-Based Bidirectional Gated Recurrent Unit–Additive Attention Mechanism Hybrid Deep Neural Networks for Short-Term Traffic Flow Prediction [J]. Sustainability, 2024, 16 (5):
[54]Tian M ,Wanwan W ,Yu C . Attention is all you need: An interpretable transformer-based asset allocation approach [J]. International Review of Financial Analysis, 2023, 90
[55]Hongshi O ,Jifeng S ,Shengzong Z . Multi-scale spatialtemporal information deep fusion network with temporal pyramid mechanism for video action recognition [J]. Journal of Intelligent & Fuzzy Systems, 2021, 41 (3): 4533-4545.
[56]Dang Jin,Tang Xiaofen & Li Shuai.(2023).HA-FPN: Hierarchical Attention Feature Pyramid Network for Object Detection.. Sensors (Basel, Switzerland)(9). doi:10.3390/S23094508.
[57]Lu Ming Y.,Chen Tiffany Y.,Williamson Drew F.K.,Zhao Melissa,Shady Maha,Lipkova Jana & Mahmood Faisal.(2022).AI-BASED PATHOLOGY PREDICTS ORIGINS FOR CANCERS OF UNKNOWN PRIMARY.Journal of Pathology Informatics. doi:10.1016/J.JPI.2022.100045.
[58]ANTOINE PIROVANO, HIPPOLYTE HEUBERGER, SYLVAIN BERLEMONT, et al. Improving Interpretability for Computer-Aided Diagnosis Tools on Whole Slide Imaging with Multiple Instance Learning and Gradient-Based Explanations[C]. //Interpretable and annotation-efficient learning for medical image computing :.:Springer,, 2020:43-53.
[59]Bo P . Classification of images using EfficientNet CNN model with convolutional block attention module (CBAM) and spatial group-wise enhance module (SGE)[C]// Beijing Jiaotong Univ. (China), 2022:
[60]RAVINDRA YADAV, ASHISH SARDANA, VINAY P NAMBOODIRI, et al. Speech Prediction in Silent Videos Using Variational Autoencoders[C]. //ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing, Vol.11: Toronto, Ontario, Canada.6-11 June 2021.:IEEE, 2021:7048-7052.
[61]CHEN, SI-BAO, ZHENG, YU-JIE, DING, CHRIS H. Q., et al. SIECP: Neural Network Channel Pruning based on Sequential Interval Estimation[J]. 2022,481(Apr.7):1-10. DOI:10.1016/j.neucom.2022.01.053.
[62]Toma A T ,Biswas S ,Miah S M , et al. Breast Cancer Detection Based on Simplified Deep Learning Technique With Histopathological Image Using BreaKHis Database [J]. Radio Science, 2023, 58 (11):
[63]QIAO NING, XIAOWEI ZHAO, ZHIQIANG MA. A Novel Method for Identification of Glutarylation Sites Combining Borderline-SMOTE With Tomek Links Technique in Imbalanced Data[J]. IEEE/ACM transactions on computational biology and bioinformatics,2022,19(5):2632-2641. DOI:10.1109/TCBB.2021.3095482.
[64]CAMPANELLA,GABRIELE,HANNA,MATTHEW G.,GENESLAW, LUKE, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images[J]. Nature medicine,2019,25(8):1301-1309. DOI:10.1038/s41591-019-0508-1.
[65]Jiahao F ,Ce L ,Jin W . CAM-TMIL: A Weakly-Supervised Segmentation Framework for Histopathology based on CAMs and MIL [J]. Journal of Physics: Conference Series, 2023, 2547 (1):
[66]RAMPRASAATH R. SELVARAJU, MICHAEL COGSWELL, ABHISHEK DAS, et al. Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization[C]. //2017 IEEE International Conference on Computer Vision: ICCV 2017, Venice, Italy, 22-29 October 2017, pages 1-744, [v.1].:Institute of Electrical and Electronics Engineers, 2017:618-626.
[67]MORID, MOHAMMAD AMIN, BORJALI, ALIREZA, DEL FIOL, GUILHERME. A scoping review of transfer learning research on medical image analysis using ImageNet[J]. Computers in Biology and Medicine,2021,128. DOI:10.1016/j.compbiomed.2020.104115.
[2]尚怡研,譚紅娜.基於MRI影像組學方法預測乳腺癌新輔助治療療效的現狀及進展[J].磁共振成像, 2023, 14 (07): 181-185+191.
[3]楊鑫.基於深度學習的人工智慧在數字病理學中的進展[J].中國癌症雜誌,2021,31(2):151-155.DOI:10.19401/j.cnki.1007-3639.2021.02.010.
