留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于多尺度特征与通道特征融合的脑肿瘤良恶性分类模型

姜林奇 宁春玉 余海涛

姜林奇, 宁春玉, 余海涛. 基于多尺度特征与通道特征融合的脑肿瘤良恶性分类模型[J]. 中国光学(中英文). doi: 10.37188/CO.2022-0067
引用本文: 姜林奇, 宁春玉, 余海涛. 基于多尺度特征与通道特征融合的脑肿瘤良恶性分类模型[J]. 中国光学(中英文). doi: 10.37188/CO.2022-0067
JIANG Lin-qi, NING Chun-yu, YU Hai-tao. Classification model based on fusion of multi-scale feature and channel feature for benign and malignant brain tumors[J]. Chinese Optics. doi: 10.37188/CO.2022-0067
Citation: JIANG Lin-qi, NING Chun-yu, YU Hai-tao. Classification model based on fusion of multi-scale feature and channel feature for benign and malignant brain tumors[J]. Chinese Optics. doi: 10.37188/CO.2022-0067

基于多尺度特征与通道特征融合的脑肿瘤良恶性分类模型

doi: 10.37188/CO.2022-0067
基金项目: 吉林省科技发展计划项目(No. 20200404219YY)
详细信息
    作者简介:

    姜林奇(1999—),女,内蒙古自治区呼伦贝尔市人,硕士研究生,2019年于南华大学获得学士学位,主要从事深度学习、图像分类等方面的研究。E-mail:631540532@qq.com

    宁春玉(1976—),女,吉林农安县人,硕士,副教授,硕士生导师,2005年于吉林大学获得计算机应用技术硕士学位,主要从事图像处理、模式识别的研究。E-mail:yeningcy@163.com

Classification model based on fusion of multi-scale feature and channel feature for benign and malignant brain tumors

Funds: Supported by the science and technology development project of Jilin province under grant number (No. 20200404219YY)
More Information
  • 摘要:

    针对脑肿瘤良恶性分类过程复杂、分类准确率不高等问题,提出了一种基于多尺度特征与通道特征融合的分类模型。该模型以ResNeXt网络为主干网络,首先,将基于空洞卷积的多尺度特征提取模块代替第一层卷积层,利用膨胀率获取不同感受野的图像信息,将全局特征与局部显著特征结合;其次,添加通道注意力机制模块,融合特征通道信息,提高对肿瘤区域的关注度,降低对冗余信息的关注度;最后,采用学习率的线性衰减策略、图像的标签平滑策略以及基于医学图像的迁移学习策略的组合优化策略提高模型的学习能力和泛化能力。在BraTS2017和BraTS2019数据集中进行实验,准确率分别达到98.11%和98.72%。与经典模型和其他先进方法相比,该分类模型能够有效地减少分类过程的复杂度,提高脑肿瘤良恶性分类的准确率。

     

  • 图 1  MDCA-ResNeXt网络结构

    Figure 1.  MDCA-ResNeXt network structure

    图 2  $ C = 32 $的ResNeXt结构[20]

    Figure 2.  ResNeXt structure with $ C = 32 $[20]

    图 3  不同膨胀率的空洞卷积

    Figure 3.  Dilated convolution with different dilation rates

    图 4  MD模块

    Figure 4.  MD module

    图 5  CA模块

    Figure 5.  CA module

    图 6  四种模态下的HGG图像

    Figure 6.  HGG images in four modalities

    图 7  四种模态下的LGG图像

    Figure 7.  LGG images in four modalities

    图 8  去除噪声前后对比图

    Figure 8.  Comparison before and after preprocessing

    图 9  分类结果评价图

    Figure 9.  Evaluation diagram of classification results

    图 10  HGG的原始图像和特征可视化图

    Figure 10.  Original image and feature visualization of HGG

    图 11  LGG的原始图像和特征可视化图

    Figure 11.  Original image and feature visualization of LGG

    表  1  实验数据集分布

    Table  1.   Distribution of experimental Datasets

    数据集肿瘤
    类别
    数据分布总数
    训练集测试集
    BraTS2017
    数据集
    HGG8402101050
    LGG9002251125
    BraTS2019
    数据集
    HGG10352601295
    LGG9152251140
    下载: 导出CSV

