基于深度学习技术的湖北钉螺视觉智能识别模型效能评价

doi:10.12140/j.issn.1000-7423.2021.06.006

中国寄生虫学与寄生虫病杂志 ›› 2021, Vol. 39 ›› Issue (6): 764-770.doi: 10.12140/j.issn.1000-7423.2021.06.006

基于深度学习技术的湖北钉螺视觉智能识别模型效能评价

施亮¹(), 熊春蓉¹, 刘毛毛², 魏秀参³, 张键锋¹, 王鑫瑶¹, 王涛¹, 杭德荣¹, 羊海涛¹, 杨坤¹^,²^,^*()

1 江苏省血吸虫病防治研究所,国家卫生和计划生育委员会寄生虫病预防与控制技术重点实验室,江苏省寄生虫与媒介控制技术重点实验室,江南大学公共卫生研究中心,无锡 214064
2 南京医科大学公共卫生学院,南京 211166
3 南京理工大学计算机科学与工程学院,高维信息智能感知与系统教育部重点实验,江苏省社会安全图像与视频理解重点实验室,南京 210094

收稿日期:2021-06-28 修回日期:2021-07-28 出版日期:2021-12-30 发布日期:2021-12-06
通讯作者: 杨坤
作者简介:施亮（1985-）,男,硕士,主管医师,从事空间流行病学与机器学习。E-mail： jipd1950sl@163.com
基金资助:
国家自然科学基金(82173586);江苏省国际科技合作项目(BZ2020003);江苏省省属公益院所能力提升项目(BM2018020-3);江苏省卫生健康委医学科研项目(M2021102);无锡市卫生健康委科研项目(M202121);江南大学公共卫生研究中心课题(JUPH201837);江南大学公共卫生研究中心课题(JUPH202008)

Evaluation of efficacy of visual intelligent recognition model for Oncomelania hupensis based on deep learning technology

SHI Liang¹(), XIONG Chun-rong¹, LIU Mao-mao², WEI Xiu-shen³, ZHANG Jian-feng¹, WANG Xin-yao¹, WANG Tao¹, HANG De-rong¹, YANG Hai-tao¹, YANG Kun¹^,²^,^*()

1 Key Laboratory of National Health Commission on Parasitic Disease Control and Prevention, Jiangsu Provincial Key Laboratory on Parasite and Vector Control Technology, Public Health Research Center of Jiangnan University, Jiangsu Institute of Parasitic Diseases, Wuxi 214064, China
2 School of Public Health, Nanjing Medical University, Nanjing 211166, China
3 Key Lab of Intelligent Perception and Systems for High-Dimensional Information of Ministry of Education, Jiangsu Key Lab of Image and Video Understanding for Social Security, School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China

Received:2021-06-28 Revised:2021-07-28 Online:2021-12-30 Published:2021-12-06
Contact: YANG Kun
Supported by:
National Natural Science Foundation of China(82173586);Jiangsu International Science and Technology Cooperation Project(BZ2020003);Capacity Enhancement Project of Jiangsu Provincial Public Welfare Institutes(BM2018020-3);Medical Research Project of Jiangsu Provincial Health Commission(M2021102);Medical Research Project of Jiangsu Provincial Health Commission(M202121);Public Health Research Center of Jiangnan University(JUPH201837);Public Health Research Center of Jiangnan University(JUPH202008)

