【摘要】 目的：探讨定量磁化率成像（QSM）容积分析在评估脑小血管病（CSVD）微出血与认知功能损害中的应用价值。方法：使用蒙特利尔认知评估量表（MoCA）对CSVD患者的认知功能进行测评。根据认知功能分为CSVD认知正常组（NI组）与CSVD伴认知障碍组（CI组），受试者接受颅脑常规MR及QSM扫描，结合半自动分割技术评估患者脑微出血（CMBs）的数量、部位和容积。采用独立样本非参数检验比较两组受试者CMBs个数、总容积及分布差异、深部白质Fazekas评分等参数。运用Spearman相关分析，探讨CMBs的数量及容积与MoCA总分之间的相关性；采用回归分析方法挖掘CSVD患者认知障碍的危险因素，最后通过受试者工作特征（ROC）曲线评估上述因素在识别认知功能障碍中的效能。结果：CI组（66例）的CMBs数量和容积均显著高于NI组（43例）（P<0.05）。CI组全脑、左侧额叶、右侧顶叶、两侧基底节区脑微出血数量和容积显著多于NI组（P<0.05）。CMBs总数量（r=-0.267, P=0.005）、CMBs总容积（r=-0.289, P=0.002）与MoCA评分负相关。深部白质Fazekas评分（OR=2.15, P=0.020）和CMBs总容积（OR=1.004, P=0.011）是导致CSVD患者认知障碍的独立危险因素。结论：QSM结合深度学习的半自动分割技术提供的CMBs总容积及其在额叶、顶叶和基底节区的特异性分布与CSVD患者的认知功能下降密切相关，该结果为理解CSVD相关的认知障碍发生机制提供了潜在的解释途径。更多还原

【Abstract】 Purpose: To explore the value of quantitative susceptibility mapping(QSM)-based volumetric analysis in evaluating cerebral microbleeds(CMBs) and cognitive impairment in patients with cerebral small vessel disease(CSVD). Methods: Cognitive function of patients with CSVD was assessed using the Montreal cognitive assessment(MoCA). Paticipants were categorized into a cognitively normal CSVD group(NI group) and a CSVD with cognitive impairment group(CI group). All subjects underwent routine brain MRI and QSM. A deep learningassisted semi-automatic segmentation approach was used to quantify the number, anatomical distribution, and total volume of CMBs. Group differences in CBM counts, total volume, anatomical distribution, deep white matter Fazekas score, and other parameters were compared using independent-sample nonparametric tests. Spearman correlation was used to examine associations between CBM burden(count and volume) and MoCA total score. Multivariable regression was performed to identify independent risk factors for cognitive impairment in CSVD patients, and receiver operating characteristic(ROC) curve analysis was used to evaluate their discriminatory performance. Results: Compared with the NI group(n=43), the CI group(n=66) showed significantly higher CBM counts and total CMB volume(P<0.05). Regionally, CMB count and volume were significantly greater in the whole brain, left frontal lobe, right parietal lobe, and bilateral basal ganglia in the CI group than in the NI group(P<0.05). Both total CMB count(r=-0.267, P=0.005) and total CMB volume(r=-0.289, P =0.002) were negatively correlated with MoCA scores. Deep white matter Fazekas score(OR=2.15, P=0.020) and CMB volume(OR=1.004, P=0.011) were identified as independent risk factors for cognitive impairment in CSVD patients. Conclusion: QSM-based volumetric quantification of CMBs using deep learning-assisted semi-automatic segmentation, particularly total CMB volume and their preferential distribution in the frontal, parietal, and basal ganglia regions, is closely associated with cognitive decline in CSVD patients. These findings may provide insight into the mechanisms underlying CSVDrelated cognitive impairment.更多还原