北京生物医学工程

基于静息态功能性磁共振成像的个体认知多标签分类

Multi-label classification of individual cognition based on resting state functional magnetic resonance imaging

作者：吴怊慧

单位：北京交通大学计算机与信息技术学院（北京１０００４４）

关键词：静息态功能性磁共振成像；个体认知；机器学习；多标签分类；算法对比

分类号：R318.04

出版年·卷·期（页码）：2020·39·2（138-144）

摘要：

目的与单标签分类相比,多标签分类在现实世界中更为常见。在神经影像学领域,对个体认知单标签分类的研究有很多,但是却没有关于个体认知多标签分类的研究。本研究尝试运用机器学习多标签分类算法,利用静息态fMRI数据,对个体认知进行多标签分类研究。方法基于390 名(≥18 岁)成年人的静息态fMRI数据,通过5种具有代表性的机器学习多标签分类算法: ML-kNN、hMuLab、LIFT、ML-LOC 和 GLOCAL 对个人认知进行多标签分类,采用十折交叉验证进行训练和测试,并用多标签分类常用的评价指标进行结果的检验。结果这5种算法都能用于静息态fMRI的多标签分类研究,且经过算法对比发现hMuLab算法的分类效果最好。结论本研究成功地将多标签分类算法应用于神经影像学领域并对个体认知进行多标签分类,且通过算法对比发现样本的邻域信息对分类结果很重要。

Objective Compared with single label classification, multi-label classification is more common in the real world.In the field of neuroimaging,there are many studies on single label classification of individual cognition,and no research on multi-label classification of individual cognition.We try to use machine learning multi-label classification algorithm and resting state fMRI data to study the multi-label classification of individual cognition.Methods Based on the resting state fMRI data of 390 adults ( ≥18 years old), five representative algorithms of machine learning multi-label classification: ML-kNN,hMuLab,LIFT,ML-LOC and GLOCAL were used to classify individual cognition,ten fold cross validation was used for training and testing, and commonly used evaluation criteria for multi-label classification were used to test the results.Results The results of each evaluation criterion showed that these five algorithms could be used in the multi-label classification of resting state fMRI, and the hMuLab algorithm had the best classification effect compared with other algorithms.Conclusions We successfully applied the multi-label classification algorithm to multi-label classification of individual cognition in the field of neuroimaging, and found that the sample neighborhood information was important by comparing the five algorithms.

参考文献：

[ 1 ] Zhang ML,Zhou ZH.A review on multi-label learning algorithms [ J]. IEEE Transactions on Knowledge and Data Engineering, 2014,26(8):1819-1837.

[ 2 ] Gibaja E,Ventura S.Multi-label learning: a review of the state of the art and ongoing research[J].Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery,2014,4(6): 411-444.

[ 3 ] McCallum AK.Multi-Label Text Classification with a Mixture Model Trained by EM [EB/ OL].[1999-12-10]http: / / www.kyriakides. net/ CBCL/ references/ Papers/ mccallum99 multilabel.pdf.

[ 4 ] Ueda N,Saito K.Parametric mixture models for multi-labeled text [C] / / Advances in Neural Information Processing Systems 15. Vancouver, British Columbia, Canada: NIPS 2002, 2002, 15: 737-744.

[ 5 ] Boutell MR,Luo J, Shen X, et al. Learning multi-label scene classification[J].Pattern Recognition,2004,37(9):1757-1771.

[ 6 ] Yu R,Zhang H,An L,et al.Connectivity strength-weighted sparse group representation-based brain network construction for MCI classification [ J ]. Human Brain Mapping, 2017, 38 ( 5 ): 2370-2383.

[ 7 ] Li Y,Liu J,Luo M, et al. Structural connectivity guided sparse effective connectivity for MCI identification [ M ] / / Machine Learning in Medical Imaging.Switzerland: Springer,Cham,2017, 10541: 299-306.

[ 8 ] Wee CY,Yap PT,Li W,et al.Enriched white matter connectivity networks for accurate identification of MCI patients [ J ]. NeuroImage,2011,54(3):1812-1822.

[ 9 ] Zhang ML,Zhou ZH. A k-nearest neighbor based algorithm for multi-label classification [ C ] / / 2005 IEEE International Conference on Granular Computing. Beijing: IEEE, 2005, 2: 718-721.

[10] Zhang ML,Zhou ZH. ML-KNN: A lazy learning approach to multi-label learning [ J ]. Pattern Recognition, 2007, 40 ( 7 ): 2038-2048.

[11] Wang P,Ge R, Xiao X, et al. hMuLab: a biomedical hybrid MUlti-LABel classifier based on multiple linear regression [ J]. IEEE/ ACM Transactions on Computational Biology and Bioinformatics,2017,14(5):1173-1180.

[12] Zhang ML.LIFT: multi-label learning with label-specific features [C] / / International Joint Conference on Artificial Intelligence. Barcelona,Catalonia,Spain: AAAI Press,2011:1609-1614.

[13] Huang SJ,Zhou ZH. Multi?label learning by exploiting label correlations locally [ C ] / / Proceedings of the 26th AAAI Conference on Artificial Intelligence. Toronto, Ontario, Canada: AAAI.12,2012: 949-955.

[14] Zhu Y,Kwok JT,Zhou ZH. Multi-label learning with global and local label correlation[ J].IEEE Transactions on Knowledge and Data Engineering,2018,34: 1081-1094.

[15] Smith SM,Jenkinson M, Woolrich MW, et al. Advances in functional and structural MR image analysis and implementation as FSL[J].NeuroImage,2004,23(Suppl 1): S208-S219.

[16] Jenkinson M,Beckmann CF, Behrens TE, et al. FSL [ J ]. NeuroImage,2012,62(2): 782-790.

[17] Power JD,Mitra A, Laumann TO, et al. Methods to detect, characterize,and remove motion artifact in resting state fMRI[ J]. NeuroImage,2014,84(1):320-341.

[18] Shine JM,Koyejo O, Poldrack RA. Temporal metastates are associated with differential patterns of time-resolved connectivity, network topology, and attention [ J]. Proceedings of the National Academy of Sciences of the United States of America,2016,113 (35): 9888-9891.

[19] Dosenbach NU,Nardos B,Cohen AL,et al.Prediction of individual brain maturity using fMRI [ J ]. Science, 2010, 329 ( 5997 ): 1358-1361.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录