基于增强图卷积神经网络的病毒形态识别方法研究

doi:10.12114/j.issn.1007-9572.2022.0123

中国全科医学 ›› 2022, Vol. 25 ›› Issue (14): 1749-1756.DOI: 10.12114/j.issn.1007-9572.2022.0123

• 论著·诊疗新技术研究 • 上一篇下一篇

基于增强图卷积神经网络的病毒形态识别方法研究

哈艳¹^,², 袁伟珵³, 孟翔杰⁴, 田俊峰²^,⁵^,^*()

1. 071002　河北省保定市，河北大学管理学院
2. 071002　河北省保定市，河北省高可信信息系统重点实验室
3. 050017　河北省石家庄市，河北医科大学基础医学院
4. 071002　河北省保定市，河北大学数学与信息科学学院
5. 071002　河北省保定市，河北大学网络空间安全与计算机学院

收稿日期:2022-03-11 修回日期:2022-03-22 出版日期:2022-03-31 发布日期:2022-04-07
通讯作者: 田俊峰
哈艳，袁伟珵，孟翔杰，等.基于增强图卷积神经网络的病毒形态识别方法研究[J].中国全科医学，2022，25（14）：1749-1756. [www.chinagp.net]
作者贡献：哈艳、孟翔杰进行文章的构思与设计，研究的实施与可行性分析，数据整理；哈艳、田俊峰进行数据收集，论文的修订，结果的分析与解释；孟翔杰进行统计学处理；田俊峰撰写论文，对文章整体负责，监督管理；哈艳负责文章的质量控制及审校。
基金资助:
国家自然科学基金资助项目(61802106); 河北省自然科学基金资助项目(F2021201049)

Research on Virus Morphology Recognition Method Based on Enhanced Graph Convolutional Network

Yan HA¹^,², Weicheng YUAN³, Xiangjie MENG⁴, Junfeng TIAN²^,⁵^,^*()

1. School of Management, Hebei University, Baoding 071002, China
2. Key Laboratory on High Trusted Information System in Hebei Province, Baoding 071002, China
3. School of Basic Medicine, Hebei Medical University, Shijiazhuang 050017, China
4. College of Mathematics and Information Science, Hebei University, Baoding 071002, China
5. School of Cyber Security and Computer, Hebei University, Baoding 071002, China

Received:2022-03-11 Revised:2022-03-22 Published:2022-03-31 Online:2022-04-07
Contact: Junfeng TIAN
About author:
HA Y, YUAN W C, MENG X J, et al. Research on virus morphology recognition method based on enhanced graph convolutional network[J]. Chinese General Practice, 2022, 25 (14) : 1749-1756.

分享到

摘要/Abstract

摘要： 背景透射电子显微镜（TEM）是检测病毒的重要手段，传统TEM检测常依靠专家手工观察，操作步骤繁琐，且已有机器学习方法易受到背景、噪声的影响，导致病毒检测方法准确率差、效率低且耗时长。目的探讨增强图卷积神经网络（EGCN）对TEM图像中的病毒形态自动识别问题，以提高TEM病毒检测的效率。方法 EGCN模型利用卷积神经网络（CNN）提取像素间的局部特征信息，并结合样本特征之间的最近邻关系利用图卷积网络（GCN）进行图特征学习。在模型优化中联合优化群体超分类损失和分类交叉熵损失以提高EGCN模型对病毒类别信息特征的提取能力，较CNN对TEM病毒图像特征具备更强的特征提取能力。结果通过不同方法在15类TEM病毒图像数据集上开展实验，EGCN达到3.40%的top-1错误率、1.88%的top-2错误率、96.65%的精确度和96.60%的召回率，并通过一系列对比实验表明EGCN模型可以有效避免TEM图像中背景、噪声等的影响，提高对病毒识别的准确率。结论 EGCN可以有效解决病毒形态识别任务，为病毒的诊断提供重要的参考价值。

关键词: 病毒形态分类, 神经网络，计算机, 卷积神经网络, 图卷积网络, 增强图卷积神经网络

Abstract:

Background

Transmission electron microscope (TEM) is an important method to detect virus. TEM detection often relies on manual observation by experts, and the operation steps are cumbersome. Moreover, existing machine learning methods are easily affected by background and noise, resulting in poor virus detection methods, low efficiencyand time consuming.

Objective

In order to improve the efficiency of TEM virus detection, an Enhanced Graph Convolution Network (EGCN) is proposed to solve the problem of automatic identification of virus morphology in TEM images.

Methods

In this model, Convolutional Neural Network (CNN) was used to extract the local feature information between pixels, and GCN was used for graph feature learning combined with the nearest neighbor relationship between sample features. In the model optimization, the group super classification loss and classification cross entropy loss were introduced to improve the feature extraction ability of the model for virus category information, and further improve the robustness of TEM virus image features compared with convolution neural network.

Results

Experiments were carried out on 15 types of TEM virus image datasets through different methods, and EGCN achieved a top-1 error rate of 3.40%, a top-2 error rate of 1.88%, a precision of 96.65%, and a recall rate of 96.60%. A series of comparative experiments demonstrated that the EGCN can effectively solve the influence of background and noise in TEM virus recognition.

