Page 63 - 南京医科大学自然版
P. 63

第46卷第4期          吴邦钰,王志霄,项景轩,等. 基于卷积神经网络的胃癌病理图像分类诊断与分级识别[J].
                  2026年4月                      南京医科大学学报(自然科学版),2026,46(4):520-532                       ·531 ·


                合框架的临床适用性。实验表明,Xception 模型在                            815,825
                分类诊断任务中表现最优(准确率 98.13%),而                        [5] 唐蛰雨,李绍钦,贾中芝. 基于CT图像的深度学习在主
                Stacking 集成框架显著提升了胃腺癌的分级识别精                            动脉夹层中的应用进展[J]. 南京医科大学学报(自然
                度(较单一模型平均提高 3.1%)。两个模型均拥有                              科学版),2024,44(8):1174-1178
                                                                       TANG Z Y,LI S Q,JIA Z Z. Research progress on the ap⁃
                优秀的识别性能,为数字化病理系统的开发提供了
                                                                       plication of deep learning based on CT images in aortic
                理论辅助依据与技术蓝图。
                                                                       dissection[J]. Journal of Nanjing Medical University(Nat⁃
                   利益冲突声明:
                                                                       ural Sciences),2024,44(8):1174⁃1175
                   全体作者声明没有利益冲突。
                                                                 [6] LEE C F,LIN J,HUANG Y L,et al. Deep learning⁃based
                   Conflict of Interests:
                                                                       breast MRI for predicting axillary lymph node metastasis:
                   The authors declared no conflict of interests.
                                                                       a systematic review and meta⁃analysis[J]. Cancer Imag⁃
                   作者贡献声明:
                                                                       ing,2025,25(1):44
                   吴邦钰负责数据收集、模型构建、试验方案与数据分析、
                                                                 [7] WU Y J,MA J,HUANG X S,et al. DeepMMSA:a novel
                论文撰写与修改;王志霄负责代码修改、模型优化、研究方法
                                                                       multimodal deep learning method for non⁃small cell lung
                优化指导、论文修改;项景轩负责论文撰写与修改、文献查
                                                                       cancer survival analysis[C]//2021 IEEE International
                找、资料收集、方案设计、数据分析;孙立负责提供研究方案、
                                                                       Conference on Systems,Man,and Cybernetics(SMC).
                进行技术支持与指导、论文修订;马玲负责提供研究方案与
                                                                       Melbourne,Australia. IEEE,2022:1468-1472
                项目基金支持,进行技术支持与指导、论文修订。
                                                                 [8] TAN M X,LE Q V. EfficientNetV2:smaller models and
                   Author’s Contributions:
                                                                       faster training[C]//International Conference on Machine
                   WU Bangyu was responsible for data collection,model con⁃
                                                                       Learning.,2021
                struction,experiments and data analysis,as well as writing and
                                                                 [9] HOWARD A,SANDLER M,CHEN B,et al. Searching for
                revising the manuscript;WANG Zhixiao was in charge of code
                                                                       MobileNetV3[C]//2019 IEEE/CVF International Confer⁃
                modification,model optimization,and guidance on the optimiza⁃
                                                                       ence on Computer Vision(ICCV). Seoul,Korea. IEEE,
                tion of research methods;XIANG Jingxuan was responsible for
                                                                       2019:1314-1324
                writing and revising the manuscript,literature search,data col⁃
                                                                 [10]WANG W C,AHN E,FENG D G,et al. A review of pre⁃
                lection,scheme design,and data analysis;SUN Li was responsi⁃
                                                                       dictive and contrastive self⁃supervised learning for medi⁃
                ble for providing the research plan,offering technical support
                                                                       cal images[J]. Mach Intell Res,2023,20(4):483-513
                and guidance,and revising the manuscript;MA Ling was respon⁃
                                                                 [11]ZAMANITAJEDDIN N,JAHANIFAR M,XU K,et al.
                sible for providing the research plan and funding support,offer⁃
                                                                       Benchmarking domain generalization algorithms in com⁃
                ing technical support and guidance,and revising the manuscript.
                                                                       putational pathology[EB/OL].[2025⁃09⁃14]. https://arx⁃
               [参考文献]                                                  iv.org/abs/2409.17063
               [1] BRAY F,LAVERSANNE M,SUNG H,et al. Global can⁃  [12]LE VUONG T T,KIM K,SONG B,et al. Joint categorical
                    cer statistics 2022:GLOBOCAN estimates of incidence  and ordinal learning for cancer grading in pathology imag⁃
                    and mortality worldwide for 36 cancers in 185 countries  es[J]. Med Image Anal,2021,73:102206
                   [J]. CA Cancer J Clin,2024,74(3):229-263      [13]MA L Y,SU X F,MA L Y,et al. Deep learning for classi⁃
               [2] HAN B F,ZHENG R S,ZENG H M,et al. Cancer inci⁃      fication and localization of early gastric cancer in endo⁃
                                                                       scopic images[J]. Biomed Signal Process Control,2023,
                    dence and mortality in China,2022[J]. J Natl Cancer
                    Cent,2024,4(1):47-53                               79:104200
               [3] LEI C D,SUN W Q,WANG K,et al. Artificial intelligence⁃  [14]JIN P,JI X Y,KANG W Z,et al. Artificial intelligence in
                    assisted diagnosis of early gastric cancer:present practice  gastric cancer:a systematic review[J]. J Cancer Res Clin
                    and future prospects[J]. Ann Med,2025,57(1):       Oncol,2020,146(9):2339-2350
                    2461679                                      [15]XIA K,HU Y H,CAI S T,et al. GastritisMIL:an interpre⁃
               [4] 徐思雨,张永杰,田         水,等. 深度学习重建算法在胰腺                  table deep learning model for the comprehensive histologi⁃
                    HASTE⁃T2WI序列中的临床应用价值[J]. 南京医科大                     cal assessment of chronic gastritis[J]. Patterns,2025,6
                    学学报(自然科学版),2025,45(6):810-815,825                 (8):101286
                    XU S Y,ZHANG Y J,TIAN S,et al. Clinical application  [16]LAN Q Z,WU Y Z,DING W P,et al. Effective latent hier⁃
                    value of deep learning reconstruction algorithm in pancre⁃  archical feature fusion in multiple instance learning for
                                                                       whole slide image classification[J]. Appl Soft Comput,
                    atic HASTE⁃T2WI sequence[J]. Journal of Nanjing Medi⁃
                    cal University(Natural Sciences),2025,45(6):810-   2025,177:113191
   58   59   60   61   62   63   64   65   66   67   68