[1]景海婷,张秦,陈曼,等.小样本域自适应的皮肤病临床影像识别方法[J].西安交通大学学报,2020,54(09):142-148+156.[doi:10.7652/xjtuxb202009016]
 JING Haiting,ZHANG Qin,CHEN Man,et al.Few-Shot Domain Adaptation for Identification of Clinical Image in Dermatology[J].Journal of Xi'an Jiaotong University,2020,54(09):142-148+156.[doi:10.7652/xjtuxb202009016]
点击复制

小样本域自适应的皮肤病临床影像识别方法
分享到:

《西安交通大学学报》[ISSN:0253-987X/CN:61-1069/T]

卷:
54
期数:
2020年第09期
页码:
142-148+156
栏目:
出版日期:
2020-09-10

文章信息/Info

Title:
Few-Shot Domain Adaptation for Identification of Clinical Image in Dermatology
文章编号:
0253-987X(2020)09-0142-07
作者:
景海婷1 张秦2 陈曼2 张兰3 李政霄2 祝继华1 李钟毓1
1.西安交通大学软件学院, 710049, 西安; 2.西安交通大学第二附属医院, 710004, 西安; 3.西安交通大学第一附属医院东院, 710089, 西安
Author(s):
JING Haiting1 ZHANG Qin2 CHEN Man2 ZHANG Lan3 LI Zhengxiao2 ZHU Jihua1 LI Zhongyu1
1. School of Software Engineering, Xi’an Jiaotong University, Xi’an 710049, China; 2. The Second Affiliated Hospital of Xi’an Jiaotong University, Xi’an 710004, China; 3. East Branch, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an 710089, China
关键词:
卷积神经网络 域自适应 小样本 皮肤病识别 最大相关熵准则
Keywords:
convolutional neural network domain adaption few-shot dermatological recognition maximum correntropy criterion
分类号:
TP38
DOI:
10.7652/xjtuxb202009016
文献标志码:
A
摘要:
针对公开数据集训练所得模型无法直接应用于临床上不同设备的辅助诊断,而临床获取的数据又缺少足够人力进行标注的问题,提出了一种面向皮肤病临床影像识别的小样本域自适应方法。以ISIC皮肤病公开数据集作为标签已知的源域,以实际临床采集的数据作为待识别的目标域,通过医生对极少量临床数据进行标注,建立由卷积神经网络实现的特征提取器和分类器,构建小样本域自适应模型。引入最大相关熵准则来提高识别模型的精度和泛化能力,在每类只有少量带标签目标域样本的情况下,通过交替最大最小化条件熵,在提取区别性特征的同时减小不同域之间的分布差距,提高了分类器在新域上的准确率,实现了模型的跨域迁移。对所提方法在日光性角化病和脂溢性角化病分类问题上进行了实验验证,结果表明:相比于非域自适应方法,所提方法克服了不同采集设备造成的数据分布差异问题,取得了更高的识别准确率; 相比于无监督域自适应方法,所提方法通过加入极少量标注的临床数据实现了域自适应,识别准确率为93.94%。
Abstract:
Aiming at the problem that the model trained by public data set cannot be directly applied to the auxiliary diagnosis of different clinical devices and there is not sufficient manpower to label the clinical data, few-shot domain adaptation for the identification of clinical image in dermatology is proposed. The ISIC dermatological public data set is taken as the source domain with known label, and the actual clinical dataset is taken as the target domain to be predicted, small amount of clinical data are marked by the doctor to train few-shot domain adaptive model of feature extractor and classifier implemented by convolutional neural network. The maximum correntropy criterion is adopted to improve the accuracy and generalization ability of the recognition model. If there are only a small number of labeled target samples in each class, the distribution gap between different domains is reduced while extracting discriminative features by alternating maximum and minimum conditional entropy. The accuracy of the classifier in the new domain is improved, and the model is transferred across domains. The proposed method is experimentally verified in the classification of actinic keratosis and seborrheic keratosis. Compared with the non-domain adaptive method, the proposed method solves the difficulty of difference in data distribution caused by different collection devices and achieves higher recognition accuracy. Compared with the unsupervised domain adaptive method, the proposed method achieves domain adaptation by adding a small amount of labeled clinical data, and the recognition accuracy rate is 93.94%.

