Annonce

Since its introduction in the early 2000s, three-dimensional (3D) visualization has been adopted to varying degrees in surgical planning workflows for hepatobiliary surgery worldwide. The visuospatial advantages of examining a 3D model over conventional assessment of 2D slices are intuitive, with well-documented benefits in resection planning, intraoperative navigation, and the selection of living donor procedures. Conventional medical imaging modalities often provide limited visualization of the fine vascular anatomy of the liver, making it difficult to accurately appreciate the spatial relationships between vessels and surrounding structures, such as possible lesions or tumors. Furthermore, definition of smaller parenchymal subunits does not have clear grounding in fine-vessel data, making existing clinical subsegment definitions arbitrary.

The main goal of the collaborative project surrounding this internship is to develop a pilot deep learning model capable of generating fine branch vessels from low resolution medical images (e.g. CT scans), by training on paired high-resolution datasets (micro-CT) obtained from the same subject as part of organ donation. This is a unique opportunity to work with inter-modality data bridging clinical and experimental ex-vivo imaging platforms. The intern will have the opportunity to work with various data across both scales:

• raw CT images;
• liver volume segmentations;
• vessel segmentations;
• Couinaud zoning;
• vessel graphs and landmarks.

The main objectives of this internship are the following, and may be adapted depending on the intern’s skills and progress during the project:

• Literature review about fundamentals and recent deep learning models for liver vessel segmentation.
• Address one of these tasks: 1) Co-registration between the two modalities, by using additional data (e.g. centerlines, landmarks); 2) Combined registration and segmentation between modalities; 3) Generate accurate and plausible hepatic vasculatures in CT domain, beyond imaging resolution, thanks to micro-CT data.
• Test deep models from state of the art if available.
• Design, implement, test and evaluate one or more new deep models.
• Co-author a scientific paper based on these models and findings.

The candidate should have strong skills in computer programming, preferably in Python, and have good experience with deep learning modules. Previous experience with open-source platforms (e.g. 3D slicer) and knowledge in image processing and analysis would be an added-value. We are looking for a rigorous and well-organized candidate with strong written and verbal communication skills, as well as the ability to work independently and take initiative.

• You will work in a professional environment that will draw on the skills and knowledge acquired during your training, as well as on your ability to innovate and adapt.
• You will have the opportunity to work with exceptional, high-quality medical imaging data.
• The internship environment brings together strong expertise in artificial intelligence and software development, providing you with a stimulating setting for your internship. A team of physicians is also highly involved in this international collaborative project.
• The dissemination and valorization of this work offer strong potential for scientific communications.

Please send an email to all internship supervisors as soon as possible and before 06/11/2026, including the following documents:

• cover letter;
• most recent grades;
• up-to-date CV.

An interview is expected to be arranged following your application.

Co-supervisors of this internship are:

• Ms Noémie Debroux – noemie.debroux@uca.fr
• Mr Antoine Vacavant – antoine.vacavant@uca.fr
• Dr Kah Heng Alexander Lim – kah.lim@research.uwa.edu.au

Institut Pascal, Image Guided Therapy – http://www.institutpascal.uca.fr

University of Western Australia – https://www.uwa.edu.au/

IUT Le Puy-en-Velay, France -📍 https://maps.app.goo.gl/VKqtNa6aPmaLwMTY8

The intern will be paid 4.50€/hour.