A Study on the Robustness of Brain Functional Connectivity Model in Machine Learning Classification ——Taking the Resting-State Functional Magnetic Resonance Imaging to Localize Paroxysmal Side in TLE as Example
Yang Zekun1,2, Ge Manling1,2, Fu Xiaoxuan1,2 , Chen Shenghua1,2* , Zhang Fuyi1,2 , Guo Zhitong1,2, Zhang Zhiqiang3*
1(State Key Laboratory of Reliability and Intelligence of Electrical Equipment, Hebei University of Technology, Tianjin 300130, China) 2(Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability of Hebei Province, Hebei University of Technology, Tianjin 300130, China) 3(Department of Medical Imaging, Jinling Hospital, Nanjing University School of Medicine/ General Hospital of Eastern Theater, Nanjing 210002, China)
Abstract：Currently, machine learning has promoted the application of resting-state functional magnetic resonance imaging (rfMRI) in epilepsy, where the functional connectivity model of Pearson correlation (FC) has been widely applied as a traditional imaging algorithm. However, the classification stability of functional connectivity model is rarely studied in the machine learning. To address this issue, a FC-based index model specific to the healthy people was proposed in this work, the classification stability was studied by a random cross validation in the supervised machine learning models, and compared to the results of FC, aiming to provide a new algorithm in extracting FC features input into machine learning. The rfMRI data of a total of twenty patients of medial temporal lobe epilepsy with a positive indicator of hippocampus on structure MRI (equally involved in a group of left side and a group of right side), and a total of 142 healthy people from a connectome including Southwest Adult Lifespan Dataset (SALD) in the same age group were collected. A rfMRI FC-based index model was built up, specific to the healthy people, referred as FC-based specificity index model. Thus, every FC of each brain area in an individual patient could be scored, and the brain areas sensitive to paroxysmal side could be extracted by the ROC curve. The sensitivity analysis curve was taken as the functional bio-markers, whose indexes were assigned as the feature vectors to input into the supervised machine learning models such as probabilistic neural network (PNN) and support vector machine (SVM) to classify paroxysmal side. Additionally, the classification stability was validated by a random cross validation (10 times), and the linear correlation of feature vectors between sensitive brain areas and between patients were estimated to evaluate their interdependence, aiming to find out the underlying cause to affect the classification stability. Finally, the same procedures as above were fulfilled by the FC model instead of FC-based specificity index model, and the classification stability was compared. The AUC of the feature vector of FC was 0.76, and the feature vector of specificity index was 0.84. The classification sensitivity of the FC-based specificity index model was higher than that of FC. In addition, the classification accuracy of FC fluctuated strongly between 25%~100%, the variance was as high as 25.99%, and the average correlation coefficient of the feature vector was 0.67, which had a strong correlation; while the accuracy of the exponential model was stable at 75%~100%. The variance was as low as 7.10%, and the average correlation coefficient of the feature vector was 0.28, which was relatively small. When paroxysmal side of medial temporal lobe epilepsy was subjected to the resting-state functional magnetic resonance imaging in the machine learning, the proposed FC-based specificity index model performed robustly, much better than the traditional FC model such as Pearson correlation, and the larger correlation between the feature vectors formed by the traditional FC model might be the main cause that led to the low classification stability.
杨泽坤, 葛曼玲, 付晓璇, 陈盛华, 张夫一, 郭志彤, 张志强. 脑功能连接模型在机器学习中分类鲁棒性研究——以静息态功能磁共振定位癫痫发作侧为例[J]. 中国生物医学工程学报, 2021, 40(5): 521-530.
Yang Zekun, Ge Manling, Fu Xiaoxuan, Chen Shenghua, Zhang Fuyi, Guo Zhitong, Zhang Zhiqiang. A Study on the Robustness of Brain Functional Connectivity Model in Machine Learning Classification ——Taking the Resting-State Functional Magnetic Resonance Imaging to Localize Paroxysmal Side in TLE as Example. Chinese Journal of Biomedical Engineering, 2021, 40(5): 521-530.
