Please wait a minute...
浙江大学学报(工学版)  2022, Vol. 56 Issue (11): 2232-2240    DOI: 10.3785/j.issn.1008-973X.2022.11.014
计算机技术     
基于显著稀疏强关联的脑功能连接分类方法
李明1,2,3(),段立娟1,2,3,*(),王文健1,2,3,恩擎4
1. 北京工业大学 信息学部,北京 100124
2. 可信计算北京市重点实验室,北京 100124
3. 信息安全等级保护关键技术国家工程实验室,北京 100124
4. 卡尔顿大学 计算机学院,人工智能与机器学习实验室,加拿大 渥太华K1S 5B6
Brain functional connections classification method based on significant sparse strong correlation
Ming LI1,2,3(),Li-juan DUAN1,2,3,*(),Wen-jian WANG1,2,3,Qing EN4
1. Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
2. Beijing Key Laboratory of Trusted Computing, Beijing 100124, China
3. National Engineering Laboratory for Critical Technologies of Information Security Classified Protection, Beijing 100124, China
4. Artificial Intelligence and Machine Learning (AIML) Lab, School of Computer Science, Carlton University, Ottawa K1S 5B6, Canada
 全文: PDF(949 KB)   HTML
摘要:

针对脑功能连接数据维度过高、冗余特征过多影响神经网络分类准确率的问题,提出一种基于显著稀疏强关联的脑功能连接分类方法. 该方法利用显著特征稀疏模块对原始特征进行筛选增强;采用稀疏强关联特征上下文融合模块对不同感受野内的显著特征信息进行聚合;使用全连接神经网络进行分类预测. 在ABIDE以及ADHD-200数据集上的实验结果表明,所提方法相较于现有的脑功能连接分类算法在准确率上分别提升了10.41%和12.50%. 重要特征的可视化结果表明所提方法能准确定位与疾病相关的脑区,具有一定实际应用价值.

关键词: 显著特征识别特征筛选特征增强特征融合脑功能连接分类    
Abstract:

A brain functional connectivity classification method based on significant sparse strong correlation was proposed to solve problems that data dimension of brain functional connectivity is too high and many redundant features affect the accuracy of neural network classification. In this method, the salient features sparse module was used to filter and enhance the original features. The sparse strong correlation feature context fusion module was used to aggregate the salient feature information in different receptive fields. The fully connected neural network was used for classification prediction. The results on ABIDE and ADHD-200 datasets showed that the accuracy of brain functional connection classification algorithm was improved by 10.41% and 12.5% respectively compared with the existing brain functional connection classification algorithm. The visualization results of the important features show that the proposed method can accurately locate the brain regions related to disease, which has a certain practical value.

Key words: significant feature identification    feature screening    feature fusion    feature enhancement    classification of brain functional connections
收稿日期: 2022-03-25 出版日期: 2022-12-02
CLC:  TP 391  
基金资助: 国家自然科学基金资助项目(62176009, 62106065);北京市教委重点项目(KZ201910005008)
通讯作者: 段立娟     E-mail: Liming@emails.bjut.edu.cn;ljduan@bjut.edu.cn
作者简介: 李明(1995—),男,硕士生,从事脑科学研究. orcid.org/0000-0003-4294-6614. E-mail: Liming@emails.bjut.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章  
李明
段立娟
王文健
恩擎

引用本文:

李明,段立娟,王文健,恩擎. 基于显著稀疏强关联的脑功能连接分类方法[J]. 浙江大学学报(工学版), 2022, 56(11): 2232-2240.

Ming LI,Li-juan DUAN,Wen-jian WANG,Qing EN. Brain functional connections classification method based on significant sparse strong correlation. Journal of ZheJiang University (Engineering Science), 2022, 56(11): 2232-2240.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2022.11.014        https://www.zjujournals.com/eng/CN/Y2022/V56/I11/2232

