Repository architecture

Source code is split into separate files for better readability.

File data_loader.py: incl. DICOM images loading, normalization
File graph_cut.py : incl. intercative graph cut method using OpenCV library
File local_binary_patterns.py : incl. local binary patterns using scikit-image library
File metrics.py : incl. Dice-sorensen coefficient metric
File registration.py : incl. Rigid Registration using SimpleITK library
File thresholding.py : incl. Binary and Adaptive thresholding using OpenCV library
File utils.py : incl. Utility functions
File main.py - entry point to run methods

Actual training of the U-net method was implemented using Google Colab and Jupyter notebooks.

Notebook notebooks/U-net-training.ipynb - Includes U-net model, Augmentation configuration, Model training and Evaluation.

Source code originality and references

These parts of source code were copied and modified from internet:

graph_cut.py - Interactive method of Graph Cut was modified from original OpenCV repository of samples at github.com.
notebooks/U-net-training.ipynb U-net implementation in Keras was modified from this source at github.com.
registration.py: Code for Rigid Registration was modified from external source at official page of SimpleITK library. Code was introduces as sample for Rigid Transformation in jupyter notebooks at github.io.

All other parts of source do not include any copied or modified code from other sources.