This repository contains the code for my Mathematical Modelling & Computation MSc thesis on Axon segmentation using 3D Convolutional Neural Networks, conducted at the Technical University of Denmark (DTU) between February and July 2023.
The aim of this project was to explore and analyze the 3D microstructure of the brain's white matter by segmenting axons using 3D Convolutional Neural Networks in 3D synchrotron X-ray nano-holotomography volumes from the splenium of the vervet monkey brain, aiming to overcome the current limitations of diffusion magnetic resonance imaging (MRI) methods for measuring axon diameter.
The report concentrated on analyzing and preparing volumetric image data, creating a framework to train and evaluate the 3D U-Net model and the concept of Cross-hair filters for segmenting axons, and gaining insights into the shape of predicted axons. The main goal was to accurately label axonal structures within the provided data volumes.
In this repository:
- THESIS_compressed.pdf: full report in a compressed format that may involve lower image quality.