融合Grad-CAM和卷积神经网络的COVID-19检测算法

doi:10.3778/j.issn.1673-9418.2105117

计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (9): 2108-2120.DOI: 10.3778/j.issn.1673-9418.2105117

融合Grad-CAM和卷积神经网络的COVID-19检测算法

朱炳宇, 刘朕, 张景祥()

江南大学理学院,江苏无锡 214122

收稿日期:2021-05-11 修回日期:2021-07-15 出版日期:2022-09-01 发布日期:2021-07-23
通讯作者: + E-mail: zhangjingxiang@jiangnan.edu.cn
作者简介:朱炳宇（1997—）,男,河北邯郸人,硕士研究生,主要研究方向为人工智能、模式识别、智能计算及应用。
刘朕（1998—）,男,安徽合肥人,硕士研究生,主要研究方向为人工智能、模式识别、智能计算及应用。
张景祥（1977—）,男,吉林通化人,博士,副教授,硕士生导师,主要研究方向为人工智能、模式识别、智能计算及应用。
基金资助:
国家自然科学基金(61772239);国家自然科学基金(11804123)

COVID-19 Detection Algorithm Combining Grad-CAM and Convolutional Neural Network

ZHU Bingyu, LIU Zhen, ZHANG Jingxiang()

School of Science, Jiangnan University, Wuxi, Jiangsu 214122, China

Received:2021-05-11 Revised:2021-07-15 Online:2022-09-01 Published:2021-07-23
About author:ZHU Bingyu, born in 1997, M.S. candidate. His research interests include artificial intelligence, pattern recognition, intelligent computation and its application.
LIU Zhen, born in 1998, M.S. candidate. His research interests include artificial intelligence, pattern recognition, intelligent computation and its application.
ZHANG Jingxiang, born in 1977, Ph.D., associate professor, M.S. supervisor. His research interests include artificial intelligence, pattern recognition, intelligent computation and its application.
Supported by:
National Natural Science Foundation of China(61772239);National Natural Science Foundation of China(11804123)

摘要/Abstract

摘要：

新型冠状病毒肺炎（COVID-19）检测中胸部X射线（CXR图像）和电子计算机断层扫描（CT）图像是两种主要技术手段,为医生诊断提供了重要依据。针对当前卷积神经网络（CNN）在医学放射性图像中检测COVID-19的准确率不高、算法复杂、无法标记特征区域的问题,提出了一种融合梯度加权类激活映射（Grad-CAM）颜色可视化和卷积神经网络的算法（GCCV-CNN）,对COVID-19阳性患者、COVID-19阴性患者、普通肺炎患者以及正常人的肺部CXR图像和CT扫描图像进行快速分类。通过定位到CXR图像和CT扫描图像中CNN进行分类的关键区域,再综合深度学习算法得到更准确的检测结果。为验证GCCV-CNN算法的有效性,分别在3个COVID-19阳性患者数据集上进行实验,并与已有算法进行比较。结果表明该算法对COVID-19阳性患者的CXR图像和CT扫描图像分类性能优于“新冠网络”（COVID-Net）算法及迁移学习新冠网络（DeTraC-Net）算法,准确率最高达98.06%,速度更快的同时还具有较好的鲁棒性。

关键词: CXR图像, CT扫描图像, COVID-19, Grad-CAM, 融合Grad-CAM颜色可视化和CNN的算法（GCCV-CNN）

Abstract:

In the detection of COVID-19, chest X-ray (CXR) images and CT scan images are two main technical methods, which provide an important basis for doctors' diagnosis. Currently, convolutional neural network (CNN) in detecting the COVID-19 medical radioactive images has problems of low accuracy, complex algorithms, and inability to mark feature regions. In order to solve these problems, this paper proposes an algorithm combining Grad-CAM color visualization and convolutional neural network (GCCV-CNN). The algorithm can quickly classify lung X-ray images and CT scan images of COVID-19-positive patients, COVID-19-negative patients, general pneumonia patients and healthy people. At the same time, it can quickly locate the critical area in X-ray images and CT images. Finally, the algorithm can get more accurate detection results through the synthesis of deep learning algorithms. In order to verify the effectiveness of the GCCV-CNN algorithm, experiments are performed on three COVID-19-positive patient datasets and it is compared with existing algorithms. The results show that the classification performance of the algorithm is better than the COVID-Net algorithm and the DeTraC-Net algorithm. The GCCV-CNN algorithm achieves a high accuracy of 98.06%, which is faster and more robust.

