肝豆状核变性患者默认模式网络变化及模型构建

Alterations in the default mode network and model construction in patients with Wilson’s disease

目的:基于独立成分分析(ICA),探讨肝豆状核变性(WD)患者默认模式网络(DMN)功能连接(FC)的改变及其与临床神经精神特征之间的关系。方法:将2021年1月-2021年12月在本院就诊的85例肝豆状核变性患者和年龄、性别相匹配的85例健康志愿者(HC组)纳入本研究。对每例被试采用统一肝豆状核变性评分量表(UWDRS)进行评估,包括神经功能症状(UWDRS-N)和精神症状(UWDRS-P)评分,并根据UWDRS评分将患者分为神经精神症状重度组(>10分)和轻度组(≤10分)。使用3.0T磁共振BOLD序列采集静息态fMRI数据,采用ICA方法提取DMN内各体素的FC值,并在HC组与WD组之间进行比较,对有差异脑区的FC值与临床量表评分进行Pearson相关性分析。以DMN内所有体素的FC值为特征变量采用支持向量机(SVM)的方法构建分类模型,包括正常组与WD组以及轻度与重度WD组。结果:与对照组比较,WD组的DMN内表现出广泛的FC值减低,包括前默认模式网络(aDMN)内的左内侧前额叶皮层(L_MPFC)和左侧前扣带回(L_ACC),以及后DMN(pDMN)内的左侧角回(L_ANG)、楔前叶(PCUN)、左侧顶下小叶(L_IPG)和左侧后扣带回(L_PCC)。aDMN内的L_MPFC、L_ACC和pDMN内的PCUN的FC值与UWDRS-N评分呈负相关,aDMN内的L_MPFC、L_PCC和pDMN内的L_IPG的FC值与UWDRS-P评分呈负相关。采用SVM构建的二分类器,在鉴别WD与HC组时的符合率为80.23%,AUC为0.865;在鉴别轻度与重度WD组的符合率为70.89%,AUC为0.723。结论:WD患者的默认模式网络中存在广泛的功能连接减低,其可能是导致患者出现神经精神症状(如高阶认知障碍)的潜在神经病理机制。基于DMN的FC值构建的SVM分类器可提高对WD疾病及其病情转归的评估效能。

Objective:To explore the function connectivity(FC)alteration of default mode network(DMN)in patients with Wilson’s disease(WD)and its relationship with clinical neuropsychiatric features based on independent component analysis(ICA).Methods:From January 2021 to December 2021,eighty-five patients with WD and age-and gender-matched 85 healthy controls(HC group)in our hospital were included in this study.Each subject was evaluated using the Unified Wilson's Disease Rating Scale(UWDRS),including neurological symptoms examination(UWDRS-N)and psychiatric symptoms examination(UWDRS-P).According to the UWDRS total score(UWDRS-TS),the patients with WD were divided into two groups:mild group(UWDRS-TS≤10)and severe group(UWDRS-TS>10).The resting-state fMRI data were acquired using blood oxygen level dependence(BOLD)sequence at a 3.0T magnetic resonance scanner,and the functional connectivity(FC)values of each voxel within the DMN were extracted using the ICA method,and the difference in FC values within the DMN between WD group and HC group were compared,and the correlation between the FC values of the brain areas with statistical difference in WD group and the clinical scale score were analyzed using Pearson correlation analysis.The FC values of all the voxels within the DMN were analyzed as feature variable using the method of support vector machine(SVM)for constructing classification models of normal group with WD group,and mild with severe WD group.Results:Compared with HC group,WD patients showed extensive reduction of FC values within the DMN,including the left medial prefrontal cortex(L_MPFC),left anterior cingulate gyrus(L_ACC)within the anterior default mode network(aDMN),and the left angular gyrus(L_ANG),precuneus(PCUN),left inferior parietal(L_IPG)and left posterior cingulate gyrus(L_PCC)within the posterior DMN(pDMN).In WD group,the FC values of L_MPFC and L_PCC within the aDMN and PCUN within the pDMN were found to be negatively correlated with UWDRS-N score;and the FC values of L_MPFC and L_ACC within aDMN,and L_IPG within pDMN were negatively correlated with UWDRS-P score.SVM me-thod was used to construct classifiers.For differentiating WD patients and HCs,the accuracy of the classifier was 80.23%,and the AUC was 0.865;and for differentiating mild and severe WD group,the accuracy was 70.89%,and the AUC was 0.723.Conclusion:There is extensively reduced functional connectivity in the default mode network of WD patients.The SVM classifier may improve the perfor-mance of DMN default mode network changes in identifying WD disorders and their condition transition.