[4]武明虎,馬旭茹,劉敏.基於改進的AlexNet的乳腺腫瘤智慧診斷方法[J].湖北工業大學學報, 2020, 35 (04): 1-5.
[5]嶽洋,張維,苗耀鋒.基於自適應調整VGG16模型的乳腺癌風險預測[J].信息技術, 2024, (05): 138-143. DOI:10.13274/j.cnki.hdzj.2024.05.022.
[6]王恒,李霞,劉曉芳,等.基於ResNet50網絡的乳腺癌病理圖像分類研究[J].中國計量大學學報,2019,30(1):72-77. DOI:10.3969/j.issn.2096-2835.2019.01.012.
[7]張會敏.基於深度學習的乳腺癌組織病理學影像分類方法研究[D].太原理工大學, 2023. DOI:10.27352/d.cnki.gylgu.2023.000659.
[8]王一凡,劉靜,馬金剛等.深度學習在乳腺癌影像學檢查中的應用進展[J].計算機科學與探索, 2024, 18 (02): 301-319.
[9]徐斌.基於病理圖像的癌症分析與診斷科技研究[D].電子科技大學, 2023. DOI:10.27005/d.cnki.gdzku.2023.001668.
[10]王茹.基於深度學習的病理學圖像可解釋性分類[D].山西大學, 2023. DOI:10.27284/d.cnki.gsxiu.2023.002147.
[11]焦一平.基於深度學習的數字病理圖像分析[D].東南大學,2021. DOI:10.27014/d.cnki.gdnau.2021.003855.
[12]張冰琰,徐立君,史晋宇等.全球乳腺癌BRCA1/2突變患者地域分佈研究進展[J/OL].基因組學與應用生物學,1-13[2024-06-14].
[13]何思怡,李賀,曹毛毛等.全球及我國女性乳腺癌疾病負擔年齡分佈及變化趨勢[J].中國腫瘤, 2023, 32 (01): 1-7.
[14]杨娜.自檢+篩查，讓乳腺癌無處遁形[N].中國婦女報, 2022-02-23 (008). DOI:10.28067/n.cnki.ncfnb.2022.000485.
[15]陳立.基於改進U-Net的病理圖像分割方法研究[D].陝西科技大學, 2023. DOI:10.27290/d.cnki.gxbqc.2023.000348.
[16]明濤,王丹,郭繼昌等.基於多尺度通道重校準的乳腺癌病理圖像分類[J].浙江大學學報(工學版), 2020, 54 (07): 1289-1297.
[17]程年.基於深度學習的乳腺癌病理圖像分類和分割[D].東南大學, 2019. DOI:10.27014/d.cnki.gdnau.2019.002121.
[18]胡卉.基於遷移學習的女性癌症醫療圖像識別應用研究[D].天津工業大學, 2019.
[19]汪文濤.基於深度學習和遷移學習的乳腺癌組織病理圖像分類方法研究[D].淮北師範大學,2022. DOI:10.27699/d.cnki.ghbmt.2022.000392.
[20]黎英,宋佩華.遷移學習在醫學圖像分類中的研究進展[J].中國圖像圖形學報,2022,27(3):672-686.
[21]劉世博.基於深度學習方法的乳腺病理圖像分類研究[D].杭州電子科技大學, 2023. DOI:10.27075/d.cnki.ghzdc.2023.000413.
[22]李佳盈,蔣文婷,楊林等.基於ViT的細粒度圖像分類[J].計算機工程與設計,2023,44(3):916-921. DOI:10.16208/j.issn1000-7024.2023.03.038.
[23]王國超.醫學影像的深度對比自監督分類算灋研究與應用[D]. 山西:太原科技大學,2023.
[24]周鑫鵬.基於深度學習的問題分類的研究[D].哈爾濱工業大學,2016.
[25]餘凱,賈磊,陳雨强等.深度學習的昨天、今天和明天[J].計算機研究與發展,2013,50(09):1799-1804.
[26]劉一.基於多示例弱監督學習的目標檢測方法研究.西安電子科技大學,2021.
[27]趙璐,袁立明,郝琨.多示例學習算灋綜述.計算機科學(S1),2022:93-99.
[28]劉世峰,王欣.基於對比學習的醫學圖像分類改進方法[J/OL].吉林大學學報(信息科學版), 1-8[2024-06-18].
[29]丁同賀.基於特徵金字塔融合與注意力機制的目標檢測算灋研究[D].山東師範大學, 2023. DOI:10.27280/d.cnki.gsdsu.2023.001823.
[30]李凱,林宇舜,吴晓琳.基於多尺度融合與注意力機制的小目標車輛檢測[J].浙江大學學報,2022, 56 (11): 2241-2250.