摘要/Abstract

摘要：

目的探讨基于深度学习技术建立日本血吸虫中间宿主湖北钉螺视觉智能识别模型并评价其分类识别效能。方法 2019年3月—2020年10月,根据江苏省湖沼型、山丘型和水网型地区钉螺分布规律,选择南京、镇江、扬州、苏州、常州、无锡和盐城等7个地区为样品采集现场,每个地区随机选择3个钉螺孳生环境采集螺样及螺样图像,由6名血吸虫病防治专家根据图像质量与螺类特征进行筛选、分类后建立螺类图像分类标准数据集,并将该结果作为金标准,建立的数据集按照7 ∶ 3原则划分为训练集和测试集（分内部和外部测试集）。采用MobilenetV2、ResNet50、Inception-ResNet-V2等3种主流卷积神经网络模型在训练集完成模型训练;以分类标准数据集作为金标准,绘制受试者工作特征（ROC）曲线,比较3种模型在内部测试集上的灵敏度、特异性、准确率、诊断一致率（Kappa值）和约登指数;选出最优识别模型再与20名查螺工作人员对外部测试集判断结果进行对比, 评价模型在内部测试集和外部测试集中识别钉螺结果的准确性。结果共采集钉螺和方格短沟蜷、细钻螺、真管螺、拟钉螺等4种与钉螺相似的螺类图像3 224幅,经过筛选后,螺类图像分类标准数据集共纳入2 719幅图像,涵盖钉螺图像774幅和4种相似螺类的非钉螺图像1 945幅。其中,训练集占70.2%（1 910/2 719）,测试集占29.8%（749/2 719）。内部测试中,MobilenetV2、ResNet50和Inception-ResNet-V2模型识别钉螺与金标准一致性Kappa值分别为0.78、0.83、0.88;Inception-ResNet-V2模型的灵敏度、特异性、准确率、约登指数和ROC曲线下面积（AUC）均最高,分别为92.00%、97.16%、96.13%、0.89、0.95;3种模型识别的灵敏度和特异性差异无统计学意义（χ² = 3.892、4.948,P > 0.05）,准确率差异有统计学意义（χ² = 8.607,P < 0.05）。外部测试中,最优模型Inception-ResNet-V2模型识别和工作人员鉴别与金标准一致性较好,Kappa值分别为0.80和0.83,AUC分别为0.88和0.92,差异无统计学意义（P > 0.05）。结论基于深度学习技术的钉螺视觉智能识别模型具有良好的准确性。

关键词: 湖北钉螺, 智能识别, 深度学习, 计算机视觉, 机器学习, 血吸虫病, 监测

Abstract:

Objective To investigate the performance of a deep learning-based visual intelligent recognition model for the intermediate host of Schistosoma japonicum, the snail Oncomelania hupensis, and evaluate its efficacy in recognition and classification. Methods According to the distribution pattern of O. hupensis in the topographic type of lake marsh, hill and water network in Jiangsu Province, seven regions including Nanjing, Zhenjiang, Yangzhou, Suzhou, Changzhou, Wuxi and Yancheng were selected as the fields for sample collection from March 2019 to October 2020. From each of the fields, 3 snail breeding environment sites were randomly selected for collection of snail samples and images by smart phone. The image quality and morphologic features of the sampled snails were screened and classified by six experts of schistosomiasis control to establish a standard data set of snail image classification. The data set was compiled into training set and test set (comprised of internal and external test). Three major convolutional neural network models including MobilenetV2, ResNet50 and Inception-ResNet-V2 were used to exercise the model training in the training data set. Taking the standard snail image classification data set as the gold standard, the receiver operating characteristic (ROC) curves was applied to compare the sensitivity, specificity, diagnostic concordance rate (Kappa value), and Yorden index of three model types with the internal test data; comparison was performed between the best-fit recognition model and the findings of snail search staff through the external test set, evaluating the accuracy of models in snail recognition with the internal and external test set. Results Totally, 3 224 images of 4 types of snails similar to O. hupensis were collected: Semisulcospira cancellata, Opeas gracile, Euphaedus, and Tricula. After screening, 2 719 images were included into the standard snail image classification data set, among them 774 were of O. hupensis and 1 945 of 4 being similar to but not O. hupensis. The concordance Kappa value from the snail recognition models of MobilenetV2, ResNet50 and Inception-ResNet-V2 model with the gold standard was 0.78, 0.83 and 0.88 respectively. The sensitivity, specificity, accuracy, Yorden index, and area under the ROC curve (AUC) of the Inception-ResNet-V2 were found highest, being 92.00%, 97.16%, 96.13%, 0.89 and 0.95, respectively. There was no statistically significant difference in the sensitivity and specificity among the three models (χ² = 3.892, 4.948, P > 0.05), while the difference in accuracy was statistically significant (χ² = 8.607, P < 0.05). In the external test, the best-fit model Inception-ResNet-V2 and the findings by staff showed better concordance with the gold standard, having the Kappa values of 0.80 and 0.83, and the AUCs 0.88 and 0.92, respectively, of which the difference was not statistically significant (P > 0.05). Conclusion The deep learning-based visual intelligent recognition model for O. hupensis snailshowed evident accuracy.

