Honours-Thesis-Project

Introduction

This repository includes all related codes in the honours thesis titled "Volumetric CT Small Bowel Segmentation using Deep Learning" from the University of Sydney, 2021.

Dependencies

• Python == 3.7.6
• numpy == 1.19.5
• matplotlib == 3.1.3
• SimpleITK == 2.0.2
• segmentation-models == 1.0.1
• Keras == 2.4.3
• tensorflow == 2.2.0

NOTE: You do not need all dependencies with the exact same versions to run the code.

Description

• 3D CT Registration: Implementation of the multi-step registration and segmentation generation pipeline based on SimpleITK
• 3D Patch Extraction: Implementation of the 3D image patch extraction process to generate the dataset
• IoU Calculation: Implementation of the IoU score calculation process to evaluate the generated segmentation
• UNet: Self implementation of 3D U-Net specific to this project based on Keras
• VNet: Self implementation of V-Net specific to this project based on Keras
• Train: Code used to load the dataset and train the model