【作者】 李倩；

【导师】 龚启勇；

【作者基本信息】 四川大学 ， 影像医学与核医学（专业学位）， 2021， 硕士

【摘要】 目的:基于静息态脑功能磁共振数据和控制理论,探索首发未治疗精神分裂症患者脑功能网络认知控制可控性的异常,促进对精神分裂症认知功能缺陷神经机制的进一步认识,并结合可控性指标和机器学习方法建立精神分裂症患者个体化预测的分类模型,实现特异且客观的脑影像生物标志物对精神分裂症进行临床辅助诊断的应用转化。材料和方法:本研究分两部分进行。第一部分纳入133名首发未治疗精神分裂症患者及与之年龄、性别、教育年限相匹配的135名正常对照,运用3.0T磁共振扫描仪对所有受试者进行扫描,并采集所有受试者的静息态脑功能磁共振数据以构建脑功能网络。基于控制理论计算脑功能网络的可控性指标,包括平均可控性及模式可控性,使用置换检验分析该指标在患者与正常对照之间的组间差异,并使用斯皮尔曼相关分析方法探究患者脑功能网络可控性指标与临床症状的相关性。组间比较及相关性分析均采用错误发现率进行多重比较校正,P<0.05被认为具有统计学意义。第二部分纳入134名首发未治疗精神分裂症患者及148名正常对照。运用支持向量机及支持向量机-递归特征消除算法,以可控性指标作为特征,构建首发未治疗精神分裂症个体化预测的分类模型。结果:我们发现与正常对照相比,首发未治疗精神分裂症患者背侧前扣带的脑功能网络平均可控性增加(P=0.0227),而患者背侧前扣带的脑功能网络模式可控性较正常对照减低(P=0.0227)。在患者组内,背侧前扣带的平均可控性及模式可控性与临床症状和病程无显著相关性。基于脑功能网络可控性构建的分类模型对精神分裂症患者及正常对照具有一定的区分能力,准确率为0.68(P=0.0001),敏感性为0.76,特异性为0.58,曲线下面积为0.69。最具有区分能力的特征为背侧前扣带、右侧内侧前额叶、双侧楔前叶及左侧颞上回的脑功能网络平均可控性,其中患者内侧前额叶脑功能网络平均可控性与抑郁症状呈正相关关系(r=0.24,P=0.03)。结论:本研究排除了药物及行为学治疗可能带来的对脑功能的影响,针对首发未治疗的精神分裂症患者进行了脑网络可控性和诊断模型构建的研究,上述结果表明背侧前扣带脑功能网络可控性异常在精神分裂症的神经病理机制中发挥重要作用,此外,脑功能网络可控性有利于对精神分裂症患者进行个体化预测及辅助临床诊断。更多还原

【Abstract】 Objective:The study aimed to investigate the altered controllability of functional brain networks based on the data of resting-state functional magnetic resonance imaging(rf MRI)and control theory,providing further insight into the neuropathological basis of cognitive deficits in schizophrenia.In addition,we integrated the controllability metrics and machine learning to differentiate patients with schizophrenia from healthy controls(HCs),in order to promote the clinical application of the specific and objective brain imaging biomarkers in schizophrenia and aid the clinical diagnosis.Materials and Methods:The current research included two sections.In the first part,we recruited 133first-episode untreated patients with schizophrenia and 135 age,sex,and education years matched HCs.rf MRI data were acquired from all participants using a 3.0 T MRI scanner.The controllability measurements of functional brain networks such as average controllability and modal controllability were calculated.The differences in brain controllability between patients and HCs were explored using permutation test and the associations of brain controllability and clinical symptoms in patients were tested by Spearman correlation analyses.The False Discovery Rate was applied to conduct corrections for multiple between-group comparisons and correlation analyses and P < 0.05 was considered statistically significant.In the second part,there were134 first-episode untreated patients with schizophrenia and 148 HCs.The support vector machine(SVM)and SVM-recursive feature elimination(SVM-RFE)combined with the controllability of functional brain networks were performed to establish the individual prediction model,to help identify the patients with schizophrenia in clinical practice.Results:Compared with HCs,the patients with schizophrenia showed increased average controllability(P = 0.0227)and decreased modal controllability(P = 0.0227)in the dorsal anterior cingulate cortex.There were no statistically significant correlations between the controllability of dorsal anterior cingulate cortex and clinical symptoms and illness durations in patients.The model established on the brain controllability and SVM could identify individuals with an accuracy of 0.67(P = 0.0001),a sensitivity of 0.76,a specificity of 0.58,and an area under curve of 0.69.Furthermore,the most discriminant features were the average controllability in the dorsal anterior cingulate cortex,right medial prefrontal lobe,bilateral precuneus and left superior temporal gyrus,and the average controllability of the right medial prefrontal lobe was positively correlated with depressive symptom in patients(r = 0.24,P = 0.03).Conclusion:Our study focused on the altered controllability of functional brain networks and diagnostic model establishment in first-episode untreated patients with schizophrenia without treatment effects bias.Thus,above mentioned findings suggested that the altered controllability in the dorsal anterior cingulate cortex might play a critical role in the neuropathological mechanisms of schizophrenia and the controllability of functional brain networks could be beneficial to individual prediction and aid clinical diagnosis of schizophrenia.更多还原