Retinal image-based artificial intelligence in detecting and predicting kidney diseases: Current advances and future perspectives

AI Generated Summary*

Surveying work on eye-based machine learning for kidney disease, the review first covers retinal imaging in ophthalmology and in diabetes, cardiovascular and Alzheimer disease, then the shared development and vascular biology of eye and kidney. For chronic kidney disease, it describes fundus models: one multi-ethnic algorithm of nearly 13,000 people reached internal AUCs near 0.91, falling to 0.73 to 0.86 externally, with positive predictive value of 14% and 9% outside the development set. A ResNet-50 model built from 57,672 patients reached AUC 0.918 on images alone and predicted eGFR. Later sections cover dialysis hypotension, anemia, progression, and mortality. Cited limits are image quality, single-task design, opacity, privacy, and dataset bias.

*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.