ben-heil / DICOM-CNN

Convolutional neural network solution for deep learning from DICOM images

DICOM CNN

Scripts useful for doing deep learning on DICOM images. Used to differentiate between images with and without lung nodules for my semester project in machine learning.

Note:

While you are welcome to borrow ideas and code from this repo, other models performed much better in the competition. If you are interested in DICOM image segmentation in general, you may want to read about some of the winning entries in the 2017 Data Science Bowl

utilityFunctions.py

A series of functions for converting DICOM images into numpy arrays

CNN.py

The most successful architecture I tested

writeup.pdf

A paper detailing background information and different methods of lung nodule detection