A Systematic Review and Meta-Analysis of Applying Deep Learning in the Prediction of the Risk of Cardiovascular Diseases From Retinal Images
Problem
Evidence on whether deep learning can predict cardiovascular disease risk from retinal images was scattered and needed systematic synthesis.
Why it matters
Reliable CVD risk prediction from retinal images could aid cardiovascular risk assessment.
What the AI did
A systematic review and random-effects meta-analysis pooled performance of deep learning models predicting CVD risk from retinal images across 26 studies, graded with QUADAS-2.
Results
26 studies included; 42 CVD risk-related outcomes identified (33 risk factors, 4 cardiac imaging biomarkers, 2 risk scores, presence and incident CVD); pooled AUC values truncated in abstract.
Impact
Synthesizes pooled performance evidence across 26 studies of DL-based CVD risk prediction from retinal images.
Record
| Venue | Translational Vision Science & Technology |
|---|---|
| Published | 2023 |
| Work type | review |
| Authors | Wenyi Hu, Fabian S. L. Yii, Ruiye Chen, Xinyu Zhang, Xianwen Shang, Katerina Kiburg, Ekaterina Woods, Algis Jonas Vingrys et al. |
| Citations | 36 (OpenAlex, at import time) |
| References | 76 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1167/tvst.12.7.14 |
| Open access | gold |
| Keywords | deep learning, retinal images, cardiovascular disease, meta-analysis, risk prediction |
Importance rating
significant — thousands to millions affected
CVD risk prediction from retinal images addresses a major disease area; meta-analysis of 26 studies and 42 outcomes, but abstract states no population-scale impact figures.