METHODS: Three hundred sixty cognitively normal (CN) subjects from the Alzheimer\'s Disease Neuroimaging Initiative (ADNI) database and 105 CN subjects from the Parkinson\'s Progression Markers Initiative database were used to develop the model. In ADNI, 230 mild cognitive impairment (MCI) subjects were matched with 230 CN old-aged subjects to evaluate their heterogeneity difference. One hundred four MCI subjects with 48 months of follow-up were divided into low and high heterogeneity groups. Kaplan-Meier survival curve analysis was used to observe the importance of heterogeneity results for predicting MCI progression.
RESULTS: The area under the receiver operating characteristic curve of the model in the training, internal test and external test sets was 0.7503, 0.7512 and 0.7514, respectively. There was a significantly positive correlation between age and radscore of CN subjects (r = 0.501; P < .001). The radscore of MCI subjects was significantly higher than that of matched CN subjects (P < .001). The median radscore ratios of MCI to CN from four age groups (66-70y, 71-75y, 76-80y and 81-85y) were 1.611, 1.760, 1.340 and 1.266, respectively. The probability to progression of low and high heterogeneity groups had a significant difference (P = .002).
CONCLUSIONS: When radscore is significantly higher than that of normal ageing, it is necessary to alert the possibility of cognitive impairment and deterioration.
方法:使用来自阿尔茨海默病神经影像学计划(ADNI)数据库的160名认知正常(CN)受试者和来自帕金森进展标志物计划数据库的105名CN受试者来开发模型。在ADNI,将230名轻度认知障碍(MCI)受试者与230名CN老年受试者进行匹配,以评估其异质性差异。将104名MCI受试者进行48个月的随访,分为低异质性和高异质性组。采用Kaplan-Meier生存曲线分析观察异质性结果对预测MCI进展的重要性。
结果:训练中模型的接收器工作特性曲线下的面积,内部测试和外部测试集分别为0.7503、0.7512和0.7514。CN受试者的年龄与radscore之间存在显着正相关(r=0.501;P<.001)。MCI受试者的radcore显著高于匹配的CN受试者(P<.001)。四个年龄组的MCI与CN的中位数radscore比率(66-70岁,71-75y,76-80y和81-85y)分别为1.611、1.760、1.340和1.266。低异质性和高异质性组的进展概率有显著差异(P=0.002)。
结论:当radscore明显高于正常老化时,有必要警惕认知障碍和恶化的可能性。