Three-Dimensional Quantification of Macular OCT Alterations Improves the Diagnostic Performance of Artificial Intelligence Models

This repository provides the model checkpoints and dummy data of our abovementioned work.

Prerequisites

• Python3 >= 3.11.6 Get Python
• Conda (optional) Get Conda
• GPU (optional, recommended)

Getting started

1. Clone the repository

git clone https://github.com/TIO-IKIM/AMD_SemSeg.git

2. (Optional) Create a conda environment.
3. Install the requirements using pip or conda

pip install -r requirements.txt

Relabeled data

You can find a relabeled version of Chiu et al.'s data here.

Please find the original raw images here.

Inference

nn-UNet

For a detailed overview on how to use nnU-Net's checkpoints to make predictions please refer to the original repository.

Encoder only (embeddings)

You can find an example on how to load and use torch models here.

Inquiries

If you have any questions regarding the code, collaborations or different encoders, please raise an issue.

Citation

@article{10.1167/tvst.14.7.8,
    author = {Heine, Lukas and Vahldiek, Anna and Vahldiek Benja and Hörst, Fabian and Seibold, Constantin and Lever, Mael and Pauleikhoff, Laurenz and Bechrakis, Nikolaos and Pauleikhoff, Daniel and Kleesiek, Jens},
    title = {Three-Dimensional Quantification of Macular OCT Alterations Improves the Diagnostic Performance of Artificial Intelligence Models},
    journal = {Translational Vision Science & Technology},
    volume = {14},
    number = {7},
    pages = {8-8},
    year = {2025},
    month = {07},
    issn = {2164-2591},
    doi = {10.1167/tvst.14.7.8},
    url = {https://doi.org/10.1167/tvst.14.7.8},
    eprint = {https://arvojournals.org/arvo/content\_public/journal/tvst/938722/i2164-2591-14-7-8\_1752659133.67454.pdf},
}