Mastery thesis entitled Segmentation and characterization of white matter hyperintensities in brain MRI images
This project is a teamwork of two people. The contributors:
The master's thesis was supported by lecturers from Poznan University of Technology.
The master’s thesis is based on artificial intelligence and neuronal network, which have been used to make segmentation models. The project consists of two parts. One of them has to separate brain tissue from the skull. The second task is a modification of the previous model. It has to mark every white hyperintensity matter in the brain. Data, which have been used in the project, comes from magnetic resonance imaging. The thesis addresses the proper way to calibrate the network to get the best metrics (recall, dice etc.). All results are presented as graphics. Effects are shown in 2D (two-dimensional) or 3D three-dimensional) images. Moreover, the project contains theoretical chapters, providing the information necessary to understand the topic of the thesis.
Our project is based on PyTorch Lightning. All data are represented via written scripts by ourselves and 3Dslicer, which is popular in medical aspects. All scripts to inspect the working of the Deep Learning are in the scripts directory. All required packages to train and run Artificial Neural Network are specified in install.txt.