Planimetric and linear MRI markers for progressive supranuclear palsy classification: a large multicohort international study

BACKGROUND: Differentiating progressive supranuclear palsy (PSP) from Parkinson disease and other parkinsonisms is challenging. MRI measures have been investigated, but most studies included small samples, limiting result reliability. PURPOSE: To compare the performance of planimetric and linear MRI measurements in differentiating PSP from other parkinsonisms and identify an optimized linear marker. MATERIALS AND METHODS: Participants with PSP and non-PSP parkinsonisms and controls were included in this secondary analysis of multiple international prospective studies (enrollment: 2006–2024). The previously established midbrain line, midbrain area, pons to midbrain area ratio, and MR parkinsonism index were compared with a new simple linear marker based on two midbrain measures performed on midsagittal T1-weighted sections (dual-line midbrain PSP index [DMPI]). Logistic regression including DMPI, age, and sex was used to differentiate participants with PSP (PSP–Richardson syndrome, PSP variants) from those with non-PSP parkinsonisms and controls in two large independent cohorts and in a small cohort of participants with pathologically proven diagnoses. RESULTS: A total of 2111 participants (mean age, 67.8 years ± 8.4 [SD]; 54% male) were included (Italian: 136 PSP, 238 non-PSP, 85 controls; international: 520 PSP, 564 non-PSP, 525 controls). All markers were compared in a subcohort (PSP [n = 161]; non-PSP parkinsonisms [n = 203]) representative of the overall cohort. All measures showed area under the receiver operating characteristic curve (AUC) values over 0.90 for differentiating PSP from non-PSP parkinsonisms, with the DMPI and midbrain area performing best (AUCs, 0.97 [95% CI: 0.95, 0.98] and 0.95 [95% CI: 0.93, 0.97], respectively) and the DMPI showing the smallest percentage of uncertain cases (gray zone, 29 of 364 participants [7.97%]). In the entire Italian and international cohorts, the DMPI distinguished participants with PSP from those with non-PSP, with AUCs of 0.97 (95% CI: 0.97, 0.98) and 0.96 (95% CI: 0.95, 0.97), respectively. Excellent DMPI performance was also observed in participants with early-stage disease (AUC, 0.97 [95% CI: 0.95, 0.99]) and those with pathologically confirmed diagnoses (n = 43) (AUC, 0.94 [95% CI: 0.86, 1.00]). CONCLUSION: The DMPI and midbrain area performed well for differentiating PSP from other neurodegenerative parkinsonisms.