Research Questions:
1. Can Alzheimer’s disease (AD) be reliably detected through ocular biomarkers (Fundus and OCT imaging) with comparable accuracy to traditional neuroimaging methods?
2. What are the optimal AI architectures and preprocessing methods for analyzing Vietnamese population-specific imaging data for AD detection?
3. How does the integration of multimodal data (MRI, OCT, Fundus, and clinical data) improve the performance of AI-based AD diagnosis compared to single-modality approaches in the Vietnamese population?
4. How effectively can AI models predict AD progression at 12 and 24 months based on baseline imaging and clinical data, and which biomarkers best correlate with cognitive decline?
Objectives:
1. To develop a comprehensive clinical dataset, including Fundus images, OCT images, and structural brain MRI from 400 Vietnamese subjects to train and validate AI models.
2. To develop and evaluate AI diagnostic models achieving: !80% accuracy from ocular imaging alone; !85% accuracy from MRI alone; !90% accuracy from combined imaging data; and !93% accuracy when adding clinical data.
3. To develop progression prediction models estimating cognitive decline at 12 and 24 months based on baseline data.
Scientific Rationale: The scientific foundation for this research includes:
1. Established biological connections between retinal and brain pathology in AD, where shared neurovascular pathways and amyloid/tau accumulation manifest in both tissues
2. Documented performance variation when AI models trained on Western populations are applied to Asian cohorts (accuracy dropping from 96.8% to 79.2%)
3. Complementary information from different imaging modalities addressing distinct aspects of AD pathology
4. Ocular imaging advantages in capturing early vascular and neural changes preceding cognitive symptoms
5. Evidence that baseline imaging biomarkers correlate with future cognitive trajectory