Key words: Oncomelania hupensis, Intelligent recognition, Deep learning, Computer vision, Machine learning, Schistosomiasis, Surveillance

中图分类号:

R383.241

施亮, 熊春蓉, 刘毛毛, 魏秀参, 张键锋, 王鑫瑶, 王涛, 杭德荣, 羊海涛, 杨坤. 基于深度学习技术的湖北钉螺视觉智能识别模型效能评价[J]. 中国寄生虫学与寄生虫病杂志, 2021, 39(6): 764-770.

SHI Liang, XIONG Chun-rong, LIU Mao-mao, WEI Xiu-shen, ZHANG Jian-feng, WANG Xin-yao, WANG Tao, HANG De-rong, YANG Hai-tao, YANG Kun. Evaluation of efficacy of visual intelligent recognition model for Oncomelania hupensis based on deep learning technology[J]. Chinese Journal of Parasitology and Parasitic Diseases, 2021, 39(6): 764-770.

图/表 7

图1

图2

表1

表2

图3

图4

表3

参考文献 33

[1]	Liao ZW, Wang SQ. Prevalence and prevention of major tropical diseases in China, 2000—2019[J]. Chin Trop Med, 2020, 20(3): 193-201. (in Chinese)
	(廖志武, 王善青. 我国2000—2019年主要热带病的流行与防治概况[J]. 中国热带医学, 2020, 20(3): 193-201.)
[2]	Deng WC, Li YS, Cheng XH, et al. Implications, spiritual characteristics and practical significance of Chinese schistosomiasis control culture[J]. Chin J Schisto Control, 2020, 32(3): 222-224, 229. (in Chinese)
	(邓维成, 李岳生, 程湘晖, 等. 论中国血防文化的内涵与精神特质及其现实意义[J]. 中国血吸虫病防治杂志, 2020, 32(3): 222-224, 229.)
[3]	Mao SB. The biology of schistosomiasis and the control of schistosomiasis[M]. Beijing: People’s Medical Publishing House, 1990: 699. (in Chinese)
	(毛守白. 血吸虫生物学与血吸虫病的防治[M]. 北京: 人民卫生出版社, 1990: 699.)
[4]	Zhou XN. Science on oncomelania snail[M]. Beijing: Science Press, 2005: 1. (in Chinese)
	(周晓农. 实用钉螺学[M]. 北京: 科学出版社, 2005: 1.)
[5]	Jiang TT, Yang K. Progresses of research on patterns and monitoring approaches of Oncomelania hupensis spread[J]. Chin J Schisto Control, 2020, 32(2): 208-212. (in Chinese)
	(蒋甜甜, 杨坤. 钉螺扩散规律与监测方法研究进展[J]. 中国血吸虫病防治杂志, 2020, 32(2): 208-212.)
[6]	Chinese Academy of Medical Sciences. Problems and improvement opinions on the method of Oncomelania survey[J]. Med Health Exp, 1960, 1: 18-19. (in Chinese)
	(中国医学科学院. 钉螺调查方法上存在的问题和改进意见[J]. 医药卫生快报, 1960, 1: 18-19.)
[7]	Chen M, Peng XW, Zhang HM, et al. Systematic field survey on the snail density by space sectioning method[J]. J Trop Dis Parasitol, 2015, 13(1): 23-25. (in Chinese)
	(陈美, 彭孝武, 张华明, 等. 按钉螺密度分级设定系统抽样查螺间距的探讨[J]. 热带病与寄生虫学, 2015, 13(1): 23-25.)
[8]	LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. doi: 10.1038/nature14539
[9]	He KM, Zhang XY, Ren SQ., et al. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification [C]//2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile: IEEE, 2015: 1026-1034.
