This repository contains the code used in my master project at the university of Copenahgen supervised by Christian Igel and Erik Dam, in collaboration with Biomediq.
The aim of this project was to compare 2D, 3D and pseudo-3D Convolutional Neural Networks for the semantic segmentaion of knee cartilage from MRI scans.
This used the Lasagne library in order to build various architectures for segmentaion. Other data sampling and pre-processing was also employed to increase segmentation accuracy measured by Dice Sørensen Coefficent.
Full details of the project and architecture can be read in the project report.