基于特权信息集成学习的精神分裂症单模态神经影像计算机辅助诊断

沈璐; 王倩婷; 施俊

doi:10.7507/1001-5515.201905029

基于特权信息集成学习的精神分裂症单模态神经影像计算机辅助诊断

doi: 10.7507/1001-5515.201905029

详细信息

通讯作者:
施俊，Email：junshi@shu.edu.cn

计量
- 文章访问数: 394
- HTML全文浏览量: 245
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2019-05-10
- 修回日期: 2019-11-27
- 发布日期: 2020-03-17

Single-modal neuroimaging computer aided diagnosis for schizophrenia based on ensemble learning using privileged information

More Information

Corresponding author: SHI Jun, Email: junshi@shu.edu.cn

摘要

摘要: 神经影像技术目前已经应用于精神分裂症的诊断。为了提升基于单模态神经影像的精神分裂症计算机辅助诊断（CAD）的性能，本文提出一种基于特权信息学习（LUPI）分类器的集成学习算法。该算法首先对单模态数据采用极限学习机-自编码器（ELM-AE）进行特征二次学习，然后通过随机映射算法将高维特征随机分成多个子空间，并进行两两组合形成源领域和目标领域数据对，用于训练多个支持向量机+（SVM+）弱分类器，最终通过集成学习获得一个强分类器，实现有效的模式分类。本算法在公开的精神分裂症神经影像数据库中进行了实验，包括结构磁共振成像和功能磁共振成像数据。结果表明该算法取得了最优的诊断结果，其在基于结构磁共振成像诊断的分类精度、敏感性和特异性分别可以达到 72.12% ± 8.20%、73.50% ± 15.44% 和 70.93% ± 12.93%，而基于功能磁共振成像诊断的分类精度、敏感性和特异性分别为 72.33% ± 8.95%、68.50% ± 16.58%、75.73% ± 16.10%。本文算法的主要创新点在于克服了传统的 LUPI 分类器需要额外的特权信息模态的不足，可以直接应用于单模态数据分类问题，而且还提升了分类性能，因此具有较为广泛的应用前景。
- 单模态神经影像 /
- 精神分裂症 /
- 深度学习 /
- 极限学习机-自编码器 /
- 集成特权信息学习
Abstract: Neuroimaging technologies have been applied to the diagnosis of schizophrenia. In order to improve the performance of the single-modal neuroimaging-based computer-aided diagnosis (CAD) for schizophrenia, an ensemble learning algorithm based on learning using privileged information (LUPI) was proposed in this work. Specifically, the extreme learning machine based auto-encoder (ELM-AE) was first adopted to learn new feature representation for the single-modal neuroimaging data. Random project algorithm was then performed on the learned high-dimensional features to generate several new feature subspaces. After that, multiple feature pairs were built among these subspaces to work as source domain and target domain, respectively, which were used to train multiple support vector machine plus (SVM+) classifier. Finally, a strong classifier is learned by combining these SVM+ classifiers for classification. The proposed algorithm was evaluated on a public schizophrenia neuroimaging dataset, including the data of structural magnetic resonance imaging (sMRI) and functional MRI (fMRI). The results showed that the proposed algorithm achieved the best diagnosis performance. In particular, the classification accuracy, sensitivity and specificity of the proposed algorithm were 72.12% ± 8.20%, 73.50% ± 15.44% and 70.93% ± 12.93%, respectively, on the sMRI data, and it also achieved the classification accuracy of 72.33% ± 8.95%, sensitivity of 68.50% ± 16.58% and specificity of 75.73% ± 16.10% on the fMRI data. The proposed algorithm overcomes the problem that the traditional LUPI methods need the additional privileged information modality as source domain. It can be directly applied to the single-modal data for classification, and also can improve the classification performance. Therefore, it suggests that the proposed algorithm will have wider applications.
- single-modal neuroimaging /
- schizophrenia /
- deep learning /
- extreme learning machine based auto-encoder /
- ensemble learning using privileged information