图 1  基于显著稀疏强关联的脑功能连接分类方法整体框架(SSSC)
图 2  显著特征稀疏模块结构图
图 3  稀疏强关联特征上下文特征融合模块
数据集 年龄段 患者被试(男/女) 正常被试(男/女) 合计
ASD (6,12] 130/21 114/28 293
(12,17] 156/23 159/43 381
(17,64] 157/18 162/24 361
ADHD200 [7,26] 288/74 306/279 947
表 1  不同数据的性别以及年龄阶段分布人数
数据集 模型 ACC SEN SPE AUC
ABIDE ICA-SVM 63.24 59.51 66.55 63.03
PCA-SVM 58.77 54.34 62.5 58.42
CNN-EW 64.99 62.75 66.76 64.75
Full-BiLSTM 69.29 56.65 79.67 68.16
SCNN 70.94 66.93 73.52 70.23
SSSC(ours) 81.35 78.47 87.88 86.18
ADHD200 ICA-SVM 62.23 36.18 81.7 66.02
PCA-SVM 56.08 42.58 65.84 54.21
CNN-EW 68.71 42.67 84.55 63.61
Full-BiLSTM 69.11 59.36 74.52 66.94
SCNN 69.97 65.28 73.63 69.46
SSSC(ours) 82.47 86.81 75.26 81.43
表 2  衡量分类性能的评价指标结果对比
ASD vs TC
Model ACC/% SEN/% SPE/% AUC/%
SSSC-5% 64.87 13.16 85.22 56.57
SSSC-10% 81.35 78.47 87.88 86.18
SSSC-20% 83.42 76.55 85.69 82.12
SSSC-50% 70.15 19.77 96.15 57.97
SSSC-80% 53.57 0 100 50.00
SSSC-all 53.57 0 100 50.00
ADHD vs TC
Model ACC/% SEN/% SPE/% AUC/%
SSSC-5% 67.76 20.31 93.61 70.16
SSSC-10% 82.47 86.81 75.26 81.43
SSSC-20% 82.55 84.55 73.12 76.12
SSSC-50% 72.11 23.55 89.11 63.22
SSSC-80% 63.15 15.26 92.33 54.22
SSSC-all 61.77 0 100 50.00
表 3  不同脑功能连接数量下的消融实验结果对比
图 4  融合不同尺度关联特征的分类性能对比
图 5  重要脑功能连接的原始功能连接强度
图 6  在不同数据集中得到的重要脑功能连接对分类任务的贡献程度
1 ITAHASHI T, YAMADA T, WATANABE H, et al Altered network topologies and hub organization in adults with autism: a resting-state fMRI study[J]. PloSone, 2014, 9 (4): e94115
doi: 10.1371/journal.pone.0094115
2 LIANG X, WANG J Human connectome: structural and functional brain networks[J]. Chinese Science Bulletin, 2010, 55 (16): 1565- 1583
doi: 10.1360/972009-2150
3 SPORNS O Structure and function of complex brain networks[J]. Dialogues in Clinical Neuroscience, 2013, 15 (3): 247- 262
doi: 10.31887/DCNS.2013.15.3/osporns
4 STAM C Modern network science of neurological disorders[J]. Nature Reviews Neuroscience, 2014, 15 (10): 683
doi: 10.1038/nrn3801
5 WEE C, YAP P, ZHANG D, et al Group-constrained sparse fMRI connectivity modeling for mild cognitive impairment identification[J]. Brain Structure and Function, 2014, 219 (2): 641- 656
doi: 10.1007/s00429-013-0524-8
6 CORTES C, VAPNIK V Support-vector networks[J]. Machine Learning, 1995, 20 (3): 273- 297
7 ZHOU J, LU Z, SUN J, et al. Feafiner: biomarker identification from medical data through feature generalization and selection[C]// Acm Sigkdd. Chicago: ACM, 2013: 1034-1042.
8 KHAZAEE A, EBRAHIMZADEH A, BABAJANIFEREMI A Application of advanced machine learning methods on resting-state fMRI network for identification of mild cognitive impairment and Alzheimer's disease[J]. Brain Imaging and Behavior, 2016, 10 (3): 799- 817
doi: 10.1007/s11682-015-9448-7
9 CRADDOCK R, HOLTZHEIMER P, HU X, et al Disease state prediction from resting state FMRI[J]. Neuro Image, 2009, 47 (6): S80
10 MESZLÉNYI R, BUZA K, VIDNYÁNSZKY Z Resting state fMRI functional connectivity-based classification using a convolutional neural network architecture[J]. Frontiers in Neuroinformatics, 2017, 11: 61
