Visual Classification of Tau-PET Detects 4 Subtypes With Different Long-Term Outcomes

Area of Science:

• Neuroimaging
• Alzheimer's Disease Research
• Biomarker Discovery

Background:

• Alzheimer's disease (AD) exhibits variable tau accumulation patterns.
• A data-driven approach identified four distinct tau spatiotemporal trajectories (SuStaIn).
• Clinical translation requires a visual method to identify these tau subtypes.

Purpose of the Study:

• To establish a standardized visual topographic method for identifying tau patterns using tau-PET.
• To validate visual tau subtype classification in a clinical setting.
• To assess the clinical utility of visual tau pattern identification.

Main Methods:

• Prospective study of 245 individuals with tau-PET scans from a memory clinic.
• Classification of tau-PET scans into four subtypes (S1-S4) using visual rating and the SuStain algorithm.
• Statistical analysis of inter-rater agreement (Cohen's κ) and differences in clinical/biomarker features between subtypes.

Main Results:

• Substantial agreement between raters for visual tau subtype interpretation (κ > 0.65).
• Fair agreement between visual and automated tau subtype classification (κ = 0.39).
• Visual classification identified subtypes with differing global tau load, clinical features, and cognitive decline rates.

Conclusions:

• Visual tau-PET classification reliably identifies four distinct tau patterns in Alzheimer's disease.
• These patterns correlate with clinical features and long-term outcomes, indicating clinical usefulness.
• This method supports personalized diagnosis and prognosis for higher-risk AD variants.