[10]	Carin L, Pencina MJ. On deep learning for medical image analysis[J]. JAMA, 2018, 320(11): 1192. doi: 10.1001/jama.2018.13316
[11]	Chae S, Kwon S, Lee D. Predicting infectious disease using deep learning and big data[J]. Int J Environ Res Public Health, 2018, 15(8): 1596. doi: 10.3390/ijerph15081596
[12]	Wallis C. How artificial intelligence will change medicine[J]. Nature, 2019, 576(7787): S48. doi: 10.1038/d41586-019-03845-1
[13]	National Health and Family Planning Commission of the People’s Republic of China. Survey of oncomelanid snails: WS/T 563—2017[S]. Beijing: China Standard Press, 2017. (in Chinese)
	(中华人民共和国国家卫生和计划生育委员会. 钉螺调查 WS/T 563—2017[S]. 北京: 中国标准出版社, 2017.)
[14]	Liu YY, Zhang WZ, Wang YX. Medical malacology[M]. Beijing: Ocean Press, 1993: 1-157. (in Chinese)
	(刘月英, 张文珍, 王耀先. 医学贝类学[M]. 北京: 海洋出版社, 1993: 1-157.)
[15]	Shi L, Xiong CR, Liu MM, et al. Establishment of a deep learning-visual model for intelligent recognition of Oncomelania hupensis[J]. Chin J Schisto Control, 2021, 33(5): 445-451. (in Chinese)
	(施亮, 熊春蓉, 刘毛毛, 等. 基于深度学习技术的湖北钉螺视觉智能识别模型的建立[J]. 中国血吸虫病防治杂志, 2021, 33(5): 445-451.
[16]	Wei XS. Analyzing deep learning: convolutional neural network principle and vision practice[M]. Beijing: Publishing House of Electronics Industry, 2018. (in Chinese)
	(魏秀参. 解析深度学习: 卷积神经网络原理与视觉实践[M]. 北京: 电子工业出版社, 2018.
[17]	Howard AG, Zhu ML, Chen B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv: 1704.04861, 2017.
[18]	He KM, Zhang XY, Ren SQ, et al. Deep residual learning for image recognition[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[19]	Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, Inception-ResNet and the impact of residual connections on learning[C]. AAAI Conference on Artificial Intelligence, 2016.
[20]	Zhou ZH. Machine learning[M]. Beijing: Tsinghua University Press, 2016: 28-36. (in Chinese)
	(周志华. 机器学习[M]. 北京: 清华大学出版社, 2016: 28-36.)
[21]	Sokolova M, Lapalme G. A systematic analysis of performance measures for classification tasks[J]. Inf Process Manag, 2009, 45(4): 427-437. doi: 10.1016/j.ipm.2009.03.002
[22]	Zhou YB, Zhao GM. Reliability of measurement and the methods of estimating reliability[J]. Chin J Epidemiol, 2003, 24(12): 1146-1149. (in Chinese)
	(周艺彪, 赵根明. 测量的可靠性及其估计方法[J]. 中华流行病学杂志, 2003, 24(12): 1146-1149.)
