Objective: This study aimed to develop a visual reading algorithm dedicated to the evaluation of ¹⁸F-Florzolotau positron emission tomography (PET) scans in PSP.

Background: The precise identification of tau accumulation within living brains holds significant potential in facilitating accurate diagnosis of progressive supranuclear palsy (PSP) and informing the enrollment of patients for clinical trials. While visual assessment is frequently employed, standardized visual interpretation methods for tau PET images specifically in PSP are absent.

Method: Patients with PSP (n = 148) and healthy volunteers (n = 30) with ¹⁸F-Florzolotau PET imaging were retrospectively included, and were equally and randomly divided into a development set and a testing set (n = 89 in each). For differential diagnosis, 55 patients with α-synucleinopathies were additionally included into the testing set. To establish the visual read method, an experienced assessor (Reader 0) was involved. The method was then validated using the testing set by Reader 0 and two additional readers. The regional binding was assessed as either positive (present) or negative (absent) in a binary manner. A positive binding in both the midbrain and globus pallidus/putamen regions was characterized as a PSP-like pattern, whereas any other pattern was classified as non-PSP-like.

Results: Utilizing Reader 0 as the benchmark, we observed excellent agreement for the overall binary determination (distinguishing between PSP-like and non-PSP-like patterns) for Reader 1 (94.4%) and Reader 2 (93.8%). Excellent agreement for the regional binary determinations of key PSP-related regions (midbrain and globus pallidus) were found as kappa values for inter-rater reliability consistently exceeded 0.80. The overall binary evaluation demonstrated reproducibility of 86.1%, 94.4% and 77.8% for Reader 0, Reader 1 and Reader 2, respectively. Additionally, the proposed visual reading algorithm showed high agreement with previously reported regional standardized uptake value ratios and clinical diagnoses.

Conclusion: Through the application of the suggested visual reading algorithm, ¹⁸F-Florzorotau PET imaging demonstrated a robust and consistent performance for the imaging diagnosis of PSP.

« Back to 2024 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/a-visual-interpretation-algorithm-for-assessing-18f-florzolotau-pet-scans-in-progressive-supranuclear-palsy/