farris / Capstone2021

Estimating intracranial pressure using OCT scans of the eyeball

Members: Eric He, Farris Atif, Nasser Al-Rayes, Zixiao Chen

Estimate intracranial pressure (ICP) given OCT scans and intraocular pressure (IOP) values. See our project poster.

Repository Organization

---

├── LICENSE
├── README.md          <- The top-level README for developers using this project.
├── data
│   └── raw            <- The original, immutable data dump.
│
├── docs               <- Presentation information
│
├── models             <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks          <- Jupyter notebooks for cleaning data and exploratory data analysis
│
├── requirements.txt   <- The requirements file for reproducing the analysis environment, e.g.
│                         generated with `pip freeze > requirements.txt`
│
├── setup.py           <- makes project pip installable (pip install -e .) so src can be imported
├── src                <- Source code for use in this project.
│   ├── __init__.py    <- Makes src a Python module
│   ├── data           <- Scripts to download or generate data
│   ├── models         <- Scripts to train models and then use trained models to make
│
└── tox.ini            <- tox file with settings for running tox; see tox.readthedocs.io

---

Important links

Google Drive: holds raw data, reports

• Raw data: copy of the original monkey scans
• PyTorch reduction: PyTorch arrays of reduced monkey scans, NOT STANDARDIZED
• PyTorch standardized images: Pytorch tensors of reduced and standardized monkey scans (what we use for training)
• Image samples: Examples of images after downsizing (but not standardizing) - taken from the PyTorch reduction folder
• Master Dataset: holds master mappings from the raw data to our image samples, with IOP values filled in. Replicated to the repository
• Models: holds trained models from Google Colab

Data processing

PyTorch reduction: code to downsize OCTs into PyTorch reduced arrays

Model training

From-scratch training: runs training using our forked 3-D resnet training code

Pre-trained: runs training using our forked MedicalNet model

Self-supervised: use the MoCo self-supervised learning to pre-train model