基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类

doi:10.13229/j.cnki.jdxbgxb20210347

吉林大学学报(工学版) ›› 2022, Vol. 52 ›› Issue (11): 2653-2661.doi: 10.13229/j.cnki.jdxbgxb20210347

• 计算机科学与技术 • 上一篇

基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类

王生生¹(),李晨旭¹,王翔宇¹,姚志林¹(),刘一申²,吴佳倩²,杨晴然²

^1.吉林大学计算机科学与技术学院，长春 130012
^2.东北师范大学附属中学，长春 130021

收稿日期:2021-04-22 出版日期:2022-11-01 发布日期:2022-11-16
通讯作者: 姚志林 E-mail:wss@jlu.edu.cn;yaozl@jlu.edu.cn
作者简介:王生生（1974-），男，教授，博士.研究方向：机器视觉，人工智能. E-mail： wss@jlu.edu.cn
基金资助:
国家重点研发计划项目(2020YFA0714103);国家自然科学基金区域创新发展联合基金项目(U19A2061);吉林省发展和改革委员会创新能力建设项目(2021FGWCXNLJSSZ10)

Brain tumor image classification based on improved residual capsule network and sparrow search

Sheng-sheng WANG¹(),Chen-xu LI¹,Xiang-yu WANG¹,Zhi-lin YAO¹(),Yi-shen LIU²,Jia-qian WU²,Qing-ran YANG²

^1.College of Computer Science and Technology，Jilin University，Changchun 130012，China
^2.High School Attached to Northeast Normal University，Changchun 130021，China

Received:2021-04-22 Online:2022-11-01 Published:2022-11-16
Contact: Zhi-lin YAO E-mail:wss@jlu.edu.cn;yaozl@jlu.edu.cn

摘要/Abstract

摘要：

本文提出了基于缩放重构残差胶囊网络和麻雀搜索的核磁共振成像（MRI）脑瘤图像分类方法。首先，针对图像质量差的MRI脑瘤图像，采用基于麻雀搜索的图像增强方法提升图片质量；其次，采用胶囊网络解决医疗图像数据量小、数据集不平衡的问题；最后，针对胶囊网络对于大尺寸图像产生的梯度消失和梯度爆炸问题，采用改进的残差网络提取尺寸较大图片的关键特征，使用缩放重构，降低模型体积，在避免过拟合的同时提高计算速度。实验结果验证了本文提出的模型在小样本、低质量、大尺寸MRI脑瘤图像分类问题上的有效性。

关键词: 计算机应用技术, 脑瘤图像分类, 麻雀搜索算法, 残差网络, 胶囊网络, 动态路由算法

Abstract:

A magnetic resonance imaging（MRI） brain tumor image classification method based on sparrow search algorithm and scaled reconstruction residual capsule network was proposed. Firstly， for MRI brain tumor images with poor image quality， an image enhancement method based on sparrow search was taken to improve the image quality. Secondly， the capsule network was used to achieve better results on the small data volume and unbalanced medical dataset. Finally， in view of the gradient disappearance and gradient explosion problems of the capsule network for large size images， an improved residual network was used to extract the key features of large size images. Meanwhile， by using scaled reconstruction， the volume of model decreased while the calculation speed increased. The experimental results verify the effectiveness of the proposed method in the classification of small samples， low-quality， large size MRI brain tumor images.

Key words: computer application technology, brain tumor classification, sparrow search algorithm, residual network, capsule network, dynamic routing agreement

中图分类号:

TP399

王生生,李晨旭,王翔宇,姚志林,刘一申,吴佳倩,杨晴然. 基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类[J]. 吉林大学学报(工学版), 2022, 52(11): 2653-2661.

Sheng-sheng WANG,Chen-xu LI,Xiang-yu WANG,Zhi-lin YAO,Yi-shen LIU,Jia-qian WU,Qing-ran YANG. Brain tumor image classification based on improved residual capsule network and sparrow search[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(11): 2653-2661.

