Application of deep learning to retinal-image-based oculomics for evaluation of systemic health: a review

AI Generated Summary*

Across deep learning studies indexed through July 2022, color fundus photography dominated over optical coherence tomography (OCT). Internally, fundus models estimated age within 2.43 to 3.55 years and classified sex at AUCs near 0.95, dropping to 0.80 to 0.91 externally; cardiovascular event and risk AUCs spanned 0.70 to 0.88. Blood pressure, HbA1c, lipids and body mass index were predicted weakly, while thyroid function, C-reactive protein and blood cell counts proved unpredictable in one series. Adding clinical or demographic data improved kidney disease and anemia estimates. Multiple sclerosis classifiers using OCT measures, one combined with clinical data, reached 88 to 90 percent accuracy, though OCT work was generally less robust and seldom externally validated. Whether these tools improve outcomes remains untested.

*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.