진학프로석·박사 채용 접수 플랫폼

Scientific context

Unlike conventional external‑beam radiotherapy, targeted radionuclide therapy (TRT) delivers radiation at the cellular rather than anatomical scale by administering a radioactive compound linked to a cell‑targeting vector. In TRT, the therapeutic effect arises from the energy deposited locally by the emissions of the radionuclide. Personalized TRT requires patient‑specific dosimetry. With the advent of hybrid SPECT/CT and PET/CT systems, detailed information on the in‑vivo spatial distribution of tissue density and radiopharmaceutical activity is now available. In principle, this enables the estimation of the three‑dimensional absorbed‑dose distribution within the patient.

In this context, many recent approaches explore the use of artificial intelligence, particularly deep learning, to extract patient‑specific information from medical imaging and to predict clinically relevant quantities such as personalized dose, post‑treatment image enhancement, or protocol optimization. The overarching goal is to develop tools that personalize treatment while remaining reliable, fast, and compatible with clinical workflows. However, deep‑learning methods require large amounts of annotated data. Because TRT is a relatively new therapeutic modality, clinical datasets remain scarce, and for several radionuclides the field is still in the early clinical‑trial phase. Access to sufficiently large datasets is therefore impossible. The only viable solution is to generate synthetic, realistic data through large‑scale simulations. The ability to simulate and interact with a TRT treatment on a virtual patient is also essential for improving existing protocols or designing new ones for clinical studies. Such a framework makes it possible to explore therapeutic strategies safely, assess their impact under controlled conditions, and optimize clinical decision‑making before applying them to real patients.

To address this challenge, we have developed methods to create digital twins derived from real patient data, in which anatomical and physiological parameters can be modified to generate a virtual population. We have also designed simulation pipelines capable of producing synthetic PET and SPECT images (fig 1.), as well as absorbed‑dose maps using either Monte Carlo methods (high accuracy) or deterministic physics‑driven approaches (high speed). This TRT simulation platform enables the exploration of different imaging systems, radionuclides, and clinical protocols, and can generate training datasets for any AI‑based method.

The platform is still in its early stages. Our objective is to extend and refine it to further increase the realism of the simulations and to validate the generated data against clinical measurements.

Job description and missions

The main objective is to improve the generation of virtual patients, taking into account both anatomical realism and the pharmacokinetics of the radiotracer. A second objective is to evaluate and quantify the accuracy of the simulated data by comparing it with clinical datasets available from several collaborating institutions. This work will involve several key steps:

• • Gaining expertise in TRT clinical protocols and current clinical practice.
• • Enhancing the fidelity of virtual‑patient modelling, including organ‑level anatomical refinement, integration of radiotracer biodistribution derived from PET imaging, and incorporation of our pharmacokinetic model.
• • Improving the simulation of medical imaging (PET, SPECT) across different systems and acquisition protocols.
• • Evaluating and adjusting the platform by comparing simulated data with clinical SPECT, PET, and dosimetry datasets, particularly for 177Lu‑PSMA.
• • Exploring protocol optimization for personalized dosimetry and imaging in β‑emitter TRT (177Lu‑PSMA) and α‑emitter TRT (225Ac‑PSMA) using the simulation platform.
• • Publishing scientific papers on this new open‑source platform, its validation, and its application to various clinical questions related to personalized dosimetry in TRT.

Profile

PhD in medical physics. Good programming skills is a requisite, not to develop but just to be able to play and tune the existing program. Autonomy, open-mindedness and motivation, as well as good English speaking/writing skills, are also expected.

Position context

The postdoc will join the INSERM UMR1101 Laboratory of Medical Information Processing (LaTIM, Brest, France, https://latim.univ-brest.fr). The future recruited postdoc will work in collaboration with different academic and hospital partners within the context of different European projects. Access will be given to the PLACIS infrastructure (http://placis.univ-brest.fr/english) and to clinical data from our partners.

The position is available as soon as possible for one year. Salary is around 2100 € net/month, depending on the candidate’s experience.