Justin Engelmann, Miguel O. Bernabeu. "Training a high-performance retinal foundation model with half-the-data and 400 times less compute." Nature Communications (2025). https://doi.org/10.1038/s41467-025-62123-z