Key words: chest X-ray (CXR) images, CT scan images, COVID-19, Grad-CAM, algorithm combining Grad-CAM color visualization and convolutional neural network (GCCV-CNN)

中图分类号:

TP183

朱炳宇, 刘朕, 张景祥. 融合Grad-CAM和卷积神经网络的COVID-19检测算法[J]. 计算机科学与探索, 2022, 16(9): 2108-2120.

ZHU Bingyu, LIU Zhen, ZHANG Jingxiang. COVID-19 Detection Algorithm Combining Grad-CAM and Convolutional Neural Network[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(9): 2108-2120.

图/表 19

图1 GCCV-CNN分类示意图

Fig.1 GCCV-CNN classification diagram

图2 AlexNet网络处理图像示意图

Fig.2 Diagram of AlexNet network processing images

表1 AlexNet网络参数

Table 1 AlexNet network parameters

层名称	输入	输出	参数量
Input	224×224×3	227×227×3	N/A
Block1-Conv1	227×227×3	55×55×96	34 848
Block1-ReLU1	55×55×96	55×55×96
Block1-MaxPooling1	55×55×96	27×27×96
Block2-Conv2	27×27×96	27×27×256	307 456
Block2-ReLU2	27×27×256	27×27×96
Block2-MaxPooling2	27×27×96	13×13×96
Block3-Conv3	13×13×96	13×13×384	885 120
Block4-Conv4	13×13×384	13×13×384	663 936
Block5-Conv5	13×13×384	13×13×256	442 624
Block5-MaxPooling3	13×13×256	6×6×256
Block5-Flatten	6×6×256	4 096
FC6	4 096	4 096	37 752 832
FC7	4 096	4 096	16 781 312
FC8	4 096	1 000	4 096 000
Output	1 000	1 000	N/A
Sum			60 964 128

图3 Grad-CAM结构示意图假设最后一层Softmax中输出目标类的概率为 y c,最后一层中的FeatureMap所有像素为 A i j,那么 y c对 A i j求偏导,即：

Fig.3 Grad-CAM structure diagram （10） α i j = ∂ y c ∂ A i j

图4 数据集图像数量直方图

Fig.4 Histogram of the number of images in dataset

图5 图像加噪、翻转前后对比

Fig.5 Comparison of images before and after adding noise and flipping

图6 三组实验Loss & Accuracy曲线图

Fig.6 Loss & Accuracy of three groups of experiments

图7 三组实验混淆矩阵

Fig.7 Confusion matrix of three groups of experiments

表2 COVID-19 VS NORMAL混淆矩阵

Table 2 COVID-19 VS NORMAL confusion matrix

Actual class	Predicted class
Actual class	COVID-19	NORMAL
COVID-19	TP=180	FN=0
NORMAL	FP=12	TN=168

表3 COVID-19 VS NORMAL参数值

Table 3 COVID-19 VS NORMAL parameters

参数	数值
Batch size	144
Step-per-epoch	10
Stride	2
Input size	224×224×3
Learning rate	0.000 1
Dropout	0.2
Epochs	25

图8 实验1 Grad-CAM结果

Fig.8 Grad-CAM results of experiment 1

表4 COVID-19 VS PNEUMONIA混淆矩阵

Table 4 COVID-19 VS PNEUMONIA confusion matrix

Actual class	Predicted class
Actual class	COVID-19	PNEUMONIA
COVID-19	TP=175	FN=5
PNEUMONIA	FP=2	TN=178

表5 COVID-19 VS PNEUMONIA参数值

Table 5 COVID-19 VS PNEUMONIA parameters

参数	数值
Batch size	72
Step-per-epoch	20
Stride	2
Input size	224×224×3
Learning rate	0.000 1
Dropout	0.2
Epochs	25

