Early detection of dementia through retinal imaging and trustworthy AI
Problem
Alzheimer's disease lacks a simple, affordable early detection method, leaving early-onset AD and mild cognitive impairment hard to identify.
Why it matters
Early detection of dementia could enable earlier intervention for a growing global healthcare challenge.
What the AI did
Eye-AD, a deep learning framework using multilevel graph representation of retinal OCTA images, detects early-onset Alzheimer's disease and mild cognitive impairment from retinal microvasculature.
Results
EOAD AUC 0.9355 internal, 0.9007 external; MCI AUC 0.8630 internal, 0.8037 external; 5751 OCTA images, 1671 participants, multi-center.
Impact
Multi-center external validation supports retinal OCTA imaging plus AI as a potential early dementia detection tool.
Record
| Venue | npj Digital Medicine |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Jinkui Hao, William Robert Kwapong, Ting Shen, Huazhu Fu, Yanwu Xu, Qinkang Lu, Shouyue Liu, Jiong Zhang et al. |
| Citations | 58 (OpenAlex, at import time) |
| References | 41 |
| Model or method | Eye-AD (deep learning, multilevel graph representation) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41746-024-01292-5 |
| Open access | gold |
| Keywords | Alzheimer's detection, OCTA, retinal imaging, deep learning, graph neural network, mild cognitive impairment |
Importance rating
significant — thousands to millions affected
Abstract calls AD a global healthcare challenge lacking simple, affordable detection; multi-center validation (1671 participants) of a noninvasive retinal approach suggests a genuinely new capability.