Segmentation methods for low-field MRI scans
Ksenia Slepova

Supervisor TU Delft: Martin van Gijzen

start of the project: January 2023

Summary of the master project:
Tens of thousands of children suffer from hydrocephalus in sub-saharan Africa. The most effective tool of diagnosing this disease is Magnetic Resonance Imaging (MRI), however this requires an expensive and complicated hardware, which is rarely accessible in resource-poor countries. Recently, a prototype of inexpensive and sustainable device has been designed in a joint project of TU Delft and multiple organisations in Uganda, the Netherlands and the USA, which would make testing for this disease accessible to more patients. However, scans obtained from this MRI tools are noisy and need to undergo enhancement process to be used in diagnostics.

The main goal of this thesis project is to develop segmentation methods for low-field MRI scans which facilitate hydrocephalus detection. Segmentation is a process of extracting the desired object (organ, blood vessel, etc.) from a medical image. The first stage of the method will consist of noise reduction of the input images which are often blurry and cannot be immediately used for segmentation. Next, developed segmentation algorithm will be applied to sharpened images. Finally, the whole method will be validated by testing it on MRI scans obtained by researchers in Uganda.

Contact information: Kees Vuik

Back to the home page or the Master students page of Kees Vuik