{"title":"系统性红斑狼疮异常脑功能网络:图论、基于网络的统计和机器学习研究。","authors":"Yifan Yang, Ru Bai, Shuang Liu, Shu Li, Ruotong Zhao, Xiangyu Wang, Yuqi Cheng, Jian Xu","doi":"10.1093/braincomms/fcaf130","DOIUrl":null,"url":null,"abstract":"<p><p>Systemic lupus erythematosus patients' brain functional network impairments are incompletely clarified. This study investigates the brain functional network topological alterations in systemic lupus erythematosus and the application of machine learning to the classification of systemic lupus erythematosus and healthy controls. Resting-state functional MRI data from 127 systemic lupus erythematosus patients and 102 healthy controls were used. The pre-processing process involved using automated anatomical labelling atlas to compute time series data for 116 brain regions. A functional connectivity network was then created by assessing the Pearson correlation between the time series of these brain regions. The GRETNA toolbox was used to compute the difference in topological attributes between groups. Variations in regional networks among groups were evaluated using non-parametric permutation tests that rely on network-based statistical analysis. With the functional connectivity network metrics as features and network-based statistical analysis as the feature selection method, network-based statistical analysis Predict software was used to classify systemic lupus erythematosus from controls by support vector machine. The subnets that contributed the most to systemic lupus erythematosus classification were also identified. For global indicators, the systemic lupus erythematosus group exhibited significantly lower values for the normalized clustering coefficient (<i>P</i> = 0. 0317) and small-world index (<i>P</i> = 0.0364) compared to the healthy controls group. After false discovery rate correction, the differences in Betweeness Centrality, Degree Centrality, Node Efficiency, Node Local Efficiency and other local indexes between the two groups were not retained. No correlation was found between clinical data and network indicators. Systemic lupus erythematosus group had a significantly reduced connection with a 12-node, 11-edge subnetwork (<i>P</i> = 0.024). In conclusion, systemic lupus erythematosus patients exhibit suboptimal global brain functional connectivity network topology and the presence of a subnetwork with abnormally reduced connectivity.</p>","PeriodicalId":93915,"journal":{"name":"Brain communications","volume":"7 2","pages":"fcaf130"},"PeriodicalIF":4.1000,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11979335/pdf/","citationCount":"0","resultStr":"{\"title\":\"Abnormal brain functional networks in systemic lupus erythematosus: a graph theory, network-based statistic and machine learning study.\",\"authors\":\"Yifan Yang, Ru Bai, Shuang Liu, Shu Li, Ruotong Zhao, Xiangyu Wang, Yuqi Cheng, Jian Xu\",\"doi\":\"10.1093/braincomms/fcaf130\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Systemic lupus erythematosus patients' brain functional network impairments are incompletely clarified. This study investigates the brain functional network topological alterations in systemic lupus erythematosus and the application of machine learning to the classification of systemic lupus erythematosus and healthy controls. Resting-state functional MRI data from 127 systemic lupus erythematosus patients and 102 healthy controls were used. The pre-processing process involved using automated anatomical labelling atlas to compute time series data for 116 brain regions. A functional connectivity network was then created by assessing the Pearson correlation between the time series of these brain regions. The GRETNA toolbox was used to compute the difference in topological attributes between groups. Variations in regional networks among groups were evaluated using non-parametric permutation tests that rely on network-based statistical analysis. With the functional connectivity network metrics as features and network-based statistical analysis as the feature selection method, network-based statistical analysis Predict software was used to classify systemic lupus erythematosus from controls by support vector machine. The subnets that contributed the most to systemic lupus erythematosus classification were also identified. For global indicators, the systemic lupus erythematosus group exhibited significantly lower values for the normalized clustering coefficient (<i>P</i> = 0. 0317) and small-world index (<i>P</i> = 0.0364) compared to the healthy controls group. After false discovery rate correction, the differences in Betweeness Centrality, Degree Centrality, Node Efficiency, Node Local Efficiency and other local indexes between the two groups were not retained. No correlation was found between clinical data and network indicators. Systemic lupus erythematosus group had a significantly reduced connection with a 12-node, 11-edge subnetwork (<i>P</i> = 0.024). In conclusion, systemic lupus erythematosus patients exhibit suboptimal global brain functional connectivity network topology and the presence of a subnetwork with abnormally reduced connectivity.</p>\",\"PeriodicalId\":93915,\"journal\":{\"name\":\"Brain communications\",\"volume\":\"7 2\",\"pages\":\"fcaf130\"},\"PeriodicalIF\":4.1000,\"publicationDate\":\"2025-04-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11979335/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Brain communications\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1093/braincomms/fcaf130\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"2025/1/1 0:00:00\",\"PubModel\":\"eCollection\",\"JCR\":\"Q1\",\"JCRName\":\"CLINICAL NEUROLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Brain communications","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1093/braincomms/fcaf130","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2025/1/1 0:00:00","PubModel":"eCollection","JCR":"Q1","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}
Abnormal brain functional networks in systemic lupus erythematosus: a graph theory, network-based statistic and machine learning study.
Systemic lupus erythematosus patients' brain functional network impairments are incompletely clarified. This study investigates the brain functional network topological alterations in systemic lupus erythematosus and the application of machine learning to the classification of systemic lupus erythematosus and healthy controls. Resting-state functional MRI data from 127 systemic lupus erythematosus patients and 102 healthy controls were used. The pre-processing process involved using automated anatomical labelling atlas to compute time series data for 116 brain regions. A functional connectivity network was then created by assessing the Pearson correlation between the time series of these brain regions. The GRETNA toolbox was used to compute the difference in topological attributes between groups. Variations in regional networks among groups were evaluated using non-parametric permutation tests that rely on network-based statistical analysis. With the functional connectivity network metrics as features and network-based statistical analysis as the feature selection method, network-based statistical analysis Predict software was used to classify systemic lupus erythematosus from controls by support vector machine. The subnets that contributed the most to systemic lupus erythematosus classification were also identified. For global indicators, the systemic lupus erythematosus group exhibited significantly lower values for the normalized clustering coefficient (P = 0. 0317) and small-world index (P = 0.0364) compared to the healthy controls group. After false discovery rate correction, the differences in Betweeness Centrality, Degree Centrality, Node Efficiency, Node Local Efficiency and other local indexes between the two groups were not retained. No correlation was found between clinical data and network indicators. Systemic lupus erythematosus group had a significantly reduced connection with a 12-node, 11-edge subnetwork (P = 0.024). In conclusion, systemic lupus erythematosus patients exhibit suboptimal global brain functional connectivity network topology and the presence of a subnetwork with abnormally reduced connectivity.