您的位置：山东大学 -> 科技期刊社 -> 《山东大学学报（工学版）》

山东大学学报 (工学版) ›› 2018, Vol. 48 ›› Issue (6): 1-7.doi: 10.6040/j.issn.1672-3961.0.2018.205

• 机器学习与数据挖掘 •    下一篇

基于多尺度残差神经网络的阿尔茨海默病诊断分类

刘振丙1(),方旭升1,杨辉华1,2,蓝如师1   

  1. 1. 桂林电子科技大学电子工程与自动化学院, 广西 桂林 541000
    2. 北京邮电大学自动化学院, 北京 100876
  • 收稿日期:2018-05-31 出版日期:2018-12-20 发布日期:2018-12-26
  • 作者简介:刘振丙(1980—)，男，山东济宁人，教授，博士生导师，博士，主要研究方向为机器学习与图像处理.E-mail:zbliu@guet.edu.cn
  • 基金资助:
    国家自然科学基金项目(61562013);国家自然科学基金项目(61866009);广西自然科学基金(2017GXNFDA198025)

The diagnosis of Alzheimer's disease classification based on multi-scale residual neutral network

Zhenbing LIU1(),Xusheng FANG1,Huihua YANG1,2,Rushi LAN1   

  1. 1. School of Electronic Engineering and Automation, Guilin University of Electronic Technology, Guilin 541000, Guangxi, China
    2. School of Automation, Beijing University of Electronic Technology, Beijing 100876, China
  • Received:2018-05-31 Online:2018-12-20 Published:2018-12-26
  • Supported by:
    国家自然科学基金项目(61562013);国家自然科学基金项目(61866009);广西自然科学基金(2017GXNFDA198025)

RichHTML

150

PDF (PC)

2488

摘要：

提出多尺度残差神经网络(multi-scale resnet, MSResnet)。采用不同大小的卷积核对图像进行多尺度信息采集,并对神经网络进行残差学习,避免网络退化。对核磁共振图像(magnetic resonance imaging, MRI)进行标准化处理,利用MSResnet模型在阿尔茨海默症(Alzheimer's disease, AD)和正常受试者(normal control, NC)获得的分类准确率为99.41%,在AD和轻度认知障碍(mild cognitive impairment, MCI)获得分类准确率为97.35%。与已有的算法相比,本研究提出的算法的分类准确率得到了明显的提高。

关键词: 多尺度残差神经网络, 核磁共振图像, 阿尔茨海默病, 灰度标准化

Abstract:

A multi-scale resnet (MSResnet) method was proposed in this paper, which employed multi-scale convolution kernel to extract multi-scale information of structural magnetic resonance imaging MRI, and carried out residual learning for neural network, so as to avoid network degradation. After the gray scale standardization of MRI, the 99.41% classification precision was obtained by using the MSResnet model between Alzheimer's disease (AD) and normal control (NC), and the classification accuracy between AD and mild cognitive impairment (MCI) was 97.35%. Compared with the existing approaches, the algorithm proposed in this paper improved the classification accuracy significantly.

Key words: MSResnet, magnetic resonance imaging, Alzheimer's disease, gray scale standardization

中图分类号: 

  • TP399

图1

残差学习:残差块"

图2

多尺度残差网络"

图3

多尺度残差学习块"

表1

多尺度残差模型的框架"

类型 卷积核 尺寸/步长 输出尺寸
卷积 64 7×7/2 112×112×64
最大池化 64 3×3/2 56×56×64
块1 64 Incep+Incep 56×56×64
块2 64 Incep+Incep 56×56×64
块3 128 3×3/2+Incep 28×28×128
块4 128 Incep+Incep 28×28×128
块5 256 3×3/2+Incep 14×14×256
块6 256 Incep+Incep 14×14×256
块7 512 3×3/2+Incep 7×7×512
块8 512 Incep+Incep 7×7×512
平均池化 512 7×7/1 1×1×512

表2

数据集的统计特性"

Subject AD MCI NC
Female/male 98/89 178/113 166/143
Age 76.90±6.15 75.26±7.74 76.71±6.79
MMSE ≤24 ≥24 ≥26
CDR ≥1.0 0.5~1.0 0

图4

灰度转换示意图"

图5

图像标准化前后图"

图6

模型的测试精度和损失"

图7

各模型的ROC曲线"

表3

每一种模型的准确率"

