Deep learning fundus photo analysis screens for anemia with high accuracy.
Transforming Fundus Photography for Deep Learning-Based Anemia Screening.
Taeseen Kang, Kiyup Nam
Journal of clinical medicine · 2026
Background
This study aimed to develop and evaluate a noninvasive anemia screening method using fundus photographs analyzed by deep learning models, by transforming circular fundus photographs into square images suitable for convolutional neural network analysis.
Purpose
This study aimed to develop and evaluate a noninvasive anemia screening method using fundus photographs analyzed by deep learning models, by transforming circular fundus photographs into square images suitable for convolutional neural network analysis.
Methods
A total of 39,036 fundus images and clinical data collected from 2011 to 2023 were used.
n = 39,036 fundus images
Results
0.893
higher score means the eye-scan test more reliably distinguishes anemia from no anemia
n = 39,036 fundus images
More results
Among preprocessing methods, the stretching method yielded the most accurate predictions.
“
Conclusion
Deep learning analysis of fundus photographs demonstrates potential as a noninvasive screening method for anemia. This approach may be particularly beneficial for patients requiring regular hemoglobin monitoring.