【摘要】 目的：脑小血管病(cerebral small vessel disease,CSVD)患者眼球运动异常的机制尚未明确。本研究旨在探讨CSVD患者眼球运动与白质高信号(white matter hyperintensity,WMH)严重程度及位置分布的潜在关系，探索眼球运动评估作为特异性诊断工具的可能性。方法：采用横断面研究设计，纳入湘雅医院2022年9月7日至2023年10月27日共计161例门诊与住院CSVD患者为研究对象。采集患者人口学特征、既往史、用药史以及影像学数据；采用蒙特利尔认知评估量表(Montreal Cognitive Assessment,MoCA)评估患者的认知功能，分别采用汉密尔顿焦虑量表(Hamilton Anxiety Scale,HAMA)、汉密尔顿抑郁量表(Hamilton Depression Scale,HAMD)评估患者的焦虑、抑郁状态。所有患者在入组1周内完成EyeKnow智能眼动分析评价系统测试，记录眼球运动(间隔朝向扫视、平滑追踪、注视、反扫视)数据。基于年龄相关白质改变(age-related white matter changes,ARWMC)的量化评估与Fazekas分级2种评估方法对WMH进行评分，并根据严重程度进行分组。采用Kruskal-Wallis秩和检验进行组间差异分析，采用Spearman相关性分析、多元线性回归分析探讨眼球运动特征与WMH的相关性。针对全脑与额叶、颞叶、顶枕叶、幕下区、基底节区5个脑区是否存在WMH进行受试者操作特征(receiver operating characteristic,ROC)曲线分析，并采用基于加权随机森林的分析方法构建模型，评估眼球运动特征对不同位置WMH严重程度的分类准确率。结果：本研究共纳入161例CSVD患者，均收集了基线资料。根据ARWMC总分将患者分为轻度WMH组(0～10分)、中度WMH组(11～20分)和重度WMH组(21～30分)。其中，轻度WMH组100例，男66例(66.00%)，女34例(34.00%)，年龄(61.40±9.45)岁；中度WMH组46例，男32例(69.57%)，女14例(30.43%)，年龄(63.72±8.77)岁；重度WMH组15例，男12例(80.00%)，女3例(20.00%)，年龄(63.47±10.40)岁。3组患者MoCA得分(P=0.008)、脑微出血(cerebral microbleeds,CMB)严重程度(P<0.001)、基底节区血管周围间隙(basal ganglia perivascular spaces,BG-PVS)严重程度(P<0.001)、全脑皮质萎缩(global cortical atrophy,GCA)分级系统评分(P=0.003)的差异均有统计学意义。眼动特征的组间差异分析结果显示，WMH严重程度越高，间隔朝向扫视最快反应时长越长(P=0.008)，平滑追踪偏移量(>4°)越小(P=0.013)，注视偏移(>2°)次数越少(P=0.025)，差异均有统计学意义。事后两两比较中，中度WMH组与重度WMH组患者眼动特征比较，差异均无统计学意义(均P>0.05)。Spearman相关性分析结果显示，Fazekas总分与ARWMC总分在评估WMH严重程度方面存在较好的一致性(r=0.867,P<0.01);MoCA与Fazekas总分(r=-0.302,P<0.01)、ARWMC总分(r=-0.245,P<0.01)均呈显著负相关。基于ARWMC总分的多元线性回归分析显示，在排除多重共线性后，眼动特征中的平滑追踪启动时长(β=-0.001,P=0.009)、平滑追踪偏移量(β=-1.212,P=0.001)、注视偏移(>2°)次数(β=-0.102,P=0.011)、反扫视平均反应时长(β=0.016,P=0.018)与结局变量的关联仍有统计学意义。在调整性别、年龄、受教育年限、MoCA、HAMA、HAMD以及是否存在其他影像学标志物后，平滑追踪启动时长(β<0.001,P=0.010)、平滑追踪偏移量(β=-1.066,P=0.002)、反扫视平均反应时长(β=0.013,P=0.034)与结局变量的关联仍具有统计学意义。在WMH位置分布中，基于所有眼动特征对全脑、额叶、颞叶、顶枕叶、幕下区是否存在WMH进行ROC曲线分析，其曲线下面积值分别为0.933、0.928、0.758、0.784、0.881；基底节区的二元Logistic回归分析结果无统计学意义，无法进行ROC曲线分析。基于加权随机森林算法进一步对不同位置WMH严重程度进行分类，在调整了人口学特征、MoCA、HAMA、HAMD以及其他影像学标志物后，模型对额叶、幕下区、顶枕叶的分类准确率分别提升至85.71%、81.63%、75.51%。结论：CSVD患者眼球运动表现随着WMH严重程度的加剧而愈发恶化，在额叶和幕下区尤为显著。认知水平可能在很大程度上独立于影像学改变而影响眼球运动表现。更多还原

【Abstract】 Objective: The mechanism of abnormal eye movements in patients with cerebral small vessel disease(CSVD) remains unclear. This study aims to explore the potential link between eye movement in CSVD patients and the severity and distribution of white matter hyperintensities(WMH), and to evaluate the possibility of using eye movement assessment as a tool for specific diagnosis.Methods: This retrospective cross-sectional study was conducted at Xiangya Hospital, Central South University between September 7 th, 2022 and October 27 th, 2023, enrolling a total of 161 patients with CSVD. Demographic characteristics, past medical history, medication history, and imaging data were collected. The Montreal Cognitive Assessment(MoCA) was used to evaluate patients’ cognitive function, and the Hamilton Anxiety Scale(HAMA) and Hamilton Depression Scale(HAMD) were used to assess patients’ anxiety and depressive symptoms. All participants completed the EyeKnow Intelligent Eye Movement Analysis System within one week of enrollment, with data recorded on the following eye movement paradigms: saccade, smooth pursuit, fixation, and antisaccade. WMH were scored using both the Age-Related White Matter Change(ARWMC) scale and the Fazekas grading system. Based on the scores, patients were categorized into three severity groups(mild, moderate, severe). The Kruskal-Wallis rank sum test was used to analyze intergroup differences, while Spearman correlation analysis and multiple linear regression were used to explore the relationship between eye movement characteristics and WMH. Receiver Operating Characteristic(ROC) curve analysis was performed to evaluate the ability of eye movement characteristics to discriminate