Classifying Cancer
Classifying Cancer is a Python3 project to classify cancer data using Google's TensorFlow library and Neural Networks. The goal of this project was to validate and demonstrate that modern machine learning techniques in neural nets could prove to be useful in classifying cancer datasets.
The motivation for applying neural nets at cancer in particular came from Cancer Research's Citizen Science. This is a project that relied on volunteers to classify images of breast cancer tumours. The images themselves contained a mixture of different looking cells. Despite having over 2,000,000 contributions, the project struggled to differentiate cancer cells from non-cancer cells. Relying on volunteers to manually classify cancer seemed both inefficient and ineffective, and I believed that neural nets could provide a better method for classifying cancer.
This repo contains two main sections:
- dnn_data_classifier - A Deep Neural Network implementation to classify breast cancer tumours as benign or malignant depending on measurements taken directly from tumours.
- cnn_image_classifier - A Convolutional Neural Network implementation to classify breast tumours as benign or malignant using images of histology slides.
Getting Started
Prerequisites
Installation
-
Install Python3 on your Operating System as per the Python Docs. Continuum's Anaconda distribution is recommended.
-
Clone the repo:
git clone https://github.com/jhole89/classifying-cancer.git -
Set the environment:
pip install -r requirements.txt(Optional: If applicable you can compile Tensorflow for GPU to achieve significant performance increases)
Execution
-
Go to the sub-project directory:
cd classifying-cancer/dnn_data_classifierorcd classifying-cancer/cnn_image_classifierdepending on whether you want to classify tumour measurements or tumour images (it is recommended that you read the relevant README's first). -
Run the relevant main.py script:
python main.pyorpython cnn_image_classifier/main.py.
Coding style
Classifying-Cancer is PEP8 complaint but uses a max-line-length=100. This can be checked from the command line with:
pep8 --statistics --max-line-length=120 classifying-cancer
Built With
Contributing
As I consider this project to be closed I will not be looking to add any additional features into this project. However if you feel like contributing then feel free to issue Pull Requests. Any further development or Fork of this project is bound by the sample license of its parent.
Authors
- Joel Lutman
License
This project is licensed under the GNU GPL3 License - see the LICENSE file for details
Acknowledgments
This project makes use of two core data sets for our model training. Both are brilliant resources for machine learning and I highly suggest reading the relevant papers listed.