[23]	Zhang LJ, Xu ZM, Dang H, et al. Endemic status of schistosomiasis in People’s Republic of China in 2019[J]. Chin J Schisto Control, 2020, 32(6): 551-558. (in Chinese)
	(张利娟, 徐志敏, 党辉, 等. 2019年全国血吸虫病疫情通报[J]. 中国血吸虫病防治杂志, 2020, 32(6): 551-558.)
[24]	Zhang LJ, Zhu HQ, Wang Q, et al. Assessment of schistosomiasis transmission risk along the Yangtze River basin after the flood disaster in 2020[J]. Chin J Schisto Control, 2020, 32(5): 464-468, 475. (in Chinese)
	(张利娟, 祝红庆, 王强, 等. 2020年长江流域洪涝灾害后血吸虫病传播风险分析[J]. 中国血吸虫病防治杂志, 2020, 32(5): 464-468, 475.)
[25]	Zhang YE. Search of snail recognition and counting based on template matching of color and shape[J]. J Gannan Norm Univ, 2012, 33(6): 33-36. (in Chinese)
	(章银娥. 基于颜色和形状的模板匹配的钉螺识别计数研究[J]. 赣南师范学院学报, 2012, 33(6): 33-36.)
[26]	Yan SH, Huang XT. The Oncomelania digital image identification based on SIFT & SVM[J]. J Gannan Norm Univ, 2011, 32(6): 58-61. (in Chinese)
	(严深海, 黄贤通. 基于SIFT与SVM的钉螺数字图像识别[J]. 赣南师范学院学报, 2011, 32(6): 58-61.)
[27]	Wang H. Research on snail image recognition technology based on neural network[D]. Wuhan: Hubei University, 2010: 1-49. (in Chinese)
	(汪浩. 基于神经网络的钉螺图像识别技术研究[D]. 武汉: 湖北大学, 2010: 1-49.)
[28]	Wei XS, Xie CW, Wu JX, et al. Mask-CNN: localizing parts and selecting descriptors for fine-grained bird species categorization[J]. Pattern Recognit, 2018, 76: 704-714. doi: 10.1016/j.patcog.2017.10.002
[29]	Wei XS, Luo JH, Wu J, et al. Selective convolutional descriptor aggregation for fine-grained image retrieval[J]. IEEE Trans Image Process, 2017, 26(6): 2868-2881. doi: 10.1109/TIP.2017.2688133
[30]	Smith KP, Kirby JE. Image analysis and artificial intelligence in infectious disease diagnostics[J]. Clin Microbiol Infect, 2020, 26(10): 1318-1323. doi: 10.1016/j.cmi.2020.03.012
[31]	Shi F, Wang J, Shi J, et al. Review of artificial intelligence techniques in imaging data acquisition, segmentation, and diagnosis for COVID-19[J]. IEEE Rev Biomed Eng, 2021, 14: 4-15. doi: 10.1109/RBME.4664312
[32]	Song J, Gao C, Han Q, et al. Construction and clinical preliminary validation of an automaticbone age assessment model based on deep learning[J]. Chin J Radiol, 2019, 53(11): 974-978. (in Chinese)
	(宋娟, 高畅, 韩青, 等. 基于深度学习的儿童骨龄智能评估模型构建及初步临床验证[J]. 中华放射学杂志, 2019, 53(11): 974-978.)
[33]	Yang F, Poostchi M, Yu H, et al. Deep learning for smartphone-based malaria parasite detection in thick blood smears[J]. IEEE J Biomed Heal Informatics, 2019, 24(5): 1427-1438.