参考文献/References:

[1] TSCHANDL P, ROSENDAHL C, KITTLER H. The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions [J]. Scientific Data, 2018, 5(1): 180161.
[2] COMBALIA M, CODELLA N C F, ROTEMBERG V, et al. BCN20000: dermoscopic lesions in the wild [EB/OL]. [2020-02-01]. https:∥arxiv.org/abs/1908.02288.
[3] CODELLA N C F, GUTMAN D, CELEBI M E, et al. Skin lesion analysis toward melanoma detection: a challenge at the 2017 International Symposium on Biomedical Imaging(ISBI), hosted by the International Skin Imaging Collaboration(ISIC)[C]∥2018 IEEE 15th International Symposium on Biomedical Imaging(ISBI). Piscataway, NJ, USA: IEEE, 2018: 168-172.
[4] LIU W F, POKHAREL P P, PRINCIPE J C. Correntropy: properties and applications in non-Gaussian signal processing [J]. IEEE Transactions on Signal Processing, 2007, 55(11): 5286-5298.
[5] SAITO K, KIM D, SCLAROFF S, et al. Semi-supervised domain adaptation via minimax entropy [C]∥2019 IEEE/CVF International Conference on Computer Vision(ICCV). Piscataway, NJ, USA: IEEE, 2019: 8049-8057.
[6] RUSSAKOVSKY O, DENG J, SU H, et al. Imagenet large scale visual recognition challenge [J]. International Journal of Computer Vision, 2015, 115(3): 211-252.
[7] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition [C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway, NJ, USA: IEEE, 2016: 770-778.
[8] DAHLMEIER D, NG H T. Domain adaptation for semantic role labeling in the biomedical domain [J]. Bioinformatics, 2010, 26(8): 1098-1104.
[9] DAI W Y, YANG Q, XUE G R, et al. Boosting for transfer learning [C]∥Proceedings of the 24th International Conference on Machine Learning. Princeton, NJ, USA: International Machine Learning Society(IMLS), 2007: 193-200.
[10] JIANG J, ZHAI C X. Instance weighting for domain adaptation in NLP [C]∥Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics. Stroudsburg, PA, USA: Association for Computational Linguistics(ACL), 2007: 264-271.
[11] DAUMé III H. Frustratingly easy domain adaptation [EB/OL]. [2020-02-01]. https:∥arxiv.org/abs/0907.1815.
[12] SCHWEIKERT G, WIDMER C, SCH?LKOPF B, et al. An empirical analysis of domain adaptation algorithms for genomic sequence analysis [C]∥Proceedings of the 2008 Advances in Neural Information Processing Systems. Vancouver, Canada: NIPS, 2009: 1433-1440.
[13] 缑水平, 焦李成, 姚瑶, 等. 基于迁移核匹配追踪的医学影像诊断系统及诊断方法: CN101609485 [P]. 2009-12-23.
[14] KAMNITSAS K, BAUMGARTNER C, LEDIG C, et al. Unsupervised domain adaptation in brain lesion segmentation with adversarial networks [C]∥International Conference on Information Processing in Medical Imaging. Cham, Germany: Springer, 2017: 597-609.
[15] SHI J, ZHOU S C, LIU X, et al. Stacked deep polynomial network based representation learning for tumor classification with small ultrasound image dataset [J]. Neurocomputing, 2016, 194: 87-94.
[16] MONDAL A K, DOLZ J, DESROSIERS C. Few-shot 3D multi-modal medical image segmentation using generative adversarial learning [EB/OL]. [2020-02-01]. https:∥arxiv.org/abs/1810.12241.
[17] MEDELA A, PICON A, SARATXAGA C L, et al. Few shot learning in histopathological images: reducing the need of labeled data on biological datasets [C]∥2019 IEEE 16th International Symposium on Biomedical Imaging(ISBI). Piscataway, NJ, USA: IEEE, 2019: 1860-1864.
[18] YE D H, POHL K M, DAVATZIKOS C. Semi-supervised pattern classification: application to structural MRI of Alzheimer’s disease [C]∥2011 International Workshop on Pattern Recognition in NeuroImaging. Piscataway, NJ, USA: IEEE, 2011: 1-4.
[19] RAHHAL M M A, BAZI Y, ALHICHRI H, et al. Deep learning approach for active classification of electrocardiogram signals [J]. Information Sciences, 2016, 345: 340-354.
[20] BAI W, OKTAY O, SINCLAIR M, et al. Semi-supervised learning for network-based cardiac MR image segmentation [C]∥2017 International Conference on Medical Image Computing and Computer Assisted Intervention. Cham, Germany: Springer, 2017: 253-260.
[21] CODELLA N, ROTEMBERG V, TSCHANDL P, et al. Skin lesion analysis toward melanoma detection 2018: a challenge hosted by the International Skin Imaging Collaboration(ISIC)[EB/OL]. [2020-02-20]. https:∥arxiv.org/abs/1902.03368.
[22] GIDARIS S, KOMODAKIS N. Dynamic few-shot vis-ual learning without forgetting [C]∥Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2018: 4367-4375.
[23] GANIN Y, USTINOVA E, AJAKAN H, et al. Domain-adversarial training of neural networks [J]. Journal of Machine Learning Research, 2016, 17(1): 189-209.
[24] DAMODARAN B B, KELLENBERGER B, FLAMARY R, et al. DeePJDOT: deep joint distribution optimal transport for unsupervised domain adaptation [C]∥Proceedings of the European Conference on Computer Vision(ECCV). Cham, Germany: Springer, 2018: 447-463.
[本刊相关文献链接]
许美凤,国雷达,宋盼盼,等.多卷积神经网络模型融合的皮肤病识别方法.2019,53(11):125-130.[doi:10.7652/xjtuxb 201911018]
崔云博,杜友田,王航.面向图像的有效目标区域提取方法.2019,53(5):52-57.[doi:10.7652/xjtuxb201905008]
袁帅,秦贵和,晏婕.应用残差生成对抗网络的路况视频帧预测模型.2018,52(10):146-152.[doi:10.7652/xjtuxb201810 020]
张西宁,向宙,唐春华.一种深度卷积自编码网络及其在滚动轴承故障诊断中的应用.2018,52(7):1-8.[doi:10.7652/xjtuxb201807001]
李策,张亚超,蓝天,等.一种高分辨率遥感图像视感知目标检测算法.2018,52(6):9-16.[doi:10.7652/xjtuxb201806 002]