[31]吳勝垚,陳星.圖像任務中空間和通道注意力機制研究綜述[J/OL].微電子學與計算機,1-13[2024-05-23].http://kns.cnki.net/kcms/detail/61.1123.tn.20240507.1734.014.html.
[32]李浩鵬,周琬婷,陳玉等.基於域無關迴圈生成對抗網絡的跨模態醫學影像生成[J].數據與計算發展前沿(中英文),2024, 6 (02):80-88.
[33]吳宇杭.基於注意力機制的多尺度融合肝臟器官分割[J].现代计算机,2021,(13):90-96.
[34]石晶晶,周緒川,蔣鳳霞.基於孿生網絡的自監督細粒度分類度量[J].西南民族大學學報（自然科學版）,2023,49(3):297-302. DOI:10.11920/xnmdzk.2023.03.009.
[35]朱雪冬.基於深度卷積網絡的乳腺癌數字化病理圖像的自動診斷研究[D].江蘇:東南大學,2022.
[36]於營,楊婷婷,楊博雄.混淆矩陣分類性能評估及Python實現[J].現代計算機,2021(20):70-73,79. DOI:10.3969/j.issn.1007-1423.2021.20.014.
[37]許月.基於多尺度特徵融合的深度學習目標檢測科技研究[D].安徽工程大學, 2023. DOI:10.27763/d.cnki.gahgc.2023.000274.
[38]Smith, R. A., et al. (2017).Cancer screening in the United States, 2017: A review of current American Cancer Society guidelines and current issues in cancer screening. CA: A Cancer Journal for Clinicians, 67(2), 100-121.
[39]Lu, M. Y., Williamson, D. F. K., Chen, T. Y., Chen, R. J., Barbieri, M., & Mahmood, F. (2021).Data-efficient and weakly supervised computational pathology on whole-slide images. Nature Biomedical Engineering. doi:10.1038/s41551-020-00682-w.
[40]Maximilian Ilse, Jakub Tomczak, Max Welling. Attention-based Deep Multiple Instance Learning[C].In International Conference on Machine Learning,pages 21272136.PMLR,July 2018.ISSN:2640-3498.
[41]Mahmood T ,Arsalan M ,Owais M , et al. Artificial Intelligence-Based Mitosis Detection in Breast Cancer Histopathology Images Using Faster R-CNN and Deep CNNs [J]. Journal of Clinical Medicine, 2020, 9 (3): 749-749.
[42]Xu G ,Ren M ,Li G . Efficient online surface defect detection using multiple instance learning [J]. Expert Systems With Applications, 2024, 252 (PB): 124244-.
[43]Bolei X ,Jingxin L ,Xianxu H , et al. Attention by Selection: A Deep Selective Attention Approach to Breast Cancer Classification [J]. IEEE Transactions on Medical Imaging, 2019, 39 (6): 1-1.
[44]Pawer M M, Pujari S D, Pawar S P, et al. MuSCF-Net: Multi-scale, Multi-Channel Feature Network Using Resnet-based Attention Mechanism for Breast Histopathological Image Classification[A].//Machine Learning and Deep Learning Techniques for Medical Science[M].CRC Press, 2022: 243-261.
[45]K. H ,H. F ,Y. W , et al. Momentum Contrast for Unsupervised Visual Representation Learning [J]. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2020, 9726-9735.
[46]Xinlei Chen, Saining Xie, Kaiming He. An Empirical Study of Training Self-Supervised Vision Transformers.[C], IEEE International Conference on Computer Vision, 2021, abs/2104.02057: 9620-9629.
[47]TING CHEN, SIMON KORNBLITH, MOHAMMAD NOROUZI, et al. A Simple Framework for Contrastive Learning of Visual Representations[C]. //37th International Conference on Machine Learning: ICML 2020, Online, 13-18 July 2020, Part 3 of 15.:Curran Associates, Inc., 2021:1597-1607.
[48]]Li Bin,Li Yin & Eliceiri Kevin W.(2021).Dual-stream Multiple Instance Learning Network for Whole Slide Image Classification with Self-supervised Contrastive Learning.. Conference on Computer Vision and Pattern Recognition Workshops. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Workshops. doi:10.1109/CVPR46437.2021.01409.
[49]Łukasz Struski,Szymon Janusz,Jacek Tabor. Multiple instance learning for medical image classification based on instance importance [J]. Biomedical Signal Processing and Control, 2024, 91 105874-.
[50]Songpol B ,Natdanai T ,Krisna P , et al. Self-supervised learning advanced plant disease image classification with SimCLR [J]. Advances in Computational Intelligence, 2023, 3 (5).