Conclusion

By using the enhanced graph convolutional neural network, the task of virus morphology recognition can be effectively solved, providing important reference value for virus diagnosis.

Key words: Virus morphological classification, Neural networks, computer, Convolutional neural network, Graph convolutional network, Enhanced graph convolutional network

哈艳,袁伟珵,孟翔杰,等. 基于增强图卷积神经网络的病毒形态识别方法研究[J]. 中国全科医学, 2022, 25(14): 1749-1756. DOI: 10.12114/j.issn.1007-9572.2022.0123.
Yan HA,Weicheng YUAN,Xiangjie MENG, et al. Research on Virus Morphology Recognition Method Based on Enhanced Graph Convolutional Network[J]. Chinese General Practice, 2022, 25(14): 1749-1756. DOI: 10.12114/j.issn.1007-9572.2022.0123.

图/表 9

图1 EGCN算法模型整体结构

Figure 1 Overall structure of EGCN algorithm model

表1 RepVGG模型

Table 1 The model of Rep VGG network

阶段	输出尺寸（像素）	层数
1	32×32	1
2	16×16	4
3	8×8	6
4	4×4	16
5	2×2	1

表2 TEM病毒数据集

Table 2 TEM virus dataset.

病毒类别	训练图像	测试图像	总数
腺病毒	75	25	100
星状病毒	75	25	100
刚果出血热病毒	75	25	100
牛痘病毒	75	25	100
登革热病毒	75	25	100
埃博拉病毒	75	25	100
流行性感冒病毒	75	25	100
拉沙病毒	75	25	100
马尔堡病毒	75	25	100
诺沃克病毒	75	25	100
口疮病毒	75	25	100
乳头状瘤病毒	75	25	100
裂谷热病毒	75	25	100
轮状病毒	75	25	100
西尼罗河病毒	75	25	100
总计	1 125	375	1 500

表3 不同模型对病毒形态分类的定量分析结果汇总（%）

Table 3 Summary of quantitative analysis results of virus morphological classification by different models

方法	top-1错误率	top-2错误率	精确度	召回率
VGG-19	14.67	9.69	86.23	85.33
ResNet-50	16.27	13.89	84.87	83.73
DenseNet-101	17.33	10.36	83.93	82.67
RepVGG	12.27	8.56	88.45	87.73
RMAN	4.67	2.52	95.41	95.33
EGCN	3.40	1.88	96.65	96.60

图2 所有对比方法中每个类别的top-1错误率

Figure 2 Top-1 error rates of each category in comparison methods

图3 混淆矩阵

Figure 3 The confusion matrix

图4 t-SNE可视化效果图注：a~c表示监督模型，d~e表示病毒识别算法和f表示EGCN算法

Figure 4 Thet-SNEplotsofourmethodandcomparisonalgorithms

图5 损失函数中平衡参数　对EGCN算法的影响注：横轴表示参数值，纵轴表示top-1错误率；以Lcp为主要损失函数，λ1，λ2和λ3和分别为Lca、Lgc和Lcon的平衡参数

Figure 5 The influence of balance parameters in loss function on our EGCN algorithm

表4 EGCN方法的消融实验结果

Figure 4 Ablation experimental results of EGCN method.

