基于动脉自旋标记成像分析帕金森病神经代谢网络拓扑属性的改变

The topological alterations of metabolic neural network in Parkinson disease:an arterial spin labeling imaging study

目的:应用动脉自旋标记(ASL)灌注成像(PWI)从代谢视角探讨神经网络的无创性构建,并分析其拓扑属性改变在帕金森病(PD)神经损伤中的作用机制。方法:基于磁共振ASL-PWI获取52例PD患者及相匹配的55例健康志愿者(HC组)的脑血流信息,构建组水平的神经代谢网络,采用图论分析网络的全局和节点拓扑属性(稀疏度为10%~50%,步长为1%)并进行组间比较。全局拓扑属性包括小世界属性(γ、λ、σ)、最短路径长度(Lp)、聚类系数(Cp)、全局效率(Eglob)和局部效率(Eloc),节点拓扑属性包括节点介数(BC)、节点度(DC)和节点效率(NE)。结果:两组受试者年龄、性别分布、受教育程度及认知功能(MMSE)评分的差异均无统计学意义(P>0.05)。PD组运动功能(UPDRS-Ⅲ)评分为24.60±14.00、疾病H-Y分期为1.39±0.49。在全局拓扑属性方面,PD组和HC组的神经代谢网络均显示出小世界属性(γ>1、λ≈1且σ>1);PD组的Eglob值显著高于HC组(0.25 vs.0.23),而λ(0.42 vs.0.46)、Lp(0.67 vs.0.76)、Cp(0.26 vs.0.30)和Eloc值(0.32 vs.0.34)均显著低于HC组(P均<0.05)。在节点拓扑属性方面(三种节点属性中任一异常),相较于HC组,PD组在双侧楔前叶的BC(左:436.56 vs.60.71;右:529.37 vs.92.12)、双侧中央后回的NE(左:0.26 vs.0.31;右:0.24 vs.0.30)和DC(右:10.92 vs.18.95)、左侧岛叶BC(14.23 vs.81.00)、左侧舌回BC(59.55 vs.3.84)、左侧梭状回BC(75.32 vs.5.83)、左侧颞中回BC(37.19 vs.221.12)、DC(15.04 vs.22.59)和NE(0.26 vs.0.32)、左侧额下回BC(127.39 vs.4.02)和双侧额上回BC(左:4.02 vs.127.39;右:3.47 vs.71.77)均存在显著改变(P均<0.05,FDR校正)。结论:基于ASL-PWI显示的脑血流信息无创性构建神经代谢网络具备可行性,且PD患者的信息处理能力在全局及节点水平均存在代谢性病理异常。因此,本研究结果有助于加深对PD神经网络退行性损伤的机制阐述。

Objective:This study was aimed to explore the noninvasive construction of neural network using magnetic resonance arterial spin labeled perfusion imaging(ASL-PWI)from the perspective of metabolism,and to investigate the mechanism of its topological property changes in the neurodegeneration of Parkinson’s disease(PD).Methods:The data of cerebral blood flow(CBF)in both PD group and matched healthy control(HC)group were obtained by ASL-PWI.The construction of metabolic neural networks at group level was performed by covariation approach,followed by calculation of network properties(global and nodal)based on graph theory and statistical analysis between groups.Global properties include small-worldness(γ,λandσ),characteristic path length(Lp),clustering coefficient(Cp),global efficiency(Eglob)and local efficiency(Eloc).Nodal properties include nodal betweenness(BC),nodal degree centrality(DC)and nodal efficiency(NE).Results:A total of 52 PD patients and 55 HC subjects were enrolled.There were no significant intergroup differences in age,gender distribution,years of education and cognitive function(MMSE)(P>0.05);For PD patients,the motor function was scored as 24.60±14.00 by UPDRS-Ⅲ,and H-Y stage was 1.39±0.49.In terms of global properties,metabolic neural networks in both PD and HC groups were featured with small-world property(γ>1,λ≈1 andσ>1).The value of Eglob in PD group was significantly higher than that of HC group(0.25 vs.0.23,P<0.05),while the values ofλ(0.42 vs.0.46),Lp(0.67 vs.0.76),Cp(0.26 vs.0.30)and Eloc(0.32 vs.0.34)were all significantly lower than those of HC group(all P<0.05).In terms of nodal property,when compared with HC group(any of the three nodal properties),PD group showed significant changes in bilateral precuneus(left BC:436.56 vs.60.71;right BC:529.37 vs.92.12),bilateral postcentral gyrus(left NE:0.26 vs.0.31;right NE;0.24 vs.0.30;right DC:10.92 vs.18.95),left insula(BC:14.23 vs.81.00),left lingual gyrus(BC:59.55 vs.3.84),left fusiform gyrus(BC:75.32 vs.5.83),left middle temporal gyrus(BC:37.19 vs.221.12;DC:15.04 vs.22.59;NE:0.26 vs.0.32),left inferior frontal gyrus(BC:127.39 vs.4.02)and bilateral superior frontal gyrus(left BC:4.02 vs.127.39;right BC:3.47 vs.71.77)with statistical difference(P<0.05,FDR corrected).Conclusion:It is feasible to construct metabolic neural network noninvasively based on information of cerebral blood flow that derived from MR ASL-PWI.The metabolic ability of information processing in PD patients is pathologically impaired at both the global and nodal levels.The results of this study may contribute to further elucidation of the neurodegenerative mechanism in PD neural network.