螺类 Snails	训练集 Training set	测试集Test set		合计 Total
螺类 Snails	训练集 Training set	内部测试集 Internal test set	外部测试集 External test set	合计 Total
钉螺 Oncomelania hupensis	594	150	30	774
非钉螺 Non-Oncomelania hupensis	1 316	599	30	1 945
合计Total	1 910	749	60	2 719

模型Model	灵敏度/%Sensitivity/%	特异性/%Specificity/%	准确率/% Accuracy/%	约登指数 Youden index	Kapaa值 Kappa value	ROC曲线下面积 AUC
MobilenetV2	84.67	94.66	92.66^a	0.79	0.78	0.90
Resnet50	88.00	96.16	94.53^ab	0.84	0.83	0.92
Inception-ResNet-V2	92.00	97.16	96.13^a	0.89	0.88	0.95
χ²	3.892	4.948	8.607
P值P value	> 0.05	> 0.05	< 0.05

分组 Group	灵敏度/% Sensitivity/%	特异性/% Specificity/%	准确率/% Accuracy/%	约登指数 Youden index	Kappa值 Kappa value	AUC
工作人员Staff	86.67	96.67	91.67	0.83	0.83	0.92
模型Model	86.67	93.33	90.00	0.80	0.80	0.88

基于深度学习技术的湖北钉螺视觉智能识别模型效能评价

Evaluation of efficacy of visual intelligent recognition model for Oncomelania hupensis based on deep learning technology

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 33

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李慧敏, 刘婧姝, 王希涵, 赵翰卿, 解艺, 尹静娴, 吕超, 周楠, 蒋天哥, 谷思雨, 殷堃, 周晓农, 郭晓奎, 胡沁沁. 基于全健康理念的新发传染病综合监测预警体系：结构与创新[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(5): 572-578.
[2]	王可艺, 舒黄芳, 方悦怡, 曾庆生, 宋铁. 2016—2020年广东省华支睾吸虫病高流行地区人群感染监测分析[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(5): 629-634.
[3]	冯家鑫, 公衍峰, 罗卓韦, 汪伟, 曹淳力, 许静, 李石柱. 我国血吸虫病防治策略的科学基础与“十四五”展望[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(4): 428-435.
[4]	曹磊, 马琳, 朱妮, 张义, 王安礼, 王舒, 李欣欣. 2011—2020年陕西省疟疾流行特征[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(4): 454-459.
[5]	陈兵, 张国莉, 张高红. 血吸虫病候选疫苗临床研究进展[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(4): 511-515.
[6]	谢贤良, 陈云虹, 江典伟, 高澜琳, 郑丹, 张志魁, 魏美羡, 谢汉国. 2016—2020年漳州市国家监测点土源性线虫感染情况[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(4): 545-548.
[7]	江莉, 张耀光, 刘红霞, 王真瑜, 朱民, 吴寰宇. 疟疾蚊媒监测多重PCR方法的建立[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(2): 159-167.
[8]	于维君, 王子江, 王博, 毛玲玲, 吴琪俊, 姚文清, 孙英伟. 2016—2020年辽宁省土源性线虫感染现状及流行趋势[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(2): 211-215.
[9]	吴向林, 闫芳, 段红菊, 齐蓉婷, 高建炜. 2020年宁夏回族自治区棘球蚴病监测结果分析[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(2): 242-246.
[10]	陈云虹, 谢汉国, 谢贤良, 江典伟, 张榕燕. 2017—2019年福建省人群土源性线虫感染现状及防治效果[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(1): 84-87.
[11]	高元, 胡媛, 曹建平. 免疫细胞对血吸虫病肝纤维化作用的研究进展[J]. 中国寄生虫学与寄生虫病杂志, 2022, 40(1): 88-93.
[12]	夏志贵, 丰俊, 张丽, 冯欣宇, 黄芳, 尹建海, 周水森, 周升, 杨恒林, 王善青, 高琪, 汤林华, 严俊. 中国消除疟疾：监测响应系统的实施与成效分析[J]. 中国寄生虫学与寄生虫病杂志, 2021, 39(6): 733-740.
[13]	郭苏影, 祝红庆, 曹淳力, 邓王平, 鲍子平, 贾铁武, 李银龙, 吕超, 秦志强, 张利娟, 冯婷, 杨帆, 吕山, 许静, 李石柱. 2020年长江中下游地区洪涝灾害后血吸虫病传播风险评估[J]. 中国寄生虫学与寄生虫病杂志, 2021, 39(6): 753-758.
[14]	万孝玲, 吕国丽, 蒋智华, 区方奇, 韦海艳, 唐雯茜, 林源, 张伟尉, 黄铿凌, 石云良, 刘多, 刘健, 孟军, 杨益超. 2016—2020年广西壮族自治区人群土源性线虫感染监测结果[J]. 中国寄生虫学与寄生虫病杂志, 2021, 39(6): 800-806.
[15]	崔延雯, 黄芳, 尹建海, 夏志贵. 疟疾控制与消除中血清学监测方法的研究与应用进展[J]. 中国寄生虫学与寄生虫病杂志, 2021, 39(6): 836-841.