Evaluation of a Sensitive Visual Read Algorithm for Assessing 3R/4R Tau PET Images

Area of Science:

• Neuroimaging
• Alzheimer's Disease Diagnostics
• Radiopharmaceutical Imaging

Background:

• Accurate assessment of Alzheimer's disease (AD)-related tau in tau Positron Emission Tomography (PET) imaging is crucial for clinical trials and practice.
• Visual read algorithms are being developed to interpret tau PET scans, necessitating validation against quantitative measures and biomarkers.

Purpose of the Study:

• To compare the accuracy of two visual read algorithms (BioFINDER visual read [BF-VR] and [18F]flortaucipir visual read [FTP-VR]) for tau PET imaging.
• To evaluate these algorithms against semiquantitative measurements and plasma phospho-tau 217 (p-tau217) status.
• To assess the reliability of the BF-VR algorithm.

Main Methods:

• 1,654 participants from the BioFINDER-2 cohort underwent [18F]RO948 PET scans; 37 also had [18F]flortaucipir scans.
• PET scans were analyzed using the BF-VR protocol and the established FTP-VR method.
• Comparative analyses included semiquantitative SUV ratios (SUVRs) in the entorhinal cortex (ERC) and temporal meta-region, and plasma p-tau217 status.

Main Results:

• Both visual read methods showed strong concordance with semiquantitative SUVRs.
• BF-VR demonstrated superior accuracy for tau in the ERC (93.9%) compared to FTP-VR (89.9%).
• BF-VR exhibited excellent interrater (κ=0.87) and intrarater (κ=0.94) reliability, with high concordance between [18F]RO948 and [18F]flortaucipir assessments (κ=0.94).

Conclusions:

• Visual reads provide a simple method for evaluating tau PET scans.
• The BF-VR algorithm is a more accurate method for detecting tau uptake in the ERC than FTP-VR.
• BF-VR shows excellent reliability and comparable performance with different tau PET tracers ([18F]RO948 and [18F]flortaucipir).