HTML全文

图 1 结合 ELM-AE 的集成 SVM+分类算法流程图

Figure 1. Flowchart of ensemble SVM+ classifier with ELM-AE

下载: 全尺寸图片幻灯片

图 2 ELM-AE 的网络结构图

Figure 2. Network structure of ELM-AE

下载: 全尺寸图片幻灯片

图 3 基于 sMRI 数据不同算法的 ROC 曲线

Figure 3. ROC curves of different algorithms based on sMRI data

下载: 全尺寸图片幻灯片

图 4 基于 fMRI 数据不同算法的 ROC 曲线

Figure 4. ROC curves of different algorithms based on fMRI data

下载: 全尺寸图片幻灯片

表 1 基于 sMRI 数据不同算法分类的结果

Table 1. Classification results of different algorithms based on sMRI data

算法	精度（%）	敏感度（%）	特异性（%）	AUC
EA-SVM	69.41 ± 10.25	67.90 ± 16.06	70.92 ± 15.53	0.727 2
EA-Voting-SVM	70.46 ± 8.38	68.00 ± 16.54	72.53 ± 13.49	−
EA-MDO-SVM	70.59 ± 7.16	70.50 ± 17.26	71.24 ± 11.82	−
EA-MKB-SVM	70.73 ± 8.43	70.50 ± 21.11	70.75 ± 16.63	0.733 6
EA-Voting-SVM+	71.42 ± 7.05	74.00 ± 14.40	70.00 ± 10.87	−
EA-MDO-SVM+	71.69 ± 8.87	72.50 ± 14.43	70.78 ± 12.69	−
EA-MKB-SVM+	72.12 ± 8.20	73.50 ± 15.44	70.93 ± 12.93	0.767 7

下载: 导出CSV

表 2 基于 fMRI 数据不同算法分类的结果

Table 2. Classification results of different algorithms based on fMRI data

算法	精度（%）	敏感度（%）	特异性（%）	AUC
EA-SVM	68.42 ± 10.92	67.50 ± 16.57	69.20 ± 15.60	0.707 1
EA-Voting-SVM	70.01 ± 9.10	64.00 ± 16.66	75.42 ± 17.38	−
EA-MDO-SVM	70.96 ± 8.38	64.00 ± 16.66	77.16 ± 16.21	−
EA-MKB-SVM	70.95 ± 7.29	65.00 ± 18.33	74.80 ± 14.28	0.728 0
EA-Voting-SVM+	71.91 ± 9.68	69.50 ± 16.77	74.09 ± 15.52	−
EA-MDO-SVM+	71.43 ± 9.13	69.50 ± 16.17	73.20 ± 15.61	−
EA-MKB-SVM+	72.33 ± 8.95	68.50 ± 16.58	75.73 ± 16.10	0.735 7

下载: 导出CSV

参考文献(20)

