Establishment of diagnostic system for progressive supranuclear palsy using in vivo tau imaging

[Objective] We aim to establish a diagnostic system for progressive supranuclear palsy (PSP) by tau positron emission tomography (PET) with 18F-PM-PBB3. [Methods] We employed the data of 24 healthy controls (HC; age 67.5 ± 5.1 [mean ± SD] y, 9 males); 30 PSP (71.4 ± 8.2 y, 21 males, PSP rating scale [PSPRS] 41.1 ± 17.7) who met the MDS-PSP criteria and showed a typical topological pattern of 18F-PM-PBB3 PET observed in PSP. All data were corrected by age and sex, and standardized for analysis. Standardized uptake value ratios using the cerebellar cortex as reference region were obtained in 112 volumes of interests (VOIs) by the multi-atlas method. The Elastic Net cross validation analysis was applied to the set of VOIs to determine which VOIs useful for discriminating PSP from HC. The obtained coefficients were applied to each individual data to calculate and define as PSP score. We tested PSP score for validation to 10 HC (68.9 ± 7.4 y, 7 males) and 5 PSP cases (73.0 ± 7.8 y, 3 males, PSPRS 46.6 ± 11.3) in a new cohort. We also evaluated association between PSP score and disease severity measured by PSPRS. [Results] Globus pallidus, putamen and midbrain were selected as pivotal VOIs calculating PSP score. PSP score for training and validation data showed accuracy of 94.4% and 80%, precision of 93.3% and 100%, and recall of 96.6% and 62.5%. Moreover, PSPRS correlated well with PSP score (Spearman’s rs = 0.56, p < 0.001). [Conclusions] Automated analysis system of tau PET with 18F-PM-PBB3 would be a promising tool both for diagnosing and predicting disease severity in PSP.第62回日本神経学会学術大