基于脑结构网络分析技术对卒中后认知障碍的预测研究

Predictive Value of Topological Attributes of Brain Structural Network for Post Stroke Cognitive Impairment

目的 探究卒中事件后患者大脑结构网络变化与卒中后认知功能障碍(PSCI)的关系。方法 前瞻性搜集江苏大学附属医院2020年6月至2022年5月收治的急性脑梗死患者209例,同期搜集无认知障碍病史的健康对照者50名(HC组)。所有研究对象均接受系统MRI扫描和相关量表评估,并进行12个月的纵向随访。依据随访时是否发生认知障碍将患者纳入卒中后认知障碍组(PSCI组)和卒中后认知正常组(PSNCI组)。采用方差分析探究脑结构网络属性的组间差异。构建Logistic回归模型分析差异性脑结构网络属性对于PSCI的预测效能。基于Pearson相关分析探究脑结构网络属性与蒙特利尔认知量表(MoCA)评分之间的关联。结果 方差分析结果显示3组间特征路径长度(F=3.47,P=0.033)、归一化聚类系数(F=3.60,P=0.028)、归一化特征路径长度(F=9.47,P<0.001)、全局效率(F=9.41,P<0.001)、左侧额中回节点效率(F=6.01,P=0.002)、右侧海马节点效率(F=8.24,P<0.001)、右侧眶部额上回节点效率(F=4.31,P=0.015)存在显著差异。Logistic回归模型显示基线时归一化特征路径长度(OR=1.87,95%CI 1.84~1.93,P<0.001)和右侧海马节点效率(OR=0.71,95%CI 0.67~0.79,P<0.001)对PSCI的发生具有良好的预测效能,绘制受试者工作特征(ROC)曲线,曲线下面积(AUC)分别为0.72、0.73。二者联合指标预测效能更佳,AUC为0.82。相关分析结果显示PSCI组归一化特征路径长度(r=-0.61,P=0.002)和右侧海马节点效率(r=0.59,P=0.006)与MoCA评分具有显著相关性。结论 存在归一化特征路径长度和右侧海马节点效率改变的脑卒中患者发生认知障碍的风险更大。

Objective To explore the relationship between changes in brain structural network and the development of post-stroke cognitive impairment(PSCI)in patients after a stroke incident.Methods Prospective collection of data from 209 patients with acute ischemic stroke admitted to Jiangsu University Affiliated Hospital from June 2020 to May 2022.A healthy control group(HC)consisting of 50 individuals without a history of cognitive impairment was also collected during the same period.All participants underwent systematic magnetic resonance imaging(MRI)scans and relevant scale assessments,followed by a 12-month longitudinal follow-up.According to the presence of cognitive impairment during follow-up,patients were divided into the post-stroke cognitive impairment group(PSCI)and the post-stroke normal cognitive group(PSNCI).Analysis of variance was used to investigate differences in brain structural network attributes between the groups.Logistic regression models were constructed to investigate the predictive efficacy of brain structural network attributes for PSCI.Pearson correlation analysis was used to explore the correlations between brain structural network attributes and Montreal Cognitive Assessment(MoCA)scores.Results Analysis of variance results showed significant differences in feature path length(F=3.47,P=0.033),normalized clustering coefficient(F=3.60,P=0.028),normalized feature path length(F=9.47,P<0.001),global efficiency(F=9.41,P<0.001),Left middle frontal gyrus node efficiency(F=6.01,P=0.002),right hippocampal node efficiency(F=8.24,P<0.001),and Right orbital superior frontal gyrus node efficiency(F=4.31,P=0.015)among the three groups.The logistic regression model demonstrated that baseline normalized feature path length(OR=1.87,95%CI 1.84-1.93,P<0.001)and right hippocampal node efficiency(OR=0.71,95%CI 0.67-0.79,P<0.001)had good predictive efficacy for PSCI,with area under the receiver operating characteristic curve(AUC)of 0.72 and 0.73,respectively.The combined index had a better predictive efficacy with an AUC of 0.82.The cor-relation analysis results showed that the normalized feature path length(r=-0.61,P=0.002)and right hippocampal node efficiency(r=0.59,P=0.006)of the PSCI group were significantly correlated with the MoCA score.Conclusion Stroke patients with changes in normalized feature path length and right hippocampal node efficiency have a higher risk of cognitive impairment.