[1]	管丽丽, 杜立哲, 马弘. 精神分裂症的疾病负担. 中国心理卫生杂志, 2012, 26(12): 913-919. doi: 10.3969/j.issn.1000-6729.2012.12.008
[2]	Birur B, Kraguljac N V, Shelton R C, et al. Brain structure, function, and neurochemistry in schizophrenia and bipolar disorder-a systematic review of the magnetic resonance neuroimaging literature. NPJ Schizophrenia, 2017, 3(1): 15. doi: 10.1038/s41537-017-0013-9
[3]	Shi Jun, Zheng Xiao, Li Yan, et al. Multimodal neuroimaging feature learning with multimodal stacked deep polynomial networks for diagnosis of Alzheimer’s disease. IEEE J Biomed Health Inform, 2018, 22(1): 173-183. doi: 10.1109/JBHI.2017.2655720
[4]	Shi Jun, Xue Zeyu, Dai Yakang, et al. Cascaded multi-column RVFL+ classifier for single-modal neuroimaging-based diagnosis of Parkinson’s disease. IEEE Trans Biomed Eng, 2019, 66(8): 2362-2371. doi: 10.1109/TBME.2018.2889398
[5]	Kasun L L C, Zhou H, Huang G B, et al. Representational learning with extreme learning machine for big data. IEEE Intell Syst, 2013, 28(6): 31-34.
[6]	Tang Jiexiong, Deng Chenwei, Huang Guangbin. Extreme learning machine for multilayer perceptron. IEEE Trans Neural Netw Learn Syst, 2016, 27(4): 809-821. doi: 10.1109/TNNLS.2015.2424995
[7]	Zhang Junjie, Yin Jie, Zhang Qi, et al. Robust sound event classification with bilinear multi-column ELM-AE and two-stage ensemble learning. EURASIP Journal on Audio, Speech, and Music Processing, 2017: 11.
[8]	Vapnik V, Vashist A. A new learning paradigm: learning using privileged information. Neural Netw, 2009, 22(5/6): 544-557.
[9]	Duan Lixin, Xu Yanwu, Li Wen, et al. Incorporating privileged genetic information for fundus image based glaucoma detection// International Conference on Medical Image Computing and Computer-Assisted Intervention. Boston: Springer, 2014: 204-211.
[10]	Zheng Xiao, Shi Jun, Ying Shihui, et al. Improving single-modal neuroimaging based diagnosis of brain disorders via boosted privileged information learning framework// International Workshop on Machine Learning in Medical Imaging. Athens: Springer, 2016: 95-103.
[11]	Zheng X, Shi J, Zhang Q, et al. Improving MRI-based diagnosis of Alzheimer’s disease via an ensemble privileged information learning algorithm// 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017). Melbourne: IEEE, 2017: 456-459.
[12]	Huang Guangbin, Zhu Qinyu, Siew C K. Extreme learning machine: Theory and applications. Neurocomputing, 2006, 70(1/3): 489-501.
[13]	Zhu P F, Zhang L, Hu Q H, et al. Multi-scale patch based collaborative representation for face recognition with margin distribution optimization// European Conference on Computer Vision. Firenze: Springer, 2012: 822-835.
[14]	Yang Fan, Lu Huchuan, Yang M H. Robust visual tracking via multiple kernel boosting with affinity constraints. IEEE Transactions on Circuits and Systems for Video Technology, 2014, 24(2): 242-254. doi: 10.1109/TCSVT.2013.2276145
[15]	Xu Lai, Groth K M, Pearlson G, et al. Source-based morphometry: the use of independent component analysis to identify gray matter differences with application to schizophrenia. Hum Brain Mapp, 2009, 30(3): 711-724. doi: 10.1002/hbm.20540
[16]	Jafri M J, Pearlson G D, Stevens M, et al. A method for functional network connectivity among spatially independent resting-state components in schizophrenia. Neuroimage, 2008, 39(4): 1666-1681. doi: 10.1016/j.neuroimage.2007.11.001
[17]	Esposito F, Scarabino T, Hyvarinen A, et al. Independent component analysis of fMRI group studies by self-organizing clustering. Neuroimage, 2005, 25(1): 193-205. doi: 10.1016/j.neuroimage.2004.10.042
[18]	Hyvärinen A, Oja E. Independent component analysis: algorithms and applications. Neural Netw, 2000, 13(4/5): 411-430.
[19]	Calhoun V D, Adali T, Pearlson G D, et al. A method for making group inferences from functional MRI data using independent component analysis. Hum Brain Mapp, 2001, 14(3): 140-151. doi: 10.1002/hbm.1048
[20]	Silva R F, Castro E, Gupta C N, et al. The tenth annual MLSP competition: Schizophrenia classification challenge// 2014 IEEE International Workshop on Machine Learning for Signal Processing (MLSP). Reims: IEEE, 2014: 1-6.