[1] Gu Lian, Liang Baoyun, Chen Qing, et al. Prevalence of epilepsy in the People's Republic of China: A systematic review[J]. Epilepsy Res, 2013, 105(1-2):195-205. [2] Téllez-Zenteno JF, Hernández-Ronquillo L. A review of the epidemiology of temporal lobe epilepsy [J]. Epilepsy Res Treat, 2012, 2012(2012): 630853. [3] 崔婧轩,崔园,刘奇,等.颞叶癫痫患者全脑皮层厚度分析[J].中国生物医学工程学报,2017,36(6):755-758. [4] 张汉勇,孟庆芳,杜蕾,等.基于加权复杂网络度熵和的癫痫发作检测方法[J].中国生物医学工程学报,2019,38(3):273-280. [5] Berg AT, Berkovic SF, Brodie MJ, et al. Revised terminology and concepts for organization of seizures and epilepsies: Report of the ILAE commission on classification and terminology, 2005-2009 [J]. Epilepsia, 2010, 51(4):676-685. [6] Elson LS, Philippe R. MRI negative epilepsy: Evaluation and surgical management [M]. Cambridge: Cambridge University Press, 2015. [7] 陈富琴,张俊然,杨冰.基于模型的功能磁共振成像方法研究综述[J].中国生物医学工程学报,2016,35(3):340-347. [8] Jin Bo, Krishnan B, Adler S, et al. Automated detection of focal cortical dysplasia type II with surface-based magnetic resonance imaging postprocessing and machine learning [J]. Epilepsia, 2018, 59(5):982-992. [9] Bakken IJ, Axelson D, Kvistad KA, et al. Applications of neural network analyses to in vivo 1H magnetic resonance spectroscopy of epilepsy patients [J].Epilepsy Res,1999,35 (3): 245-252. [10] Jae SL, Dong SL, Seok-Ki K, et al. Localization of epileptogenic zones in F-18 FDG brain PET of patients with temporal lobe epilepsy using artificial neural network [J].IEEE Trans on Med Imaging, 2000,19(4): 347-355. [11] Kerr WT, Nguyen ST, Cho AY, et al. Computer-aided diagnosis and localization of lateralized temporal lobe epilepsy using interictal FDG-PET [J]. Front Neurol, 2013, 4:31. [12] Lopes R, Steinling M, Szurhaj W, et al. Fractal features for localization of temporal lobe epileptic foci using SPECT imaging [J]. Computers in Biol & Med, 2010, 40(5):469-477. [13] Keihaninejad S, Heckemann RA, Gousias IS, et al. Classification and lateralization of temporal lobe epilepsies with and without hippocampal atrophy based on whole-brain automatic MRI segmentation[J]. PLoS ONE, 2012,7 (4): e33096. [14] Kouhei K, Shiori A, Yuichi S, et al. Machine learning of DTI structural brain connectomes for lateralization of temporal lobe epilepsy[J]. Mag Reso in Med Sci, 2015, 15(1):121-129. [15] Gaizo JD, Mofrad N, Jensen J H, et al. Using machine learning to classify temporal lobe epilepsy based on diffusion MRI [J]. Brain and Beha, 2017, 7(4):e00801. [16] Jin SH, Chung CK. Electrophysiological resting-state biomarker for diagnosing mesial temporal lobe epilepsy with hippocampal sclerosis [J]. Epilepsy Res, 2017, 129:138-145. [17] Barron DS, Fox PT, Pardoe H, et al. Thalamic functional connectivity predicts seizure laterality in individual TLE patients: Application of a biomarker development strategy [J]. NeuroImage: Clinical, 2015,7: 273-280. [18] Zhang Wei, Muravina V, Azencott R, et al. Mutual information better quantifies brain network architecture in children with epilepsy [J].Comput Math Methods Med, 2018, 2018: 6142898. [19] Lavery MR, Acharya P, Sivo SA, et al. Number of predictors and multicollinearity: What are their effects on error and bias in regression? [J]. Comm in Stat-Simu and Comp, 2019, 48(1): 27-38. [20] 刘国旗. 多重共线性的产生原因及其诊断处理[J]. 合肥工业大学学报(自然科学版), 2001,24(4):607-610. [21] Wei Dongtao, Zhuang Kaixiang, Ai Lei, et al. Structural and functional brain scans from the cross-sectional Southwest University adult lifespan dataset [J]. Sci Data, 2018, 5:180134. [22] 吴寒, 张志强, 许强, 等. 间期痫样发放对内侧颞叶癫痫脑网络的影响[J]. 磁共振成像, 2015, 6(11): 801-806. [23] Woolrich MW, Jbabdi S, Patenaude B, et al. Bayesian analysis of neuroimaging data in FSL [J]. NeuroImage, 2009, 45(S1): S173-S186. [24] Jenkinson M, Bannister P, Brady M, et al. Improved optimization for the robust and accurate linear registration and motion correction of brain images [J]. NeuroImage, 2002, 17(2): 825-841. [25] Wang Danhong, Li Meiling, Wang Meiyun, et al. Individual-specific functional connectivity markers track dimensional and categorical features of psychotic illness [J]. Molecular Psychiatry, 2020, 25 (9): 2119-2129. [26] Yeo BT, Krienen FM, Sepulcre J, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity [J]. Journal of Neurophysiol, 2011, 106(3): 1125-1165. [27] Smith SM. Fast robust automated brain extraction [J]. Human Brain Mapping, 2002,17(3): 143-155. [28] Yan Chaogan, Wang Xindi, Zuo Xinian, et al. DPABI: Data processing & analysis for (resting-state) brain imaging [J]. Neuroinform, 2016, 14(3):339-351. [29] Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain [J]. NeuroImage, 2002, 15(1): 273-289. [30] Fieller EC, Hartley HO, Pearson ES. Tests for rank correlation coefficients [J]. Biometrika, 1957, 44(3-4):470-481. [31] Xia Mingrui, Wang Jinhui, He Yong. BrainNet Viewer: A network visualization tool for human brain connectomics [J]. PLoS ONE, 2013, 8: e68910. [32] Specht DF. Probabilistic neural networks [J]. Neural Networks, 1990, 3(1):109-118. [33] Wasserman PD. Advanced methods in neural computing [M]. New York: John Wiley & Sons, Inc. 1993. [34] Wu CH, Tzeng GH, Lin RH. A novel hybrid genetic algorithm for kernel function and parameter optimization in support vector regression [J]. Expert Systems with Applications, 2009, 36 (3): 4725-4735. [35] 徐晓明. SVM参数寻优及其在分类中的应用[D]. 大连:大连海事大学,2014. [36] Bharath RD, Panda R, Raj J, et al. Machine learning identifies “rsfMRI epilepsy networks” in temporal lobe epilepsy [J]. Euro Radi, 2019, 29 (7): 3496-3505. [37] Mo J, Liu Z, Sun K, et al. Automated detection of hippocampal sclerosis using clinically empirical and radiomics feature s[J]. Epilepsia, 2019, 60(12): 2519-2529. [38] Jeong SW, Lee Sang K, Hong KS, et al. Prognostic factors for the surgery for mesial temporal lobe epilepsy: Longitudinal analysis [J]. Epilepsia, 2005, 46 (8): 1273-1279.