    表  2  优化前BraTS2017数据集的分类结果评价表

    Table  2.   Evaluation table of classification results on BraTS2017 before optimization

    网络ACC(%)SEN(%)SPE(%)PPV(%)NPV(%)
    ResNet89.15±1.8388.76±3.3889.51±3.3888.92±3.6789.60±2.58
    SENet90.44±3.2593.05±2.0989.25±7.4188.21±5.6993.19±1.77
    ResNeXt90.34±1.1489.23±3.3692.53±2.1891.85±1.9690.22±2.57
    MDCA-ResNeXt93.19±0.3593.05±1.6793.33±1.6992.91±1.5493.54±1.36
    下载: 导出CSV

    表  3  优化前BraTS2019数据集的分类结果评价表

    Table  3.   Evaluation table of classification results on BraTS2019 before optimization

    网络ACC(%)SEN(%)SPE(%)PPV(%)NPV(%)
    ResNet91.83±2.7393.31±2.1290.13±4.5991.71±3.6592.10±2.51
    SENet91.91±2.4290.54±5.6391.74±4.8393.00±3.7693.25±2.54
    ResNeXt93.57±1.5094.23±2.0292.80±2.9993.85±2.3193.35±2.15
    MDCA-ResNeXt94.10±1.4094.38±1.6793.78±2.3394.76±2.1293.60±1.59
    下载: 导出CSV

    表  4  优化后BraTS2017数据集的分类结果评价表

    Table  4.   Evaluation table of classification results on BraTS2017 after optimization

    网络ACC(%)SEN(%)SPE(%)PPV(%)NPV(%)
    Improved ResNet96.87±1.4996.76±0.7196.98±2.9096.84±2.9796.98±0.64
    Improved SENet97.56±1.0496.67±0.8998.40±1.4398.27±1.5296.94±0.83
    Improved ResNeXt97.98±1.3397.43±2.0698.49±1.2898.38±1.3997.63±1.88
    Improved MDCA-ResNeXt98.11±0.4197.43±0.2698.76±0.9198.66±0.9797.63±0.22
    下载: 导出CSV

    表  5  优化后BraTS2019数据集的分类结果评价表

    Table  5.   Evaluation table of classification results on BraTS2019 after optimization

    网络ACC(%)SEN(%)SPE(%)PPV(%)NPV(%)
    Improved ResNet97.03±1.9597.31±2.0296.62±3.0497.20±2.5596.91±2.23
    Improved SENet96.69±0.8894.38±1.8998.74±0.9398.62±1.0194.99±1.57
    Improved ResNeXt97.98±0.5797.69±0.4798.31±0.7398.53±0.6497.36±0.54
    Improved MDCA-ResNeXt98.72±0.3198.62±0.6498.85±0.5199.00±0.4498.41±0.73
    下载: 导出CSV

    表  6  先进方法分类结果对比表

    Table  6.   Comparison of classification results of advanced methods

    文献方法肿瘤分割数据集准确率(%)
    文献[7]HCS+ Multi-SVNNBraTs201493.00
    文献[15]Inception V3+POSBraTs201796.90
    文献[16]VGG16+ELMBraTs201796.90
    文献[17]3D CNN+VGG19+FNNBraTs201796.97
    文献[8]FBSOBraTs201893.85
    文献[19]3D U-NetBraTs201891.67
    本文方法Improved MDCA-ResNeXtBraTs201798.11
    本文方法Improved MDCA-ResNeXtBraTs201998.72
    下载: 导出CSV
  • [1] BRAY F, FERLAY J, SOERJOMATARAM I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA:A Cancer Journal for Clinicians, 2018, 68(6): 394-424. doi: 10.3322/caac.21492
    [2] BRAY F, FERLAY J, SOERJOMATARAM I, et al. Erratum: global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA:A Cancer Journal for Clinicians, 2020, 70(4): 313.
    [3] BAKAS S, AKBARI H, SOTIRAS A, et al. Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features[J]. Scientific Data, 2017, 4: 170117. doi: 10.1038/sdata.2017.117
    [4] MOHAN G, SUBASHINI M M. MRI based medical image analysis: survey on brain tumor grade classification[J]. Biomedical Signal Processing and Control, 2018, 39: 139-161. doi: 10.1016/j.bspc.2017.07.007
    [5] KHARRAT A, HALIMA M B, AYED M B. MRI brain tumor classification using Support Vector Machines and meta-heuristic method[C]. IEEE 2015 15th International Conference on Intelligent Systems Design and Applications (ISDA), IEEE, 2015: 446-451.
    [6] 徐立, 白金牛, 李磊民. 基于脑部MR图像GMM特征决策分类的肿瘤诊断[J]. 控制工程,2017,24(8):1718-1722. doi: 10.14107/j.cnki.kzgc.150604

