宫颈细胞细粒度分类方法

doi:10.3778/j.issn.1002-8331.2009-0146

摘要/Abstract

摘要： 针对宫颈细胞图像的相似性极高，其细粒度分类存在准确率低的问题，提出了一种基于双路径网络与局部判别损失函数的DRMNet（dense reset module net）算法。该算法在特征提取阶段以残差结构为主体，加入密集连接路径，结合两者优点，使网络对特征有着高复用率、低特征冗余度的同时，保持探索新特征的能力。在分类阶段，通过改进损失函数来挖掘图像中的细微特征，利用局部判别损失函数使网络寻找具有判别力的局部区域特征。该算法在Herlev数据集上的七分类准确率达到了98.9%，对比其他算法有一定的提升，从而验证了该算法的有效性。

关键词: 宫颈细胞, 细粒度, 局部判别损失函数, 密集连接

Abstract: Aiming at the high similarity of cervical cell images and the low accuracy of its fine-grained classification, a DRMNet（dense reset module net） algorithm based on dual-path network and local discriminant loss function is proposed. In the feature extraction stage, the residual structure is taken as the main body, and the dense connection path is added. Combining the advantages of the two, the network has high reuse rate and low feature redundancy, while maintaining the ability to explore new features. In the classification stage, the subtle features in the image are mined by improving the loss function, and the local discriminant loss function is used to make the network look for discriminative local area features. The seven classification accuracy rate of this algorithm on the Herlev data set reaches 98.9%, which is a certain improvement compared with other algorithms, which verifies the effectiveness of this algorithm.

Key words: cervical cells, fine-grained, local discriminant loss function, dense connection

苟明亮, 秦明伟, 姚远程. 宫颈细胞细粒度分类方法[J]. 计算机工程与应用, 2022, 58(4): 206-211.

GOU Mingliang, QIN Mingwei, YAO Yuancheng. Cervical Cell Fine-Grained Image Classification Method[J]. Computer Engineering and Applications, 2022, 58(4): 206-211.

参考文献

[1] HUTCHINSON M L，ZAHNISER D J，SHERMAN M E，et al.Utility of liquid-based cytology for cervical carcinoma screening：results of a population-based study conducted in a region of Costa Rica with a high incidence of cervical carcinoma[J].Cancer，1999，87（2）：48-55.
[2] RUSTAGI A S，KAMINENI A，WEINMANN S，et al.Cervical screening and cervical cancer death among older women：a population-based，case-control study[J].American Journal of Epidemiology，2014，179（9）：1107-1114.
[3] SIEGEL R L，MILLER K D，JEMAL A.Cancer statistics[J].CA：A Cancer Journal for Clinicians，2017，67（1）：7-30.
[4] NAYAR R，WILBUR D.The Bethesda system for reporting cervical cytology[M].Heidelberg：Springer International Publishing，2015.
[5] KITCHENER H C，BLANKS R，DUNN G，et al.Automation-assisted versus manual reading of cervical cytology（MAVARIC）：a randomised controlled trial[J].The Lancet Oncology，2011，12（1）：56-64.
[6] MARINAKIS Y，JANTZEN J.Pap smear diagnosis using a hybrid intelligent scheme focusing on genetic algorithm based feature selection and nearest neighbor classification[J].Computers in Biology and Medicine，2009，39（1）：69-78.
[7] PLISSITI M E，NIKOU C.Cervical cell classification based exclusively on nucleus features[C]//Proceedings of the International Conference on Image Analysis and Recognition.Heidelberg：Springer，2012：483-490.
[8] GEN?TAV A，AKSOY S，?NDER S.Unsupervised segmentation and classification of cervical cell images[J].Pattern Recognition，2012，45（12）：4151-4168.
[9] ZHAO L L，LI K，YIN J P，et al.Complete three-phase detection framework for identifying abnormal cervical cells[J].IET Image Processing，2017，11（4）：258-265.
[10] KRIZHEVSKY A，SUTSKEVER I，HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Proceedings of the Advances in Neural Information Processing Systems.Cambridge：MIT Press，2012：1097-1105.
[11] ZHANG L，LU L，NOGUES I，et al.DeepPap：deep convolutional networks for cervical cell classification[J].IEEE Journal of Biomedical and Health Informatics，2017，21（6）：1633-1643.
[12] XU T，ZHANG H，HUANG X L，et al.Multimodal deep learning for cervical dysplasia diagnosis[C]//19th International Conference on Medical Image Computing and Computer-Assisted Intervention.Cham：Springer，2016.
[13] SARWAR A，SURI J，ALI M，et al.Novel benchmark database of digitized and calibrated cervical cells for artificial intelligence based screening of cervical cancer[J].Journal of Ambient Intelligence and Humanized Computing，2016，7（4）：593-606.
[14] TAREEF A，SONG Y，HUANG H，et al.Optimizing the cervix cytological examination based on deep learning and dynamic shape modeling[J].Neurocomputing，2017，248：28-40.
[15] HUANG G，LIU Z，WEINBERGER K Q，et al.Densely connected convolutional networks[J].arXiv：1608.06993，2016.
[16] JANTZEN J，NORUP J，DOUNIAS G，et al.Pap-smear benchmark data for pattern classification[C]//Proceedings of the 1st European Symposium on Nature-Inspired Smart Information Systems，2005.
[17] BORA K，CHOWDHURY M，MAHANTA L B，et al.Automated classification of Pap smear images to detect cervical dysplasia[J].Computer Methods & Programs in Biomedicine，2017，138（1）：31-47.
[18] KIM K B，SONG D H，WOO Y W.Nucleus segmentation and recognition of uterine cervical Pap-smears[C]//Proceedings of the International Workshop on Rough Sets，Fuzzy Sets，Data Mining，and Granular-Soft Computing.Heidelberg：Springer，2007：153-160.