%
模型 AD vs NC AD vs MCI MCI vs NC AD vs NC vs MCI
BA+Resnet 96.51 93.11 84.93 82.36
ST+Resnet 97.45 94.28 85.55 83.51
BA+MSResnet 98.32 95.15 86.70 83.87
ST+SVM 90.16 88.19 80.26 75.26
ST+ELM 91.35 86.46 81.79 77.07
ST+Alexnet 93.49 91.92 83.04 80.30
ST+VGG16 95.83 93.76 84.16 81.25
ST+GoogLenet 97.86 94.49 86.05 83.57
ST+MSResnet 99.41 97.35 87.75 85.53

表4

与其他模型的准确率、敏感度和特异性的比较"

%
模型 AD vs NC AD vs MCI MCI vs NC
准确率 敏感度 特异度 准确率 敏感度 特异度 准确率 敏感度 特异度
Alexnet[17] 96.14 93.57 100.00 90.52 85.11 94.20 84.80 88.02 80.62
VGG[16] 98.33 97.78 98.89 93.89 90.00 97.78 91.67 91.11 92.22
GoogLenet[15] 98.84
CNN[14] 93.08 94.92 92.67 86.30 88.46 84.55 83.30 80.99 85.55
DBN[20] 90.09 94.10 86.12 84.00 89.12 79.12 83.14 95.09 67.26
MSResnet 99.41 97.89 99.86 97.35 93.73 98.51 87.75 84.08 88.49
1 张柏雯, 林岚, 吴水才, 等. 深度学习在轻度认知障碍转化与分类中的应用分析[J]. 医疗卫生装备, 2017, 38 (9): 105- 111.
ZHANG Baiwen , LIN Lan , WU Shuicai , et al. Application of deep learning to mild cognitive impairment conversion and classification[J]. Chinese Medical Equipment Journal, 2017, 38 (9): 105- 111.
2 ORTIZ A , RRIZ J M , REZ J , et al. LVQ-SVM based CAD tool applied to structural MRI for the diagnosis of the Alzheimer's disease[J]. Pattern Recognition Letters, 2013, 34 (14): 1725- 1733.
doi: 10.1016/j.patrec.2013.04.014
3 LIU F , SUK H I , WEE C Y , et al. High-order graph matching based feature selection for Alzheimer's disease identification[J]. Medical Image Computing and Computer-Assisted Intervention, 2013, 16 (2): 311- 318.
4 YANG J , PAN P , SONG W , et al. Voxelwise meta-analysis of gray matter anomalies in Alzheimer's disease and mild cognitive impairment using anatomic likelihood estimation[J]. Journal of the Neurological Sciences, 2012, 316 (1-2): 21- 29.
doi: 10.1016/j.jns.2012.02.010
5 李昕, 童隆正, 周晓霞, 等. 基于MR图像三维纹理特征的阿尔茨海默病和轻度认知障碍的分类[J]. 中国医学影像技术, 2011, 27 (5): 1047- 1051.
LI Xin , TONG Longzheng , ZHOU Xiaoxia , et al. Classification of 3D texture feature based on MRI image in discrimination of Alzheimer disease and mild cognitive impairment from normal controls[J]. Chinese Journal of Medical Imaging Technology, 2011, 27 (5): 1047- 1051.
6 何其佳, 刘振丙, 徐涛, 等. 基于LBP和极限学习机的脑部MR图像分类[J]. 山东大学学报(工学版), 2017, 47 (2): 86- 93.
HE Qijia , LIU Zhenbing , XU Tao , et al. MR image classification based on LBP and extreme learning machine[J]. Journal of Shandong University (Engineering Science), 2017, 47 (2): 86- 93.
7 LIU Z , XU T , MA C , et al. T-test based Alzheimer's disease diagnosis with multi-feature in MRIs[J]. Multimedia Tools & Applications, 2018, (2): 1- 17.
8 ZHANG D , WANG Y , ZHOU L , et al. Multimodal classification of Alzheimer's disease and mild cognitive impairment[J]. Neuroimage, 2011, 55 (3): 856.
doi: 10.1016/j.neuroimage.2011.01.008
9 ZHU X , SUK H I , WANG L , et al. A novel relational regularization feature selection method for joint regression and classification in AD diagnosis[J]. Medical Image Analysis, 2015, 75 (6): 570- 577.
