Mathematically-founded deep learning methods for image reconstruction in Compton camera SPECT

The CREATIS laboratory announces the opening of a 36-month, fully funded PhD position starting in September 2025. For details, https://www.creatis.insa-lyon.fr/site/en/recrutement/mathematically-founded-deep-learning-methods-image-reconstruction-compton-camera-spect .

Context and Objectives

Single Particle Emission Computed Tomography (SPECT) imaging is undergoing a significant revival, driven by new clinical needs -particularly in oncology - and recent advances in detector technology. Within this context, the TomoRadio team at CREATIS is opening a fully funded three-year PhD position aimed at developing mathematically-grounded deep learning methods for image reconstruction tailored to a novel Compton camera system. This project is part of the HORIZON-EURATOM AIDER project(Advanced Imaging DEtector for targeted Radionuclide therapy), an international collaborative project between partners in France, Spain, Italy, Germany. It aims to design and validate a new Compton camera prototype along with cutting-edge image reconstruction algorithms.
The successful candidate will contribute to the development of deep learning functionalities within CoReSi, an open-source reconstruction code, leveraging the framework of convergent Plug-and-Play methods for inverse problems.

The thesis will be carried out in the medical imaging laboratory, in the Tomoradio team, which specialises in inverse problems for imaging. The methods developed relate to physical modelling, optimisation algorithms and their connection through deep learning. The thesis will be supervised by Voichita Maxim (full professor, CREATIS laboratory and INSA Lyon) and Etienne Testa (Senior Associate Professor, IP2I laboratory and Claude Bernard University, Lyon).

We are looking for a highly motivated candidate with a solid background in applied mathematics, physics or signal/image processing. Familiarity with deep learning and scientific programming with Python and PyTorch is highly desirable. Experience with resolution of inverse problems, optimization, physical modeling will be a strong asset. Strong academic records and genuine motivation will be valued more than the specific background of the candidate.