A deep learning model for novel systemic biomarkers in photographs of the external eye: a retrospective study

AI Generated Summary*

A convolutional neural network was trained to predict kidney, liver, thyroid, mineral and blood count values from 123,130 external eye photographs of 38,398 diabetic patients at 11 Los Angeles County screening sites. Nine prespecified targets were tested in three validation sets (25,510 patients) from a Los Angeles County programme and Atlanta Veterans Affairs clinics against logistic regression baselines using clinicodemographic variables. In the set resembling the training population, AUC gains over baseline reached significance for AST, calcium, eGFR, haemoglobin, platelets, albumin-to-creatinine ratio and white cells, spanning 5.3 to 19.9 percentage points. In the two demographically distinct veteran sets, advantages persisted for ACR at or above 300 mg/g and, where available, haemoglobin under 11 g/dL (7.3 to 13.2 points).

*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.