stereocord/LWBNA-MTL: LWBNA-MTL v1.0.0 – Code for Retinal Age Prediction Study

This release corresponds to the exact code used in the manuscript: "Retinal Fundus Photographs for Simultaneous Estimation of Biological Age and HbA1c." Contents: train.py: training and validation (with k-fold and early stopping) test_batch.py: batch inference for ensemble predictions lwbna_net2.py: lightweight backbone network with bottleneck-attention customoph.py: preprocessing routines requirements.txt: environment specification Note: Dataset is not included and must be prepared externally. Users should organize fundus images and associated DataFrame (with columns FILEPATH, Eye, Age, HbA1c) as documented. Environment: Python 3.10 (recommended) PyTorch 2.x CUDA 11.x (if using GPU) Reproducibility: Experiments reported in the manuscript were conducted using this version. Model weights are not included due to data-sharing restrictions. A DOI will be issued via Zenodo upon publication. For citation, please refer to the associated manuscript.