    XU L, BAI J N, LI L M. Diagnosis of tumor in brain MR images based on GMM features and decision tree classifier[J]. Control Engineering of China, 2017, 24(8): 1718-1722. (in Chinese) doi: 10.14107/j.cnki.kzgc.150604
    [7] RAJU A R, SURESH P, RAO R R. Bayesian HCS-based multi-SVNN: a classification approach for brain tumor segmentation and classification using Bayesian fuzzy clustering[J]. Biocybernetics and Biomedical Engineering, 2018, 38(3): 646-660. doi: 10.1016/j.bbe.2018.05.001
    [8] NARMATHA C, ELJACK S M, TUKA A A R M, et al. . A hybrid fuzzy brain-storm optimization algorithm for the classification of brain tumor MRI images[J]. Journal of Ambient Intelligence and Humanized Computing,doi: 10.1007/s12652-020-02470-5.
    [9] 黄乐弘, 曹立华, 李宁, 等. 深度学习的空间红外弱小目标状态感知方法[J]. 中国光学,2020,13(3):527-536.

    HUANG L H, CAO L H, LI N, et al. A state perception method for infrared dim and small targets with deep learning[J]. Chinese Optics, 2020, 13(3): 527-536. (in Chinese)
    [10] 郑江鹏, 余平, 赵萌, 等. 利用低信噪比小样本太赫兹光谱实现心肌淀粉样变检测[J]. 中国光学,2022,15(3):443-453. doi: 10.37188/CO.2021-0223

    ZHENG J P, YU P, ZHAO M, et al. Detection of myocardial amyloidosis by a small number of terahertz spectra with low signal-to-noise ratio[J]. Chinese Optics, 2022, 15(3): 443-453. (in Chinese) doi: 10.37188/CO.2021-0223
    [11] 吴海滨, 魏喜盈, 王爱丽, 等. 八度卷积和双向门控循环单元结合的X光安检图像分类[J]. 中国光学,2020,13(5):1138-1146. doi: 10.37188/CO.2020-0073

    WU H B, WEI X Y, WANG A L, et al. X-ray security inspection images classification combined octave convolution and bidirectional GRU[J]. Chinese Optics, 2020, 13(5): 1138-1146. (in Chinese) doi: 10.37188/CO.2020-0073
    [12] 李宇, 刘雪莹, 张洪群, 等. 基于卷积神经网络的光学遥感图像检[J]. 光学 精密工程,2018,26(1):200-207. doi: 10.3788/OPE.20182601.0200

    LI Y, LIU X Y, ZHANG H Q, et al. Optical remote sensing image retrieval based on convolutional neural networks[J]. Optics and Precision Engineering, 2018, 26(1): 200-207. (in Chinese) doi: 10.3788/OPE.20182601.0200
    [13] 陈筱, 朱向冰, 吴昌凡, 等. 基于迁移学习与特征融合的眼底图像分类[J]. 光学 精密工程,2021,29(2):388-399. doi: 10.37188/OPE.20212902.0388

