(c) 2020, Mayo Clinic Radiology Informatics Lab
Project Overview
- Install Python 3.6 from: https://www.python.org/downloads/
- (Recommended) In a terminal, create a new Python environment using venv:
python3 -m venv py36
source py36/bin/activate
pip install -U pip - Install Tensorflow and Nibabel into the newly-created "py36" environment using:
pip install tensorflow==2.2.0
pip install nibabel - Clone the GitHub repository to disk.
- Download the model's weights and place them in the same folder as
z_controlboard.py
Weights for the training dataset only (40 normal examinations).
OR
Weights for the primary dataset and the iNPH dataset (50 normal examinations + 12 examinations demonstrating ventricular enlargement; recommended for routine use). - Open a terminal and type:
python /path/to/z_controlboard.py
Further instructions are found in the module.
If you would like to use the model in its training state, please comment out lines 12-24 and uncomment lines 28-56 in z_controlboard.py. We provided 3 sample volumes in the "image_data" and "mask_data" folders for this demonstration.
Please note that SciPy is required for the image augmentation module (pip install scipy).
• RIL-Contour is a medical image annotation tool developed by our lab. It can run Tensorflow Keras models through a user interface. The instructions for downloading, installing and navigating RIL-Contour are available here.
• A tutorial showing how to run our model in RIL-Contour is available here.
JC Cai, Z Akkus, KA Philbrick, A Boonrod, S Hoodeshenas, AD Weston, P Rouzrokh, GM Conte, A Zeinoddini, DC Vogelsang, Q Huang, BJ Erickson
“Fully Automated Segmentation of Neuroanatomy on Head CT Using Deep Learning”
Radiol Artif Intell. 2020 Sep; 2(5):e190183. https://doi.org/10.1148/ryai.2020190183
Click here to download citation data.
For inquires, please email jason.cai outlook com