Conventional magnetic resonance imaging in confirmed progressive supranuclear palsy and multiple system atrophy

Conventional magnetic resonance imaging (cMRI) is often used to aid the diagnosis of progressive supranuclear palsy (PSP) and multiple system atrophy (MSA), but its ability to predict the histopathological diagnosis has not been systematically studied. cMRI from 48 neuropathologically confirmed cases, including PSP (n = 22), MSA (n = 13), Parkinson's disease (PD) (n = 7), and corticobasal degeneration (n = 6), and controls (n = 9) were assessed blinded to clinical details and systematically rated for reported abnormalities. Clinical diagnosis and macroscopic postmortem findings were retrospectively assessed. Radiological assessment of MRI was correct in 16 of 22 (72.7%) PSP cases and 10 of 13 (76.9%) MSA cases with substantial interrater agreement (Cohen's kappa 0.708; P < .001); no PSP case was misclassified as MSA or vice versa. MRI was less sensitive but more specific than clinical diagnosis in PSP and both more sensitive and specific than clinical diagnosis in MSA. The "hummingbird" and "morning glory" signs were highly specific for PSP, and "the middle cerebellar peduncle sign" and "hot cross bun" for MSA, but sensitivity was lower (up to 68.4%) and characteristic findings may not be present even at autopsy. cMRI, clinical diagnosis, and macroscopic examination at postmortem have similar sensitivity and specificity in predicting a neuropathological diagnosis. We have validated specific radiological signs in pathologically confirmed PSP and MSA. However, the low sensitivity of these and macroscopic findings at autopsy suggest a need for imaging techniques sensitive to microstructural abnormalities without regional atrophy.