图9 实验2 Grad-CAM结果

Fig.9 Grad-CAM results of experiment 2

表6 COVID-19 VS NOCOVID-19混淆矩阵

Table 6 COVID-19 VS NOCOVID-19 confusion matrix

Actual class	Predicted class
Actual class	COVID-19	NOCOVID-19
COVID-19	TP=370	FN=30
NOCOVID-19	FP=66	TN=334

表7 COVID-19 VS NOCOVID-19参数值

Table 7 COVID-19 VS NOCOVID-19 parameters

参数	数值
Batch size	82
Step-per-epoch	20
Stride	2
Input size	224×224×3
Learning rate	0.000 1
Dropout	0.2
Epochs	50

图10 实验3 Grad-CAM结果

Fig.10 Grad-CAM results of experiment 3

表8 三组实验分类评价参数

Table 8 Classification evaluation parameters of three groups of experiments

实验	Precision/%	Recall/%	F- measure	Specificity/%	Sensitivity/%	OverallAcc/%
实验1	93.33	100.00	0.966	93.75	100.00	96.67
实验2	98.89	97.27	0.981	98.87	97.27	98.06
实验3	83.50	91.76	0.874	84.86	91.76	88.00

表9 Classification accuracy of three algorithms on three datasets 单位：%

Table 9

算法	数据集1	数据集2	数据集3
COVID-Net算法	94.30	92.50	91.00
DeTraC-Net算法	92.70	94.10	87.60
GCCV-CNN算法	96.67	98.06	88.00

