Predicting high coronary artery calcium score from retinal fundus images with deep learning algorithms

AI Generated Summary*

Fundus photographs fed an inception-v3 model trained to separate people with no coronary artery calcium from those with elevated scores, drawing on 44,184 images from 20,130 individuals (mean age 48.9) screened at a Korean hospital with same-day cardiac CT. Under five-fold cross-validation, AUROC for any calcium versus none was about 75%, rising to 82.3% with one eye and 83.2% with both at a threshold of 100, with no significant gain at 200, 300 or 400. Inpainting vessels cut AUROC by roughly two percentage points, the fovea by about one, and heatmaps centred on temporal vessels. Age alone gave 82.8%, age with sex and hypertension 88.3%, and adding images 88.6%; performance was judged insufficient for clinical deployment.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.