doi: 10.3389/fninf.2017.00061
11 KAWAHARA J, BROWN C, MILLER S, et al Brainnet CNN: convolutional neural networks for brain networks towards predicting neurodevelopment[J]. NeuroImage, 2017, 146: 1038- 1049
doi: 10.1016/j.neuroimage.2016.09.046
12 JIE B, LIU M, LIAN C, et al Designing weighted correlation kernels in convolutional neural networks for functional connectivity based brain disease diagnosis[J]. Medical Image Analysis, 2020, 63
13 JEON E, KANG E, LEE J, et al. Enriched representation learning in resting-state fMRI for early MCI diagnosis[C]// MICCAI. Lima: Springer, 2020: 397-406.
14 KIM J, CALHOUN V, SHIM E, et al Deep neural network with weight sparsity control and pre-training extracts hierarchical features and enhances classification performance: evidence from whole-brain resting-state functional connectivity patterns of schizophrenia[J]. Neuroimage, 2016, 124: 127- 146
doi: 10.1016/j.neuroimage.2015.05.018
15 LI H, PARIKH N, HE L A novel transfer learning approach to enhance deep neural network classification of brain functional connectomes[J]. Frontiers in Neuroscience, 2018, 12: 491
doi: 10.3389/fnins.2018.00491
16 HUANG J, ZHOU L, WANG L, et al. Integrating functional and structural connectivities via diffusion-convolution-bilinear neural network[C]// Miccai. Shenzhen: Springer, 2019: 691-699.
17 WANG M, HUANG J, LIU M, et al Modeling dynamic characteristics of brain functional connectivity networks using resting-state functional MRI[J]. Medical Image Analysis, 2021, 71: 102063
doi: 10.1016/j.media.2021.102063
18 SUBBARAJU V, SURESH M, SUNDARAM S, et al Identifying differences in brain activities and an accurate detection of autism spectrum disorder using resting state functional-magnetic resonance imaging: a spatial filtering approach[J]. Medical Image Analysis, 2017, 35: 375- 389
doi: 10.1016/j.media.2016.08.003
19 BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate [C]// Iclr. San Diego, 2015.
20 LUONG M, PHAM H, MANNING C. Effective approaches to attention-based neural machine translation [C]// Emnlp. Lisbon: the association for computational Linguistics, 2015: 1412−1421.
21 YIN W, SCHÜTZE H, XIANG B, et al Abcnn: Attention-based convolutional neural network for modeling sentence pairs[J]. Transactions of the Association for Computational Linguistics, 2016, 4: 259- 272
doi: 10.1162/tacl_a_00097
22 VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]// Advances in Neural Information Processing Systems. Long Beach: NIPS Poundation, 2017: 5998-6008.
23 CHEN M, RADFORD A, CHILD R, et al. Generative pretraining from pixels[C]// International Conference on Machine Learning. Virtual Event: PNVR, 2020: 1691-1703.
24 DOSOVITSKIYET A, BEYER L, KOLESNIKOV A, et al. An image is worth 16x16 words: Transformers for image recognition at scale [C]// Iclr. Virtual Event: Open Review Net, 2021.
