Purpose: Quantification methods for white matter hyperintensities (WMH) on Magnetic Resonance Imaging are heterogeneous, deterring their application. This study compared three WMH rating scales, varying in complexity, and a volumetric method, to evaluate trade-offs between complexity and clinical utility in differentiating dementia subgroups and in correlating with cognition.
Methods: WMH were rated using the Fazekas, Age-Related White Matter Changes (ARWMC) and Scheltens scales, and segmented by computational volumetry in 108 patients with Alzheimer's Disease (AD), 23 with Mild Cognitive Impairment (MCI) and 34 normal controls (NC). Global and hippocampal atrophy, age and education, were accounted for in correlations of WMH with cognitive domains.
Results: Intra- and inter-rater reliability were high (intraclass correlation coefficients = 0.88-0.97) across rating scales. WMH scores of all scales were highly correlated with volumes (Spearman r = 0.78-0.90, Ps < 0.001), as well as with each other (Spearman r = 0.86-0.91, Ps < 0.001). The Fazekas scale showed significant separation between AD, MCI and NC using non-parametric analysis, while the ARWMC and Scheltens' scales, and WMH volumes demonstrated significant correlations (standardized β = -0.19 to -0.24, Ps < 0.05) with cognitive domain scores using multivariate regression analysis, controlling for age, education, global and hippocampal atrophy in patients with AD.
Conclusions: This study suggests that the degree of complexity of WMH rating scales did not affect validation against WMH volumes, but did vary in validation against cognition. The simplest scale performed best in separating cognitive subgroups, but the more complex scales and quantification correlated better with cognitive measures, especially executive function. Hence the best choice of scale depends on the particular application.