Age and sex affect deep learning prediction of cardiometabolic risk factors from retinal images

AI Generated Summary*

Using a MobileNet-V2 network, fundus photographs from 3,000 Qatari citizens in the Qatar Biobank were analysed for cardiometabolic risk factors, with 7,200 of the 12,000 images used for training. Combining each person's four images (macula and disc centred, both eyes) gave a mean absolute error of 2.78 years for age (R2 0.89) and an AUC of 0.97 for sex, plus acceptable results for systolic pressure (MAE 8.96 mmHg), diastolic pressure (6.84 mmHg), HbA1c (0.61%), relative fat mass and testosterone; lipids, glucose and insulin fared poorly. Testosterone predictions appeared to track sex, age predictions did not, and age and sex only partly explained the blood pressure, HbA1c and fat mass results in this single Middle Eastern cohort.

*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.