Assessing the arterial wall compression in ultrasound image sequences using deep learning and realistic simulations
Recrutement en cours/passé: 
Recrutement en cours
M. Maciej Orkisz, maciej.orkisz[at] M. Hervé LIEBGOTT, liebgott[at]

Objectives of the project: The main goal is to develop a “2-in-1” simulation method for US image sequences combining a very good control of geometry and deformations of the modelled arterial wall and high realism of the surrounding tissues mimicking clinical images. Depending on the candidate’s interests and skills, optimization of the deep-learning-based method developed in the team for the wall segmentation in image sequences can be foreseen, to improve its ability to quantify the wall cyclic compression.

Scientific challenges: The main challenge is a seamless immersion of an in-silico artery model in the anatomic context from a real clinical image. Such a 2-in-1 method has never been developed in the vascular field. Nevertheless, a similar approach has been successfully developed in the team for the cardiac imaging, we also implemented a simulation pipeline to obtain realistic vascular images based on clinical ones, and the modelling of the artery may be inspired by the recent literature.

Expected candidate profile (prerequisite): understanding the physical processes underlying the ultrasound image acquisition, programming, image processing, deep learning, interest for biomedical field and biomechanical modeling for health sector.

More details in the attached file.