参考文献 21

[1]	邱海波, 李绪言, 杜斌, 等. 危重型新型冠状病毒肺炎的治疗思考[J]. 中华结核和呼吸杂志, 2020, 43(4): 273-277.
	QIU H B, LI X Y, DU B, et al. Thoughts on the treatment of the COVID-19[J]. Chinese Journal of Tuberculosis and Respiratory Diseases, 2020, 43(4): 273-277.
[2]	HU Q, GUAN H, SUN Z, et al. Early CT features and temporal lung changes in COVID-19 pneumonia in Wuhan, China[J]. European Journal of Radiology, 2020, 128: 109017. DOI URL
[3]	钱宝鑫, 肖志勇, 宋威. 改进的卷积神经网络在肺部图像上的分割应用[J]. 计算机科学与探索, 2020, 14(8): 1358-1367. DOI
	QIAN B X, XIAO Z Y, SONG W. Application of improved convolutional neural network in lung image segmentation[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(8): 1358-1367.
[4]	王保加, 潘海为, 谢晓芹, 等. 基于多模态特征的医学图像聚类方法[J]. 计算机科学与探索, 2018, 12(3): 411-422. DOI
	WANG B J, PAN H W, XIE X Q, et al. Medical image clustering based on multimodal features[J]. Journal of Frontiers of Computer Science and Technology, 2018, 12(3): 411-422.
[5]	李振昊, 高小玲, 杨小娟, 等. 新型冠状病毒核酸检测分析[J]. 检验医学与临床, 2020, 17(10): 7-9.
	LI Z H, GAO X L, YANG X J, et al. The nucleic acid detection and analysis of COVID-19[J]. Laboratory Medicine and Clinic, 2020, 17(10): 7-9.
[6]	WANG S, KANG B, MA J L, et al. A deep learning algorithm using CT images to screen for corona virus disease (COVID-19)[J]. European Radiology, 2021, 31(8): 6096-6104. DOI URL
[7]	SONG Y, ZHENG S J, LI L, et al. Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images[J]. IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2021, 18(6): 2775-2780. DOI URL
[8]	SETHY P K, SANTI K, BEHERA, et al. Detection of coronavirus disease (COVID-19) based on deep features and support vector machine[J]. International Journal of Mathematical, Engineering and Management Sciences, 2020, 5(4): 643-651. DOI URL
[9]	WANG L, LIN Z Q, WONG A. COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-ray images[J]. Scientific Reports, 2020, 10(1): 19549. DOI URL
[10]	ABBAS A, ABDELSAMEA M M, GABER M M. Classification of COVID-19 in chest X-ray images using DeTraC deep convolutional neural network[J]. Applied Intelligence, 2021, 51(2): 854-864. DOI URL
[11]	APOSTOLOPOULOS I D, BESSIANA T. Covid-19: automatic detection from X-ray images utilizing transfer learning with convolutional neural networks[J]. Physical and Engineering Sciences in Medicine, 2020, 43: 635-640. DOI URL
[12]	PANWAR H, GUPTA P K, SIDDIQUI M K, et al. Application of deep learning for fast detection of COVID-19 in X-Rays using nCOVnet[J]. Chaos Solitons & Fractals, 2020, 138: 109944. DOI URL
[13]	KUMAR S, MISHRA S, SINGH S K. Deep transfer learning-based COVID-19 prediction using chest X-rays[J]. Physical and Engineering Sciences in Medicine, 2020, 14(1): 356-371.
[14]	LUJÁN-GARCÍA J E, YÁÑEZ-MÁRQUEZ C, VILLUENDAS-REY Y, et al. A transfer learning method for pneumonia classification and visualization[J]. Applied Sciences, 2020, 10(8): 2908. DOI URL
[15]	TOGAÇAR M, ERGEN B, CÖMERT Z. COVID-19 detection using deep learning models to exploit social mimic optimization and structured chest X-ray images using fuzzy color and stacking approaches[J]. Computers in Biology and Medicine, 2020, 121: 103805. DOI URL
[16]	OZTURK T, TALO M, YILDIRIM E A, et al. Automated detection of COVID-19 cases using deep neural networks with X-ray images[J]. Computers in Biology and Medicine, 2020, 121: 103792. DOI URL
[17]	SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization[J]. International Journal of Computer Vision, 2020, 128(2): 336-359. DOI URL
[18]	BENGIO Y, COURVILLE A C, VINCENT P. Representation learning: a review and new perspectives[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8): 1798-1828. DOI URL
[19]	MAHENDRAN A, VEDALDI A. Visualizing deep convolutional neural networks using natural pre-images[J]. International Journal of Computer Vision, 2016, 120(3): 233-255. DOI URL
[20]	COHEN J P, MORRISON P, LAN D, et al. COVID-19 image data collection: prospective predictions are the future[J]. Asian Pacific Journal of Tropical Medicine, 2020, 13(6): 8-12. DOI URL
[21]	SOARES E, ANGELOV P, BIASO S, et al. SARS-CoV-2 CT-scan dataset: a large dataset of real patients CT scans for SARS-CoV-2 identification[J]. Applied Intelligence, 2020, 19(1): 81-84.

融合Grad-CAM和卷积神经网络的COVID-19检测算法

COVID-19 Detection Algorithm Combining Grad-CAM and Convolutional Neural Network

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 19

参考文献 21

相关文章 6

编辑推荐

Metrics

[1]	任宁, 付岩, 吴艳霞, 梁鹏举, 韩希. 深度学习应用于目标检测中失衡问题研究综述[J]. 计算机科学与探索, 2022, 16(9): 1933-1953.
[2]	萨日娜, 李艳玲, 林民. 知识图谱推理问答研究综述[J]. 计算机科学与探索, 2022, 16(8): 1727-1741.
[3]	童敢, 黄立波. Winograd快速卷积相关研究综述[J]. 计算机科学与探索, 2022, 16(5): 959-971.
[4]	王忠民, 赵玉鹏, 郑镕林, 贺炎, 张嘉雯, 刘洋. 脑电信号情绪识别研究综述[J]. 计算机科学与探索, 2022, 16(4): 760-774.
[5]	马金林, 张裕, 马自萍, 毛凯绩. 轻量化神经网络卷积设计研究进展[J]. 计算机科学与探索, 2022, 16(3): 512-528.
[6]	蒋光峰, 胡鹏程, 叶桦, 仰燕兰. 基于重构误差的同构图分类模型[J]. 计算机科学与探索, 2022, 16(1): 185-193.