Code from the UW-Madison Machine Learning for Medical Imaging (ML4MI) Boot Camp. For more information about ML4MI go to: https://www.radiology.wisc.edu/research/medical-imaging-machine-learning-initiative/
All the excercises are written in Keras which is corrently integrated into tensorflow. We use the Keras functional model which is a lot more flexible than the commonly used sequential model in examples.
Keras Documentations: https://keras.io/
- Alan McMillan ( AMcmillan@uwhealth.org )
- Jacob Johnson ( firstname.lastname@example.org )
- Kevin Johnson ( email@example.com )
- Tyler Bradshaw ( firstname.lastname@example.org )
Code has been tested on a machine with a NVIDIA K80 (11gb of GPU ram). To run this you need: python 3 ( https://www.python.org/ ) tensorflow ( https://www.tensorflow.org/install/ , install tensorflow-gpu if you have one)
We installed these with the following commands.
pip install tensorflow-gpu pip install keras pip install matplotlib pip install numpy pip install livelossplot pip install conda pip install jupyterlab conda install scikit-image conda install scipy conda install -c conda-forge --no-deps pydicom
Colab from Google Research
Some of these will run on the Google research supported Colab. This is a free cloud based enviroment supported by Google. You can click on the link in the source code or go to https://colab.research.google.com/
- FunctionFitting - Some very basic networks used for learning functions
- ImageReconstruction - Training of an neural network to reconstruct MRI images using 1D operations
- MaleFemaleRadiograph - Classify chest xrays as male or female
- ImageSegmentation - Lung segmentation from CT data (need to download data yourself)
Examples missing data (work in progress):
- AgeRegression - Regression for Age
- ImageSynthesis - Image synthesis of brats data
Note on commits:
If you aim to push changes to this repository, please edit the jupyter notebooks and then run clear_and_convert_bash (in linux or WSL). This will convert the notebooks to python and clear the ouput.