25 GAO M, TSAI F, LEE C. Learning a phenotypic-attribute attentional brain connectivity embedding for ADHD classification using rs-fMRI[C]// Embc. Montreal: IEEE, 2020: 5472-5475.
26 MA J, ZHU X, YANG D, et al. Attention-guided deep graph neural network for longitudinal alzheimer’s disease analysis[C]// Miccai. Lima: Springer, 2020: 387-396.
27 CRADDOCK C, BENHAJALI Y, CHU C, et al The neuro bureau preprocessing initiative: open sharing of preprocessed neuroimaging data and derivatives[J]. Frontiers in Neuro Informatics, 2013, 7: 27
28 BELLEC P, CHU C, CHOUINARDDECORTE F, et al The neuro bureau ADHD-200 preprocessed repository[J]. Neuroimage, 2017, 144: 275- 286
doi: 10.1016/j.neuroimage.2016.06.034
29 JI J, CHEN Z, YANG C Convolutional neural network with sparse strategies to classify dynamic functional connectivity[J]. IEEE Journal of Biomedical and Health Informatics, 2021, 26 (3): 1219- 1228
30 CUI Z, GONG G The effect of machine learning regression algorithms and sample size on individualized behavioral prediction with functional connectivity features[J]. Neuroimage, 2018, 178: 622- 637
doi: 10.1016/j.neuroimage.2018.06.001
31 XING X, JI J, YAO Y. Convolutional neural network with element-wise filters to extract hierarchical topological features for brain networks[C]// IEEE International Conference on Bioinformatics and Biomedicine. Modrid: IEEE, 2018: 780-783.
32 YAN W, ZHANG H, SUI J, et al. Deep chronnectome learning via full bidirectional long short-term memory networks for MCI diagnosis[C]// International Conference on Medical Image Computing and Computer-assisted Intervention. Granada: Springer, Cham, 2018: 249-257.
33 HEINSFELD A, FRANCO A, CRADDOCK R, et al Identification of autism spectrum disorder using deep learning and the ABIDE dataset[J]. NeuroImage: Clinical, 2018, 17: 16- 23
doi: 10.1016/j.nicl.2017.08.017
[1] 张娜,戚旭磊,包晓安,吴彪,涂小妹,金瑜婷. 基于优化预测定位的单阶段目标检测算法[J]. 浙江大学学报(工学版), 2022, 56(4): 783-794.
[2] 黄新宇,游帆,张沛,张昭,张柏礼,吕建华,徐立臻. 基于多分类及特征融合的静默活体检测算法[J]. 浙江大学学报(工学版), 2022, 56(2): 263-270.
[3] 张云佐,郭威,蔡昭权,李文博. 联合多尺度与注意力机制的遥感图像目标检测[J]. 浙江大学学报(工学版), 2022, 56(11): 2215-2223.
[4] 杨栋杰,高贤君,冉树浩,张广斌,王萍,杨元维. 基于多重多尺度融合注意力网络的建筑物提取[J]. 浙江大学学报(工学版), 2022, 56(10): 1924-1934.
[5] 陈智超,焦海宁,杨杰,曾华福. 基于改进MobileNet v2的垃圾图像分类算法[J]. 浙江大学学报(工学版), 2021, 55(8): 1490-1499.
[6] 刘嘉诚,冀俊忠. 基于宽度学习系统的fMRI数据分类方法[J]. 浙江大学学报(工学版), 2021, 55(7): 1270-1278.
[7] 陈岳林,田文靖,蔡晓东,郑淑婷. 基于密集连接网络和多维特征融合的文本匹配模型[J]. 浙江大学学报(工学版), 2021, 55(12): 2352-2358.
[8] 陈雪云,夏瑾,杜珂. 基于多线型特征增强网络的架空输电线检测[J]. 浙江大学学报(工学版), 2021, 55(12): 2382-2389.
[9] 郑浦,白宏阳,李伟,郭宏伟. 复杂背景下的小目标检测算法[J]. 浙江大学学报(工学版), 2020, 54(9): 1777-1784.
[10] 周登文,田金月,马路遥,孙秀秀. 基于多级特征并联的轻量级图像语义分割[J]. 浙江大学学报(工学版), 2020, 54(8): 1516-1524.
[11] 林泓,卢瑶瑶. 聚焦难样本的区分尺度的文字检测方法[J]. 浙江大学学报(工学版), 2019, 53(8): 1506-1516.
[12] 成翔昊,达飞鹏,汪亮. 基于融合约束局部模型的三维人脸特征点定位[J]. 浙江大学学报(工学版), 2019, 53(4): 770-776.
[13] 王万良,杨胜兰,赵燕伟,李卓蓉. 基于条件边界平衡生成对抗网络的河流表面流速估测[J]. 浙江大学学报(工学版), 2019, 53(11): 2118-2128.
[14] 鲍必赛, 伍健荣, 楼晓俊, 刘海涛. 基于二维特征矩阵的特征融合算法[J]. J4, 2012, 46(11): 2081-2088.
[15] 姚拓中 项志宇 刘济林. 基于图像多特征融合的野外水体障碍物检测[J]. , 2009, 43(4): 605-609.