SCSR Model Shows Cross-Population Transferability for Alzheimer's Detection
A study from arXiv (2608.07092) investigates the generalization of Stochastic Cortical Self-Reconstruction (SCSR), a method for mapping gray matter atrophy in neurodegenerative disorders like Alzheimer's disease (AD). SCSR estimates an individualized healthy reference from vertex-level cortical thickness, enabling detection of subtle deviations. Originally trained on UK Biobank (UKB) data, the model was tested on an independent Chinese population dataset. The study evaluates SCSR-derived Z-scores to discriminate between healthy controls, individuals with mild cognitive impairment (MCI), and AD patients. Results indicate promising cross-population transferability, suggesting SCSR's potential for personalized diagnostics across diverse populations. The work addresses limitations of conventional normative modeling, which operates at coarse regional levels and is constrained by training covariates. This research contributes to advancing neuroimaging-based biomarkers for early detection and monitoring of neurodegenerative diseases.
Key facts
- SCSR enables personalized mapping of gray matter atrophy onto high-resolution cortical surfaces.
- SCSR estimates an individualized healthy reference from vertex-level cortical thickness.
- The model was originally trained on UK Biobank (UKB) data.
- The study evaluates transferability to an independent Chinese population dataset.
- SCSR-derived Z-scores are used to discriminate between healthy, MCI, and AD patients.
- The approach detects subtle, subject-specific deviations from healthy cortical shape.
- Conventional normative modeling operates at a coarse regional level and is constrained by training covariates.
- The study is published on arXiv with identifier 2608.07092.
Entities
Institutions
- UK Biobank
- arXiv
Locations
- China
- United Kingdom