静息态脑电在阿尔茨海默病诊断中的价值

Value of resting state electroencephalogram in the diagnosis of Alzheimer's disease

背景阿尔茨海默病(AD)诊断仍面临很大挑战,脑电图检查具有便携、无创的优势,脑电诊断AD是目前的研究热点。目的探讨静息态脑电用于AD诊断的价值,为临床上AD的早期识别提供参考。方法回顾性分析2019年5月-2022年5月在深圳市康宁医院老年精神障碍科住院的AD患者(n=59)临床资料,以同期在该院门诊检查的健康老年人作为对照组(n=54)。收集8通道静息态脑电数据,使用快速傅里叶变换(FFT)计算患者在α、β、θ、δ频段脑电的绝对功率值和α/θ绝对功率比值。采用简易精神状态评价量表(MMSE)和蒙特利尔认知评估量表(MoCA)评定患者的认知功能。采用Spearman相关分析考查患者脑电变量与MMSE和MoCA评分的相关性。基于选定的脑电及临床资料,建立预测AD的Logistic回归模型,采用受试者工作特征(ROC)曲线下面积(AUC)评估模型性能。结果AD患者右额极(F4)、左右侧额极(F7、F8)θ绝对功率均高于健康对照组,差异均有统计学意义(t=-2.844、-2.825、-3.014,P<0.05或0.01);AD患者左右前额极(Fp1、Fp2)、左右额极(F3、F4)、左右侧额极(F7、F8)α/θ绝对功率比值均低于健康对照组,差异均有统计学意义(t=2.081、2.327、3.423、2.358、3.272、2.445,P<0.05或0.01)。Spearman相关分析显示,MMSE评分与α绝对功率、β绝对功率和α/θ绝对功率比值均呈正相关(r=0.206、0.288、0.372,P<0.05或0.01)。MoCA评分与β绝对功率和α/θ绝对功率比值均呈正相关(r=0.201、0.315,P<0.05或0.01),与θ绝对功率呈负相关(r=-0.218,P<0.05)。脑电组合预测AD的模型ROC曲线AUC=0.882(95%CI:0.820~0.943),灵敏度为0.966,特异度为0.673。综合变量模型预测能力最强,ROC曲线AUC=0.946(95%CI:0.905~0.986),灵敏度为0.948,特异度为0.873。结论AD患者静息态脑电与认知功能相关。静息态脑电在AD诊断中可能具有重要价值,其中θ绝对功率和α/θ绝对功率比值可能与AD的相关性最强。

Background The diagnosis of Alzheimer's disease(AD)still faces great challenges,and the advantage of electroencephalogram(EEG)diagnosis lies in its portable and non-invasive nature,so the EEG diagnosis of AD has occupied an important place in clinical research.Objective To evaluate the value of resting state EEG for AD diagnosis,and to provide references for early recognition of AD in clinical practice.Methods Clinical data of AD patients(n=59)in an Inpatient Geriatric Psychiatry Unit of Shenzhen Kangning Hospital from May 2019 to May 2022 were retrospectively analyzed,and healthy elderly individuals attending outpatient clinics at the hospital during the same period were enrolled as control group(n=54).Eight-channel resting state EEG data were acquired,and the absolute power values in theα,β,θandδfrequency bands and theα/θratio were obtained and calculated using Fast Fourier Transform(FFT).Cognitive function assessments of patients were done by Mini-Mental State Examination(MMSE)and Montreal Cognitive Assessment(MoCA).Spearman correlation analysis was used to examine the correlation between EEG findings and MMSE and MoCA scores of AD patienrs.Logistic regression prediction model for AD was built using currently available EEG and clinical variables,and the model performance was assessed using the receiver operating characteristic(ROC)curve and the area under curve(AUC).Results Theθ-band absolute powers in the right mid-frontal(F4)and mid-lateral(F7,F8)regions were higher in AD patients than those in healthy controls,with statistically significant difference(t=-2.844,-2.825,-3.014,P<0.05 or 0.01).The absolute powers ofα/θratio in prefrontal(Fp1,Fp2),mid-frontal(F3,F4)and mid-lateral(F7,F8)regions showed a notable reduction in AD patients compared with healthy controls,with statistical difference(t=2.081,2.327,3.423,2.358,3.272,2.445,P<0.05 or 0.01).Spearman correlation analysis denoted that MMSE score was positively correlated with the absolute powers ofα-band,β-band andα/θratio(r=0.206,0.288,0.372,P<0.05 or 0.01).MoCA score was positively correlated withβabsolute powers andα/θratio(r=0.201,0.315,P<0.05 or 0.01),and negatively correlated withθabsolute power(r=-0.218,P<0.05).ROC curve revealed an AUC of 0.882(95%CI:0.820~0.943),a sensitivity of 0.966 and a specificity of 0.673 for the AD prediction model based on EEG variables,while the prediction model for AD using comprehensive variables achieved better predictive efficacy,reaching an AUC,sensitivity and specificity of 0.946(95%CI:0.905~0.986),0.948 and 0.873,respectively.Conclusion Resting state EEG of AD patients is correlated with cognitive function,and are of great value in the diagnosis of AD,withθabsolute power andα/θratio in EEG being the most strongly correlated with AD.