脑白质病变MRI视觉分级方法的比较

Comparison of visual rating scales for white matter lesions on magnetic resonance imaging

目的：比较7种常用的脑白质病变（WMLs）MRI视觉分级方法，评价各分级之间的相关性和一致性，并计算两两间相互转化的参数．方法：随机选取50例MRI T2-FLAIR显示有不同程度WMLs患者，由两名医师分别用7种分级方法对其WMLs进行评分．视觉分级间的相关性评价采用Spearman等级相关分析，不同评估者之间的一致性分析采用Kappa检验，通过回归分析计算各分级间相互转化的参数．结果：不同视觉分级在WMLs评分上均有较强的相关性（相关系数rs：0．579～0．917，P〈0．05）．除Sehelten（modilified）分级和Ylikoski分级外，其它分级在不同观察者间的一致性好（K〉0．585）．除Aharon-Peretz J分级和De Groot JC分级外，通过回归分析得出了其它5个分级间相互转化的参数，其确定系数砰的范围为0．5856～0．8892．结论：7种不同白质病变的MRI视觉分级间有较好的相关性，而不同评估者间Aharon-Peretz J分级的一致性最好．

AIM： To compare 7 widely used rating scales for white matter lesions （WMLs） on MRI, to assess the correlations among the scales and the interrater agreement for each scale, and to calculate the conversion coefficients of the score between different scales. METHODS： We selected randomly 50 patients with various degrees of WMLs on T2-FLAIR imaging. The WMLs were scored by two trained raters using 7 MRI rating scales for WMLs. Spearman rank correlation, Kappa test and regression analysis were applied. RESULTS： A high correlation was observed among the 7 scales （correlation coefficients ranging from 0.579 to 0. 917, P 〈 0.05 ）. Apart from the scale of Schelten （modilitied） and Ylikoski, other rating scales demonstrated better interrater agreement （ K 〉 0.585 ）. Using regression analysis, we determined numeric parameters to transform the score of 1 scale to the corresponding score for each of the remaining scales, except for the scale of Aharon-Peretz J and De Groot JC. The predictive values of these conversions R2 ranged from 0. 5856 to 0. 8892. CONCLUSION： Our data demonstrates a high correlation among the considered visual rating scales for WMLs and a best interrater agreement for the scale of Aharon-Peretz J.