消融模块	top-1错误率（%）
无数据增强	4.17
无GCN模块	12.27
无超分类技巧	3.96
无图归一化	32.54
EGCN模型	3.40

参考文献 29

[1]	SPURGEON S R，OPHUS C，JONES L，et al. Towards data-driven next-generation transmission electron microscopy[J]. Nat Mater，2021，20（3）：274-279. DOI：10.1038/s41563-020-00833-z.
[2]	HOPFER H，HERZIG M C，GOSERT R，et al. Hunting coronavirus by transmission electron microscopy - a guide to SARS-CoV-2-associated ultrastructural pathology in COVID-19 tissues[J]. Histopathology，2021，78（3）：358-370. DOI：10.1111/his.14264.
[3]	GIBBS A J，HAJIZADEH M，OHSHIMA K，et al. The potyviruses：an evolutionary synthesis is emerging[J]. Viruses，2020，12（2）：E132. DOI：10.3390/v12020132.
[4]	YANG X L，TAN C W，ANDERSON D E，et al. Characterization of a filovirus （Měnglà virus） from Rousettus bats in China[J]. Nat Microbiol，2019，4（3）：390-395. DOI：10.1038/s41564-018-0328-y.
[5]	GRUBAUGH N D，LADNER J T，LEMEY P，et al. Tracking virus outbreaks in the twenty-first century[J]. Nat Microbiol，2019，4（1）：10-19. DOI：10.1038/s41564-018-0296-2.
[6]	XIAO C，CHEN X，XIE Q W，et al. Virus identification in electron microscopy images by residual mixed attention network[J]. Comput Methods Programs Biomed，2021，198：105766. DOI：10.1016/j.cmpb.2020.105766.
[7]	ABDALLA M A E，SEKER H. Recognition of protozoan parasites from microscopic images：Eimeria species in chickens and rabbits as a case study[C]. IEEE，2017：1517-1520. DOI：10.1109/EMBC.2017.8037124.
[8]	MARK G，APARICIO A，RAVENTóS B，et al. Leishmaniasis parasite segmentation and classification using deep learning[C]. Springer，Cham，2018：53-62.
[9]	LI S，YANG Q，JIANG H，et al. Parasitologist-level classification of apicomplexan parasites and host cell with deep cycle transfer learning （DCTL）[J]. Bioinformatics，2020，36（16）：4498-4505.
[10]	SINTORN I M，HOMMAN-LOUDIYI M，SÖDERBERG-NAUCLéR C，et al. A refined circular template matching method for classification of human Cytomegalovirus capsids in TEM images[J]. Comput Methods Programs Biomed，2004，76（2）：95-102. DOI：10.1016/j.cmpb.2004.03.006.
[11]	ONG H，CHANDRAN V. Identification of gastroenteric viruses by electron microscopy using higher order spectral features[J]. J Clin Virol，2005，34（3）：195-206. DOI：10.1016/j.jcv.2005.04.001.
[12]	WEN Z J，LI Z J，PENG Y X，et al. Virus image classification using multi-scale completed local binary pattern features extracted from filtered images by multi-scale principal component analysis[J]. Pattern Recognit Lett，2016，79：25-30. DOI：10.1016/j.patrec.2016.04.022.
[13]	SAHBI H. Skeleton-based hand-gesture recognition with lightweight graph convolutional networks[EB/OL]. （2021-12-27）[2022-01-20].
[14]	MIN C，XU J L，XIAO L，et al. Attentional graph neural network for parking-slot detection[J]. IEEE Robotics Autom Lett，2021，6（2）：3445-3450. DOI：10.1109/LRA.2021.3064270.
[15]	ZHOU L，MAO Q，DONG M. Objective class-based micro-expression recognition through simultaneous action Unit detection and feature aggregation[EB/OL]. （2020-12-24）[2022-01-20].
[16]	XIAO X，JI C，Mudiyanselage T B，et al. PK-GCN：prior knowledge assisted image classification using graph convolution networks[EB/OL]. （2020-09-24）[2022-01-20].
[17]	ADNAN M，KALRA S，TIZHOOSH H R. Representation learning of histopathology images using graph neural networks[C]. IEEE，2020：4254-4261. DOI：10.1109/CVPRW50498.2020.00502.
[18]	BAO L Q，MA B P，CHANG H，et al. Masked graph attention network for person re-identification[C]. IEEE，2019：1496-1505. DOI：10.1109/CVPRW.2019.00191.
[19]	SIMONYAN K，ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. （2014-09-04）[2022-01-20].
[20]	DING X H，ZHANG X Y，MA N N，et al. RepVGG：making VGG-style ConvNets great again[C]. IEEE，2021：13728-13737. DOI：10.1109/CVPR46437.2021.01352.
[21]	KIPF T N，WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL]. （2016-09-09）[2022-01-20].
[22]	KYLBERG G，UPPSTRÖM M，HEDLUND K O，et al. Segmentation of virus particle candidates in transmission electron microscopy images[J]. J Microsc，2012，245（2）：140-147. DOI：10.1111/j.1365-2818.2011.03556.x.
[23]	KINGMA D P，BA J. Adam：a method for stochastic optimization[EB/OL]. （2014-12-22）[2022-01-20].
[24]	ZHOU J L，GANDOMI A H，CHEN F，et al. Evaluating the quality of machine learning explanations：a survey on methods and metrics[J]. Electronics，2021，10（5）：593. DOI：10.3390/electronics10050593.
[25]	HE K M，ZHANG X Y，REN S Q，et al. Deep residual learning for image recognition[C]. IEEE，2016：770-778. DOI：10.1109/CVPR.2016.90.
[26]	SINGH D，KUMAR V，KAUR M. Densely connected convolutional networks-based COVID-19 screening model[J]. Appl Intell （Dordr），2021：2021Feb7；1-2021Feb7；8. DOI：10.1007/s10489-020-02149-6.
[27]	DENG J，DONG W，SOCHER R，et al. ImageNet：a large-scale hierarchical image database[C]. IEEE，2009：248-255. DOI：10.1109/CVPR.2009.5206848.
[28]	ZHANG L X，YAN W Q. Deep learning methods for virus identification from digital images[C]. IEEE，2020：1-6. DOI：10.1109/IVCNZ51579.2020.9290670.
[29]	SHIAELIS N，TOMETZKI A，PETO L，et al. Virus detection and identification in minutes using single-particle imaging and deep learning[J]. medRxiv，2020，DOI：10.1101/2020.10.13.20212035. DOI：10.1101/2020.10.13.20212035.

基于增强图卷积神经网络的病毒形态识别方法研究

Research on Virus Morphology Recognition Method Based on Enhanced Graph Convolutional Network

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 29

相关文章 1

编辑推荐

Metrics

留言