图/表 7

表1

三种网络结构"

Capsule	Rescap less height	Rescap less layer
Input（224×224×1， image）	Input（112×112×4， image）	Input（224×224×1， image）
8×8 Conv，256，stride 8	3×3 Conv，64	3×3 Conv，64
	$3 × 3 c o n v, 64 3 × 3 c o n v, 64 × 2$	$3 × 3 c o n v, 32 3 × 3 c o n v, 32 × 2$
	$3 × 3 c o n v, 128 3 × 3 c o n v, 128$ 1×1 Conv，128，stride 2	$3 × 3 c o n v, 64 3 × 3 c o n v, 64 × 2$
	$3 × 3 c o n v, 128 3 × 3 c o n v, 128$	$3 × 3 c o n v, 64 3 × 3 c o n v, 64 × 2$
	$3 × 3 c o n v, 256 3 × 3 c o n v, 256$ 1×1 Conv，256，stride 2	$3 × 3 c o n v, 128 3 × 3 c o n v, 128$ 1×1 Conv，128，stride 2
	$3 × 3 c o n v, 256 3 × 3 c o n v, 256$	$3 × 3 c o n v, 128 3 × 3 c o n v, 128$
	$3 × 3 c o n v, 512 3 × 3 c o n v, 512$ 1×1 Conv，512，stride 2	$3 × 3 c o n v, 256 3 × 3 c o n v, 256$ 1×1 Conv，256，stride 2
	$3 × 3 c o n v, 512 3 × 3 c o n v, 512$	$3 × 3 c o n v, 256 3 × 3 c o n v, 256$
－	特征图reshape为（28×28×128）	－
8D/3200 primary capsule	8D/288 primary capsule	8D/288 primary capsule
16D/4 final capsule	16D/4 final capsule	16D/4 final capsule
Dense 512	Dense 512	Dense 512
Dense 1024	Dense 1024	Dense 1024
Dense 224×224	Dense 784	Dense 784

表1

图1

图2

表2

表3

表4

表5

参考文献 15

1	Pan Y, Huang W, Lin Z, et al. Brain tumor grading based on neural networks and convolutional neural networks[C]∥37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Milan, Italy, 2015: 699-702.
2	Deepak S, Ameer P M. Brain tumor classification using deep CNN features via transfer learning[J]. Computers in Bbiology and Medicine, 2019, 111: 103345.
3	Navid G, Afshin S, Modjtaba R.Deep neural network with generative adversarial networks pre-training for brain tumor classification based on MR images[J]. Biomedical Signal Processing and Control, 2020, 57: 101678.
4	Javaria A, Muhammad S, Nadia G, et al. Brain tumor classification based on DWT fusion of MRI sequences using convolutional neural network[J]. Pattern Recognition Letters, 2020, 129: 115-122.
5	Anupama M A, Sowmya V, Soman K P. Breast cancer classification using capsule network with preprocessed histology images[C]∥The 8th International Conference on Communication and Signal Processing, Melmaruvathur, India, 2019: 143-147.
6	Afshar P, Oikonomou A, Naderkhani F, et al. 3D-MCN: a 3D multi-scale capsule network for lung nodule malignancy prediction[J]. Scientific Reports, 2020, 10(1): 1-11.
7	Afshar P, Heidarian S, Naderkhani F, et al. Covid-caps: a capsule network-based framework for identification of covid-19 cases from x-ray images[J]. Pattern Recognition Letters, 2020, 138: 638-643.
8	Xue J K, Shen B. A novel swarm intelligence optimization approach: sparrow search algorithm[J]. Systems Science & Control Engineering An Open Access Journal, 2020, 8(1): 22-34.
9	Chen J, Yu W, Tian J, et al. Image contrast enhancement using an artificial bee colony algorithm[J]. Swarm and Evolutionary Computation, 2018, 38: 287-294.
10	Sabour S, Frosst N, Hinton G E. Dynamic routing between capsules[C]∥The 31st Annual Conference on Neural Information Processing Systems, Long Beach, CA, USA, 2017: 3857-3867.
11	He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 770-778.
12	许骞艺, 秦贵和, 孙铭会, 等. 基于改进的ResNeSt驾驶员头部状态分类算法[J]. 吉林大学学报: 工学版, 2021, 51(2): 704-711.
	Xu Qian-yi, Qin Gui-he, Sun Ming-hui, et al. Classification of drivers' head status based on improved ResNeSt[J].Journal of Jilin University(Engineering and Technology Edition), 2021, 51(2): 704-711.
13	方明, 陈文强. 结合残差网络及目标掩膜的人脸微表情识别[J]. 吉林大学学报: 工学版, 2021, 51(1): 303-313.
	Fang Ming, Chen Wen-qiang. Face micro-expression recognition based on ResNet with object mask[J].Journal of Jilin University(Engineering and Technology Edition), 2021, 51(1): 303-313.
14	Minarno A E, Mandiri M H C, Munarko Y, et al. Convolutional neural network with hyperparameter tuning for brain tumor classification[J]. Kinetik: Game Technology, Information System, Computer Network, Computing, Electronics and Control, 2021, 6(2): 127-132.
15	Jaeyong K, Zahid U, Jeonghwan G. MRI-based brain tumor classification using ensemble of deep features and machine learning classifiers[J]. Sensors, 2021, 21(6): 21062222.

