Work

AI enabled Tau–Neurodegeneration mismatch in Alzheimer’s disease

Year: 2026
Type: preprint
Abstract: The heterogeneity of Alzheimer's disease (AD) and lack of validated markers for non-AD co-pathologies limits targeted therapies. Our prior work showed that deviations ("mismatch") from a normative Tau (T)–Neurodegeneration (N) relationship can identify co-pathologies and resilience in AD. However, T–N relationships are nonlinear and spatially complex. In this study, we implemented a T–N mismatch framework using inter-modality image translation to synthesize "expected" cortical thickness maps from tau PET, and comparing these to observed thickness as a mismatch metric. T–N mismatch was quantified as region-wise “reconstruction error” (AI-predicted versus observed thickness) and subsequently used for data-driven clustering. The AI-predicted N aligned with the observed N while deviations still commonly existed. The AI-enabled T-N mismatch identified distinct data-driven T-N phenotypes associated with specific co-pathological signatures. Our findings support that AI-based T-N mismatch for identifying non-AD modulators in AD may have advantages over a statistical linear-regression approach.
Language: English
Field: Medicine
Open Access status: green
OpenAlex ID: W7236244949
Cites: 72
Cited by:
Related to: 10