jaburke166/deepgpet

Repository storing fully automatic DL-based model and mode weights for choroid region segmentation in optical coherence tomography images.

# DeepGPET: fully automated choroid region segmentation in OCT ## DeepGPET DeepGPET is a fully automatic, deep learning based method for choroid region segmentation in optical coherence tomography (OCT) scans. This algorithm can be used to segment the choroid region, and then subsequently measure clinically relevant metrics such as choroid thickness, area and volume. The method was trained on labels from a previously published, semi-automatic algorithm [GPET](https://ieeexplore.ieee.org/document/9623378).  The description of the model can be found [here](https://tvst.arvojournals.org/article.aspx?articleid=2778573). The paper for this methodology has been published in Translations Vision Science & Technology, Association for Research and Vision in Ophthalmology (ARVO). --- ### Project Stucture ``` . ├── choseg/ # core module for carrying out choroid region segmentation and derived regional measurements ├── notebooks/ # Jupyter notebooks which stores example data and a demo of the pixel GUI to select the fovea ├── install/ # Files to help with installation ├── .gitignore ├── README.md └── usage.ipynb # Anaconda prompt commands for building conda environment with core packages ``` - The code found in `choseg` ``` . ├── choseg/ ├───── metrics/ # Code to calculate downstream regional measures such choroid thickness, area and (subregional) volume ├───── __init__.py ├───── inference.py # Inference classes for segmenting └───── utils.py # Helper functions for plotting and processing segmentations ``` - The code found in `choseg/metrics` ``` . ├── metrics/ ├───── __init__.py ├───── choroid_metrics.py # Code to calculate choroid measurements such choroid thickness, area and (subregional) volume ├───── choroid_utils.py # Utili