10 KRIZHEVSKY A , HINTON G E , SUTSKEVER I . ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25 (2): 2012.
11 SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].[2017-06-20].https://arxiv.org/pdf/1049.1556v6.pdf.
12 SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston, USA: 2015: 1-9.
13 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.
14 LIU F, SHEN C. Learning deep convolutional features for MRI based Alzheimer's disease classification [EB/OL].[2017-04-30]. https://arXiv.org/pdf./:1404.3366v1.pdf.
15 SAMAN S, GHASSEM T. Deep AD: Alzheimer's disease classification via deep convolutional neural networks using MRI and FMRI[EB/OL].[2016-03-29]. https://arXiv.org/1603.08631v1.pdf.
16 BILLONES C D, DEMETRIA O J, HOSTALLERO D E, et al. DemNet: a convolutional neural network for the detection of alzheimer's disease and mild cognitive impairment[C]//Region 10 Conference(TENCON). Singapore: IEEE, 2016: 3724-3727.
17 吕鸿蒙, 赵地, 迟学斌. 基于增强AlexNet的深度学习的阿尔茨海默病的早期诊断[J]. 计算机科学, 2017, 44 (增刊1): 50- 60.
LYU Hongmeng , ZHAO Di , CHI Xuebin . Deep learning for early diagnosis of Alzheimer's disease based on intensive AlexNet[J]. Computer Science, 2017, 44 (Suppl.1): 50- 60.
18 LIN M, CHEN Q, YAN S. Network in network[EB/OL]. [2016-01-20]. http://arxiv.org/pdf/1312.-4000v3.pdf.
19 MADABHUSHI A , UDUPA J K , SOUZA A . Generalized scale: theory, algorithms, and application to image inhomogeneity correction[J]. Proceedings of SPIE-The International Society for Optical Engineering, 2006, 101 (2): 100- 121.
20 ORTIZ A , MUNILLA J , GÓRRIZ J M , et al. Ensembles of deep learning architectures for the early diagnosis of the Alzheimer's disease[J]. International Journal of Neural Systems, 2016, 26 (07): 1650025.
doi: 10.1142/S0129065716500258
[1] 王海军,柳明. 克服灰度不均匀性的脑MR图像分割及去偏移场模型[J]. 山东大学学报(工学版), 2011, 41(3): 36-41.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
[1] 王素玉,艾兴,赵军,李作丽,刘增文 . 高速立铣3Cr2Mo模具钢切削力建模及预测[J]. 山东大学学报(工学版), 2006, 36(1): 1 -5 .
[2] 李 侃 . 嵌入式相贯线焊接控制系统开发与实现[J]. 山东大学学报(工学版), 2008, 38(4): 37 -41 .
[3] 孔祥臻,刘延俊,王勇,赵秀华 . 气动比例阀的死区补偿与仿真[J]. 山东大学学报(工学版), 2006, 36(1): 99 -102 .
[4] 来翔 . 用胞映射方法讨论一类MKdV方程[J]. 山东大学学报(工学版), 2006, 36(1): 87 -92 .
[5] 余嘉元1 , 田金亭1 , 朱强忠2 . 计算智能在心理学中的应用[J]. 山东大学学报(工学版), 2009, 39(1): 1 -5 .
[6] 陈瑞,李红伟,田靖. 磁极数对径向磁轴承承载力的影响[J]. 山东大学学报(工学版), 2018, 48(2): 81 -85 .
[7] 王波,王宁生 . 机电装配体拆卸序列的自动生成及组合优化[J]. 山东大学学报(工学版), 2006, 36(2): 52 -57 .
[8] 李可,刘常春,李同磊 . 一种改进的最大互信息医学图像配准算法[J]. 山东大学学报(工学版), 2006, 36(2): 107 -110 .
[9] 季涛,高旭,孙同景,薛永端,徐丙垠 . 铁路10 kV自闭/贯通线路故障行波特征分析[J]. 山东大学学报(工学版), 2006, 36(2): 111 -116 .
[10] 浦剑1 ,张军平1 ，黄华2 . 超分辨率算法研究综述[J]. 山东大学学报(工学版), 2009, 39(1): 27 -32 .
版权所有 © 2018 《山东大学学报(工学版)》编辑部　
地址: 济南市山大南路27号（250100）　 电话: 0531-88366735　 E-mail: xbgxb@sdu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发