[51]Kanavati Fahdi,Toyokawa Gouji,Momosaki Seiya.Tsuneki Masayuki.(2020).Weakly-supervised learning for lung carcinoma classification using deep learning.. Scientific reports(1). doi:10.1038/s41598-020-66333-x.
[52]Zhang Y ,Liu C ,Liu M , et al. Attention is all you need: utilizing attention in AI-enabled drug discovery. [J]. Briefings in bioinformatics, 2023, 25 (1):
[53]Liu S ,Lin W ,Wang Y , et al. Convolutional Neural Network-Based Bidirectional Gated Recurrent Unit–Additive Attention Mechanism Hybrid Deep Neural Networks for Short-Term Traffic Flow Prediction [J]. Sustainability, 2024, 16 (5):
[54]Tian M ,Wanwan W ,Yu C . Attention is all you need: An interpretable transformer-based asset allocation approach [J]. International Review of Financial Analysis, 2023, 90
[55]Hongshi O ,Jifeng S ,Shengzong Z . Multi-scale spatialtemporal information deep fusion network with temporal pyramid mechanism for video action recognition [J]. Journal of Intelligent & Fuzzy Systems, 2021, 41 (3): 4533-4545.
[56]Dang Jin,Tang Xiaofen & Li Shuai.(2023).HA-FPN: Hierarchical Attention Feature Pyramid Network for Object Detection.. Sensors (Basel, Switzerland)(9). doi:10.3390/S23094508.
[57]Lu Ming Y.,Chen Tiffany Y.,Williamson Drew F.K.,Zhao Melissa,Shady Maha,Lipkova Jana & Mahmood Faisal.(2022).AI-BASED PATHOLOGY PREDICTS ORIGINS FOR CANCERS OF UNKNOWN PRIMARY.Journal of Pathology Informatics. doi:10.1016/J.JPI.2022.100045.
[58]ANTOINE PIROVANO, HIPPOLYTE HEUBERGER, SYLVAIN BERLEMONT, et al. Improving Interpretability for Computer-Aided Diagnosis Tools on Whole Slide Imaging with Multiple Instance Learning and Gradient-Based Explanations[C]. //Interpretable and annotation-efficient learning for medical image computing :.:Springer,, 2020:43-53.
[59]Bo P . Classification of images using EfficientNet CNN model with convolutional block attention module (CBAM) and spatial group-wise enhance module (SGE)[C]// Beijing Jiaotong Univ. (China), 2022:
[60]RAVINDRA YADAV, ASHISH SARDANA, VINAY P NAMBOODIRI, et al. Speech Prediction in Silent Videos Using Variational Autoencoders[C]. //ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing, Vol.11: Toronto, Ontario, Canada.6-11 June 2021.:IEEE, 2021:7048-7052.
[61]CHEN, SI-BAO, ZHENG, YU-JIE, DING, CHRIS H. Q., et al. SIECP: Neural Network Channel Pruning based on Sequential Interval Estimation[J]. 2022,481(Apr.7):1-10. DOI:10.1016/j.neucom.2022.01.053.
[62]Toma A T ,Biswas S ,Miah S M , et al. Breast Cancer Detection Based on Simplified Deep Learning Technique With Histopathological Image Using BreaKHis Database [J]. Radio Science, 2023, 58 (11):
[63]QIAO NING, XIAOWEI ZHAO, ZHIQIANG MA. A Novel Method for Identification of Glutarylation Sites Combining Borderline-SMOTE With Tomek Links Technique in Imbalanced Data[J]. IEEE/ACM transactions on computational biology and bioinformatics,2022,19(5):2632-2641. DOI:10.1109/TCBB.2021.3095482.
[64]CAMPANELLA,GABRIELE,HANNA,MATTHEW G.,GENESLAW, LUKE, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images[J]. Nature medicine,2019,25(8):1301-1309. DOI:10.1038/s41591-019-0508-1.
[65]Jiahao F ,Ce L ,Jin W . CAM-TMIL: A Weakly-Supervised Segmentation Framework for Histopathology based on CAMs and MIL [J]. Journal of Physics: Conference Series, 2023, 2547 (1):
[66]RAMPRASAATH R. SELVARAJU, MICHAEL COGSWELL, ABHISHEK DAS, et al. Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization[C]. //2017 IEEE International Conference on Computer Vision: ICCV 2017, Venice, Italy, 22-29 October 2017, pages 1-744, [v.1].:Institute of Electrical and Electronics Engineers, 2017:618-626.
[67]MORID, MOHAMMAD AMIN, BORJALI, ALIREZA, DEL FIOL, GUILHERME. A scoping review of transfer learning research on medical image analysis using ImageNet[J]. Computers in Biology and Medicine,2021,128. DOI:10.1016/j.compbiomed.2020.104115.
