Author: Paras Lakhani, paras.lakhani@jefferson.edu
More details and a step-by-step guide for the tutorial can be found in the Journal of Digital Imaging Publication (DOI: 10.1007/s10278-018-0079-6; https://pubmed.ncbi.nlm.nih.gov/29725961/), which is the official journal of the Society of Imaging Informatics in Medicine (SIIM).
This is a high-level introduction into practical machine learning for purposes of medical image classification.
In this tutorial, we use the Tensorflow framework and the Keras library, which a high-level application programming interface that simplifies working with Tensorflow.
We hope that this tutorial will spark interest and provide a basic starting point for those interested in machine learning in regard to medical imaging.
A Jupyter ipython notebook is provided called "HelloWorldDeepLearning.ipynb"
We provide 75 images, 38 are chest X-rays, and 37 are abdominal X-rays. These de-identified PNGs obtained from openI, https://openi.nlm.nih.gov/, a searchable online repository of medical images from published PubMed Central articles
The goal of this tutorial is to build a deep learning classifier to accurately differentiate between the two.
You'll need a computer with the following installed:
- Tensorflow (https://www.tensorflow.org)
- Keras library (https://keras.io)
- Jupyter (http://jupyter.org)
- Download the x-rays provided in .zip file
To make things easier, there is a convenient SIIM docker that has Tensorflow / Keras / Jupyterlab already installed, located here: https://github.com/ImagingInformatics/machine-learning/tree/master/docker-keras-tensorflow-python3-jupyter
After your environment is set up, open the ipython notebook, and run the code!