Segmentation of dynamic contrast-enhanced ultrasound of the liver
Level: Master project at EPFL
The purpose of this work is to help characterize focal liver lesions from dynamic contrast-enhanced ultrasound (DCE-US) clips. DCE-US images are obtained by injecting a micro-bubble contrast agent in the blood stream. These bubbles resonate when insonated by the ultrasound imager and become visible. By imaging an organ for several minutes, one can visualize the flow of the contrast agent through its tissues and characterize their perfusion. We want to classify benign vs. malignant lesion types from a comparison of dynamic vascular patterns in lesions, vs. those in normal parenchyma (liver tissue). At present, the operator draws normal parenchyma regions manually. In this project, we want to implement a variational algorithm similar to the active contour without edges algorithm by Chan and Vese. The challenge consists of finding appropriate dynamic features in the DCE.
Requirements: The project will be developed mainly in C++ (experience in Matlab is a plus). Basic knowledge in image processing is required.
Supervision: Meritxell Bach Cuadra (Meritxell.BachCuadra@unil.ch), Olivier Cuisenaire (Olivier.Cuisenaire@unil.ch).
Direction: Prof. Jean-Philippe Thiran.