相关文章 15

[1]	白天,徐明蔚,刘思铭,张佶安,王喆. 基于深度神经网络的诉辩文本争议焦点识别[J]. 吉林大学学报(工学版), 2022, 52(8): 1872-1880.
[2]	曲福恒,丁天雨,陆洋,杨勇,胡雅婷. 基于邻域相似性的图像码字快速搜索算法[J]. 吉林大学学报(工学版), 2022, 52(8): 1865-1871.
[3]	秦贵和,黄俊锋,孙铭会. 基于双手键盘的虚拟现实文本输入[J]. 吉林大学学报(工学版), 2022, 52(8): 1881-1888.
[4]	杨怀江,王二帅,隋永新,闫丰,周跃. 简化型残差结构和快速深度残差网络[J]. 吉林大学学报(工学版), 2022, 52(6): 1413-1421.
[5]	方世敏. 基于频繁模式树的多来源数据选择性集成算法[J]. 吉林大学学报(工学版), 2022, 52(4): 885-890.
[6]	刘铭,杨雨航,邹松霖,肖志成,张永刚. 增强边缘检测图像算法在多书识别中的应用[J]. 吉林大学学报(工学版), 2022, 52(4): 891-896.
[7]	董绍江,朱朋,裴雪武,李洋,胡小林. 基于子领域自适应的变工况下滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 288-295.
[8]	王生生,陈境宇,卢奕南. 基于联邦学习和区块链的新冠肺炎胸部CT图像分割[J]. 吉林大学学报(工学版), 2021, 51(6): 2164-2173.
[9]	赵宏伟,张子健,李蛟,张媛,胡黄水,臧雪柏. 基于查询树的双向分段防碰撞算法[J]. 吉林大学学报(工学版), 2021, 51(5): 1830-1837.
[10]	曹洁,屈雪,李晓旭. 基于滑动特征向量的小样本图像分类方法[J]. 吉林大学学报(工学版), 2021, 51(5): 1785-1791.
[11]	王春波,底晓强. 基于标签分类的云数据完整性验证审计方案[J]. 吉林大学学报(工学版), 2021, 51(4): 1364-1369.
[12]	钱榕,张茹,张克君,金鑫,葛诗靓,江晟. 融合全局和局部特征的胶囊图神经网络[J]. 吉林大学学报(工学版), 2021, 51(3): 1048-1054.
[13]	周炳海,吴琼. 基于多目标的机器人装配线平衡算法[J]. 吉林大学学报(工学版), 2021, 51(2): 720-727.
[14]	许骞艺,秦贵和,孙铭会,孟诚训. 基于改进的ResNeSt驾驶员头部状态分类算法[J]. 吉林大学学报(工学版), 2021, 51(2): 704-711.
[15]	宋元,周丹媛,石文昌. 增强OpenStack Swift云存储系统安全功能的方法[J]. 吉林大学学报(工学版), 2021, 51(1): 314-322.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

模型	Loss	Accuracy/%	Parameter
Rescap less layer	0.3013	73.10	8 512 948
Rescap less height	0.3047	72.84	14 780 084
Capsule	0.3179	68.27	8 327 472

类别	Rescap less layer		Rescap less height		Capsule
类别	Sensitivity/%	Specificity/%	Sensitivity/%	Specificity/%	Sensitivity/%	Specificity/%
No tumor	100.00	78.89	100.00	76.47	97.14	76.47
Meningioma tumor	96.52	87.81	95.65	88.89	86.09	84.95
Glioma tumor	19.00	100.00	20.00	100.00	18.00	99.32
Pituitary tumor	71.62	96.56	70.27	97.50	67.57	95.94

模型	Accuracy/%
模型	不使用图像增强	使用麻雀搜索图像增强
VGG19	45.68	53.86
ResNet18	58.12	64.38
文献［14］	61.70	69.50
文献［15］	63.80	70.30
Rescap less layer	68.50	73.10

基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类

Brain tumor image classification based on improved residual capsule network and sparrow search

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 15

相关文章 15

Metrics

本文评价

推荐阅读 0