Keenihan EK, Summers BA, David FH, et al.
In order to compare the accuracy of MR sequences for diagnosis of meningeal disease, MR images of the brain, and histopathologic specimens including the meninges of 60 dogs were reviewed retrospectively by independent observers in a cross-sectional study. MR images included T1-weighted pre- and postgadolinium images, subtraction images, T2-weighted images, and T2-weighted fluid-attenuated inversion-recovery (FLAIR) images. Pathologic changes affected the pachymeninges in 16 dogs, leptomeninges in 35 dogs, and brain in 38 dogs. The meninges were normal in 12 dogs. Meninges were classified histopathologically as normal (grade 0), slightly or inconsistently affected (grade 1), or markedly affected (grade 2). When applying relaxed pathologic criteria (grades 0 and 1 considered normal), the results of ROC analysis (area under curve, AUC) were: T1-weighted postcontrast images 0.74; subtraction images 0.7; T2-weighted images 0.68; FLAIR images 0.56. The difference in AUC between T1-weighted postgadolinium images and FLAIR images was significant (P = 0.04). AUC for FLAIR images was not significantly different from 0.5. When applying strict pathologic criteria (only grade 0 considered normal), none of the MR sequences had AUC significantly different from 0.5. On the basis of T1-weighted postgadolinium images and subtraction images, correct anatomic classification of lesions occurred more often for pachymeningeal than leptomeningeal lesions (P < 0.001). Overall, MR imaging had low sensitivity for diagnosis of meningeal pathology in dogs, particularly for changes affecting the leptomeninges. Subtraction images had similar accuracy to T1-weighted postgadolinium images for meningeal lesions in dogs. T2-weighted FLAIR images appear to have limited diagnostic utility for meningeal lesions.