Effect of cataracts on systemic factor prediction from fundus images

Abstract Objective To evaluate the effect of cataracts on systemic factor predictions from fundus images by comparing predictive values before and after cataract surgery. Method This Retrospective study includes consecutive patients who underwent cataract surgery between January 2024 and July 2024 at Kagoshima University Hospital. Pre- and postoperative fundus images were obtained. Systemic factors were predicted from these images using publicly available deep learning models that had been previously trained on a Japanese health checkup dataset consisting of 8,149 images for age and approximately 128,000 images for other factors. Main outcomes were changes in predicted clinical profiles—including retinal age, body mass index, blood pressure, abdominal circumference, and blood glucose—following cataract surgery. Results We included 67 patients (101 eyes; 37.3% male) with a mean age of 72.5 ± 11.4 years. More than half of the participants were receiving treatment for hypertension, diabetes mellitus, or were using steroids. Among all analyses, only chronological age was significantly associated with retinal age, both preoperatively (β = 0.48, P = 8.6 × 10 − 8 ) and postoperatively (β = 0.47, P = 3.1 × 10 − 9 ). The mean preoperative retinal age was 72.1 ± 8.2 years (95% confidence interval [CI], 70.4–73.7), which significantly decreased to 64.5 ± 7.2 years (95% CI, 63.1–65.9) postoperatively ( P = 4.1 × 10 − 16 ). These findings remained consistent in a subgroup analysis of patients without fundus diseases who underwent only cataract surgery ( P = 3.8 × 10 − 10 ). Conclusions Our findings demonstrate that retinal age reflects not only the status of the retina and optic disc but also the presence of coexisting cataracts.