相似文献/References:

[1]李策,张亚超,蓝天,等.一种高分辨率遥感图像视感知目标检测算法[J].西安交通大学学报,2018,52(06):009.[doi:10.7652/xjtuxb201806002]
 LI Ce,ZHANG Yachao,LAN Tian,et al.An Object Detection Algorithm with Visual Perception for HighResolution Remote Sensing Images[J].Journal of Xi'an Jiaotong University,2018,52(09):009.[doi:10.7652/xjtuxb201806002]
[2]张西宁,向宙,唐春华.一种深度卷积自编码网络及其在滚动轴承故障诊断中的应用[J].西安交通大学学报,2018,52(07):001.[doi:10.7652/xjtuxb201807001]
 ZHANG Xining,XIANG Zhou,TANG Chunhua.A Deep Convolutional AutoEncoding Neural Network and Its Application in Bearing Fault Diagnosis[J].Journal of Xi'an Jiaotong University,2018,52(09):001.[doi:10.7652/xjtuxb201807001]
[3]杨一晨,张国和,梁峰,等.一种基于可编程逻辑器件的卷积神经网络协处理器设计[J].西安交通大学学报,2018,52(07):153.[doi:10.7652/xjtuxb201807022]
 YANG Yichen,ZHANG Guohe,LIANG Feng,et al.Design of FPGA Based Convolutional Neural Network CoProcessor[J].Journal of Xi'an Jiaotong University,2018,52(09):153.[doi:10.7652/xjtuxb201807022]
[4]许美凤,国雷达,宋盼盼,等.多卷积神经网络模型融合的皮肤病识别方法[J].西安交通大学学报,2019,53(11):125.[doi:10.7652/xjtuxb201911018]
 XU Meifeng,GUO Leida,SONG Panpan,et al.Skin Disease Recognition Method Based on Multi-Model Fusion of Convolutional Neural Network[J].Journal of Xi'an Jiaotong University,2019,53(09):125.[doi:10.7652/xjtuxb201911018]

备注/Memo

备注/Memo:
收稿日期: 2020-03-15。作者简介: 景海婷(1996—),女,硕士生; 李钟毓(通信作者),男,讲师。基金项目: 国家自然科学基金青年科学基金资助项目(61902310)。
更新日期/Last Update: 2020-09-10