Advancing Tau PET Quantification in Alzheimer Disease with Machine Learning: Introducing THETA, a Novel Tau Summary...

Area of Science:

• Neuroimaging
• Biomarkers
• Machine Learning

Background:

• Alzheimer disease (AD) is characterized by heterogeneous tau deposition, posing challenges for accurate quantification.
• Existing methods for assessing tau burden may not fully capture its spatial variability.

Purpose of the Study:

• To develop an automated, clinically useful method for quantifying heterogeneous tau deposition in Alzheimer disease.
• To create a single summary measure for tau burden from tau PET scans.

Main Methods:

• Utilized tau PET scans from three independent cohorts (Mayo, ADNI, OASIS-3).
• Trained a machine learning classification model to predict visual tau positivity.
• Developed the Tau Heterogeneity Evaluation in Alzheimer's Disease (THETA) score using SUV ratios and Shapley additive explanations.

Main Results:

• The machine learning model achieved high accuracy in classifying tau positivity across cohorts (AUCs 0.94-1.00).
• The THETA score demonstrated stronger correlations with cognitive assessments (MMSE, CDR-SB) than meta-ROIs.
• THETA effectively discriminated between cognitively unimpaired and mild cognitive impairment groups.

Conclusions:

• The proposed automated approach accurately identifies positive tau PET scans in AD.
• The THETA score provides a robust, quantitative measure of heterogeneous tau deposition.
• THETA shows significant potential for clinical application in quantifying tau PET in Alzheimer disease.