between patients with and without WMH in the five brain regions: frontal lobe, temporal lobe, parietal-occipital lobe, infratentorial region, and basal ganglia region. A weighted random forest model was developed to assess the performance of eye movement characteristics in predicting WMH severity at different locations.Results: This study enrolled a total of 161 patients with CSVD. Baseline data were collected for all participants. According to the total ARWMC scores, patients were divided into a mild(0 to 10 points), a moderate(11 to 20 points), and a severe(21 to 30 points) WMH groups. The mild WMH group included 100 patients [66 males and 34 females, age(61.40±9.45) years]. The moderate WMH group included 46 patients [32 males and 14 females, age(63.72±8.77) years]. And the severe WMH group included 15 patients [12 males and 3 females, age(63.47±10.40) years]. Significant differences were observed among the 3 groups in MoCA scores(P=0.008), severity of cerebral microbleeds(CMB)(P<0.001), severity of basal ganglia perivascular spaces(BG-PVS)(P<0.001), and global cortical atrophy(GCA) grading system scores(P=0.003). Analysis of intergroup differences in eye movement characteristics revealed that with increasing WMH severity, the fastest saccade reaction time increased(P=0.008), the smooth pursuit deviation(P=0.013) and the number of fixation shifts(>2°)(P=0.025) decreased. In post-hoc pairwise comparisons, there were no significant differences in any eye movement characteristics between the moderate and severe WMH group(all P>0.05). Spearman correlation analysis demonstrated a strong positive correlation between the total Fazekas and total ARWMC scores(r=0.867, P<0.01), confirming their concordance for rating WMH severity. Additionally, MoCA scores were significantly negatively correlated with both the total Fazekas scores(r=-0.302, P<0.01) and the total ARWMC scores(r=-0.245, P<0.01). Multiple linear regression analysis based on the total ARWMC scores revealed that after adjusting for multicollinearity, oculomotor features including smooth pursuit initiation time(β=-0.001, P=0.009), smooth pursuit deviation(β =-1.212, P=0.001), number of fixation shifts(>2°)(β =-0.102, P=0.011), and mean reaction time of antisaccade(β =0.016, P=0.018) remained statistically significant predictors of cognitive function. After adjusting for gender, age, years of education, MoCA, HAMA, HAMD scores, and the presence of other imaging markers, the associations of smooth pursuit initiation time(β <0.001, P=0.010), smooth pursuit deviation(β=-1.066, P=0.002), and mean reaction time of antisaccade(β=0.013, P=0.034) with the outcome variable remained statistically significant. In the distribution of WMH locations, ROC curve analysis was conducted based on all eye movement characteristics to discriminate the presence of WMH in the whole brain, frontal lobe, temporal lobe, parietal-occipital lobe, and infratentorial region, with AUC values of 0.933, 0.928, 0.758, 0.784, and 0.881, respectively. For the basal ganglia region, binary logistic regression analysis showed no significant association, and therefore ROC curve analysis was not applicable. Using a weighted random forest method, the severity of WMH at different locations was further classified. After adjusting for gender, age, years of education, MoCA, HAMA, HAMD scores, and the presence of other imaging markers, the model’s classification accuracy improved to 85.71% for the frontal lobe, 81.63% for the infratentorial region, and 75.51% for the parietal-occipital lobe.Conclusion: The eye movement performance of CSVD patients worsens with the increasing severity of WMH, especially in the frontal lobe and infratentorial region. Cognitive function exerts an influence on eye movement that appears largely independent of imaging changes.更多还原