Explainable foundation model for dementia screening and risk stratification using retinal fundus images
4 affiliations
- 1Medical Big Data Research Center, Seoul National University Medical Research Center, Seoul National University College of Medicine, Seoul, South Korea
- 2Department of Biomedical Sciences, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, South Korea
- 3XAIMED Co. Ltd, Seoul, South Korea
- 4Department of Family Medicine, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, South Korea
Abstract
Dementia is a growing global health challenge, and early identification is essential for timely intervention. We evaluated whether foundation model–based deep learning of retinal fundus photographs can detect dementia and predict future dementia incidence. Using health checkup data from over 36,000 Korean individuals, dementia was defined by diagnosis with relevant medication. Five vision foundation models were assessed with multiple fine-tuning strategies. The best-performing model, RETFound-MAE with partial fine-tuning, achieved an AUROC of 0.750 for dementia detection and a C-index of 0.812 for future incidence prediction. Model outputs remained independent risk factors after adjustment (adjusted odds ratio, 1.155; adjusted hazard ratio, 1.045). Quantitative saliency analyses highlighted biologically plausible regions, primarily the optic disc and adjacent peripapillary areas. These findings suggest that foundation model–based analysis of routinely acquired retinal fundus images may provide a scalable and interpretable approach for opportunistic dementia risk stratification, warranting validation in ethnically diverse populations.
Related
- Retinal reflectance spectra measured using hyperspectral imaging are associated with prevalent dementia in a general older Japanese population
- Retinal AI risk stratification and cardiovascular event yield across QRISK3 categories in UK Biobank
- CERVEX: A Foundation-Model Framework With Built-In Explainable AI for Automated Cervical Cytology Classification From Pap Smear Images
- Retinal microvasculature alterations are associated with mild behavioral impairment in Southeast Asian dementia-free older adults.
- Extracellular vesicles in vascular cognitive impairment and dementia: a neurovascular evidence map for patient stratification and clinical translation
BibTeX
@article{han2026explainable,
title = {{Explainable foundation model for dementia screening and risk stratification using retinal fundus images}},
author = {Changho Han and Jaewon Kim and Hyeokjong Lee and Hyunwoo Joo and Jooyoung Chang and Sang Min Park},
journal = {NPJ Digital Medicine},
year = {2026},
month = jul,
doi = {10.1038/s41746-026-02968-w},
url = {https://doi.org/10.1038/s41746-026-02968-w},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Explainable foundation model for dementia screening and risk stratification using retinal fundus images
AU - Changho Han
AU - Jaewon Kim
AU - Hyeokjong Lee
AU - Hyunwoo Joo
AU - Jooyoung Chang
AU - Sang Min Park
PY - 2026
DA - 2026/07/04
JO - NPJ Digital Medicine
DO - 10.1038/s41746-026-02968-w
UR - https://doi.org/10.1038/s41746-026-02968-w
AB - Dementia is a growing global health challenge, and early identification is essential for timely intervention. We evaluated whether foundation model–based deep learning of retinal fundus photographs can detect dementia and predict future dementia incidence. Using health checkup data from over 36,000 Korean individuals, dementia was defined by diagnosis with relevant medication. Five vision foundation models were assessed with multiple fine-tuning strategies. The best-performing model, RETFound-MAE with partial fine-tuning, achieved an AUROC of 0.750 for dementia detection and a C-index of 0.812 for future incidence prediction. Model outputs remained independent risk factors after adjustment (adjusted odds ratio, 1.155; adjusted hazard ratio, 1.045). Quantitative saliency analyses highlighted biologically plausible regions, primarily the optic disc and adjacent peripapillary areas. These findings suggest that foundation model–based analysis of routinely acquired retinal fundus images may provide a scalable and interpretable approach for opportunistic dementia risk stratification, warranting validation in ethnically diverse populations.
SN - 2398-6352
ER -