摘要目的:应用隐马尔科夫模型(Hidden Markov model，HMM)探讨帕金森病(Parkinson disease，PD)患者大脑动态功能脑网络连接改变，以及动态功能指标与临床指标的相关性。方法:2019年至2023年，纳入PD患者48例(PD组)和健康对照组33名(HC组)，使用蒙特利尔认知评估量表(Montreal cognitive assessment，MoCA)对PD患者总体认知功能进行评估，使用统一帕金森病评定量表Ⅲ(unified Parkinson's disease rating scale Ⅲ，UPDRS-Ⅲ)评估患者的运动状态。使用HMM技术进行动态功能脑网络连接分析，并获取动态高阶指标部分占用率(fractional occupancy，FO)、状态切换率(switching rate，SR)及平均居留时间(mean dwell time，MDT)。采用两独立样本 t检验计算不同状态内功能连接矩阵组间差异，采用Mann-Whitney U检验计算不同状态动态高阶指标组间差异。采用Spearman相关分析计算PD组动态高阶指标与临床指标的相关性。 结果:所有被试应用HMM构建了6种空间状态，在状态1稀疏连接中，PD组MDT[24.93(19.73)个窗口]高于HC组[17.63(14.80)个窗口]( Z＝－2.030， P＝0.042)；在状态5紧密连接中，PD组MDT[6.00(3.00)个窗口]低于HC组[9.75(7.70)个窗口]( Z＝－2.210， P＝0.027)。状态3的FO与PD组MoCA评分呈负相关( r＝－0.331， P＝0.022)，PD患者状态5的FO与UPDRS-Ⅲ评分呈正相关( r＝0.412， P＝0.004)，在状态5的MDT与UPDRS-Ⅲ评分呈正相关( r＝0.448， P＝0.001)。 结论:HMM可捕捉大脑动态脑网络的瞬态变化，为帕金森病患者动态脑网络研究提供一定价值。

abstractsObjective:To investigate the changes of dynamic functional brain network connectivity in patients with Parkinson disease(PD) using Hidden Markov model(HMM), and to analyze the correlation between dynamic functional parameters and clinical parameters.Methods:Forty-eight PD patients(PD group) and thirty-three healthy controls(HC group) were included from 2019 to 2023. The cognitive function was assessed using the Montreal cognitive assessment (MoCA), and motor status was assessed using the unified Parkinson's disease rating scale Ⅲ(UPDRS-Ⅲ) in PD group.HMM technique was used to analyze the dynamic functional brain network connectivity, and the dynamic higher-order index fractional occupancy(FO), switching rate(SR), and mean dwell time(MDT) were obtained. Two independent samples t-test was used to calculate the differences between groups of functional connectivity matrices in different states, and Mann-Whitney U test was used to calculate the differences between groups of dynamic higher-order indicators in different states. Spearman correlation analysis was used to calculate the correlation between dynamic higher-order parameters and clinical parameters in the PD group. Results:The HMM was used to construct 6 spatial states for all subjects.MDT was significantly higher in PD group(24.93(19.73)) in state 1 sparse junctions than that in HC group(17.63(14.80)) ( Z=-2.030, P=0.042), but significantly lower MDT was showed in PD group(6.00(3.00)) in state 5 tight junctions than that in HC group(9.75(7.70)) ( Z=-2.210, P=0.027).FO in state 3 was negatively correlated with MoCA score in PD group( r=-0.331, P=0.022).FO in state 5 was positively correlated with UPDRS-Ⅲ score in PD patients( r=0.412, P=0.004), and MDT in state 5 was positively correlated with UPDRS-Ⅲ score( r=0.448, P=0.001). Conclusion:HMM can capture the transient changes of dynamic brain network, which can provide some value for the study of dynamic brain network in patients with Parkinson disease.