    CHEN X, ZHU X B, WU CH F, et al. Research on fundus image classification based on transfer learning and feature fusion[J]. Optics and Precision Engineering, 2021, 29(2): 388-399. (in Chinese) doi: 10.37188/OPE.20212902.0388
    [14] GNANASEKARAN V S, JOYPAUL S, SUNDARAM P M, et al. Deep learning algorithm for breast masses classification in mammograms[J]. IET Image Processing, 2020, 14(12): 2860-2868. doi: 10.1049/iet-ipr.2020.0070
    [15] SHARIF M I, LI J P, KHAN M A, et al. Active deep neural network features selection for segmentation and recognition of brain tumors using MRI images[J]. Pattern Recognition Letters, 2020, 129: 181-189. doi: 10.1016/j.patrec.2019.11.019
    [16] KHAN M A, ASHRAF I, ALHAISONI M, et al. Multimodal brain tumor classification using deep learning and robust feature selection: a machine learning application for radiologists[J]. Diagnostics, 2020, 10(8): 565. doi: 10.3390/diagnostics10080565
    [17] REHMAN A, KHAN M A, SABA T, et al. Microscopic brain tumor detection and classification using 3D CNN and feature selection architecture[J]. Microscopy Research and Technique, 2021, 84(1): 133-149. doi: 10.1002/jemt.23597
    [18] SEETHA J, RAJA S S. Brain tumor classification using Convolutional Neural Networks[J]. Biomedical and Pharmacology Journal, 2018, 11(3): 1457-1461. doi: 10.13005/bpj/1511
    [19] 赵尚义, 王远军. 基于3D CNN的脑胶质瘤分类算法[J]. 光学技术,2019,45(6):749-755.

    ZHAO SH Y, WANG Y J. Brain glioma classification algorithm based on 3D CNN[J]. Optical Technique, 2019, 45(6): 749-755. (in Chinese)
    [20] XIE S N, GIRSHICK R, DOLLÁR P, et al. . Aggregated residual transformations for deep neural networks[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2017: 5987-5995.
    [21] SZEGEDY C, VANHOUCKE V, IOFFE S, et al. . Rethinking the inception architecture for computer vision[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2016: 2818-2826.
    [22] HE K M, ZHANG X Y, REN SH Q, et al. . Deep residual learning for image recognition[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2016: 770-778.
    [23] YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[C]. 4th International Conference on Learning Representations, 2016. (查阅所有网上资料, 未找到本条文献出版者信息, 请联系作者确认并补充)
    [24] WOO S, PARK J, LEE J Y, et al. . CBAM: convolutional block attention module[C]. Proceedings of the 15th European Conference on Computer Vision, Springer, 2018: 3-19.
    [25] JIANG L Q, NING CH Y, LI J Y. Glioma classification framework based on SE-ResNeXt network and its optimization[J]. IET Image Processing, 2022, 16(2): 596-605. doi: 10.1049/ipr2.12374
    [26] CHENG J, HUANG W, CAO SH L, et al. Enhanced performance of brain tumor classification via tumor region augmentation and partition[J]. PLoS One, 2015, 10(12): e0140381.
    [27] VOVK U, PERNUS F, LIKAR B. A review of methods for correction of intensity inhomogeneity in MRI[J]. IEEE Transactions on Medical Imaging, 2007, 26(3): 405-421. doi: 10.1109/TMI.2006.891486
    [28] TUSTISON N J, AVANTS B B, COOK P A, et al. N4ITK: improved N3 bias correction[J]. IEEE Transactions on Medical Imaging, 2010, 29(6): 1310-1320. doi: 10.1109/TMI.2010.2046908
    [29] ZHONG ZH, ZHENG L, KANG G L, et al. . Random erasing data augmentation[C]. Proceedings of the 34th AAAI Conference on Artificial Intelligence, AAAI, 2020: 13001-13008.
    [30] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2018: 7132-7141.
  • 加载中
图(11) / 表(6)
计量
  • 文章访问数:  34
  • HTML全文浏览量:  28
  • PDF下载量:  48
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-04-12
  • 录用日期:  2022-08-24
  • 修回日期:  2022-05-03
  • 网络出版日期:  2022-08-24

目录

    /

    返回文章
    返回