Missions

The postdoc will be working on the ultra-high-field platform at Poitiers University Hospital, within the labcom I3M laboratory and the Laboratoire de Mathématiques et Applications LMA CNRS 7348, University of Poitiers, as part of the MoGLIA regional project. This project focuses on the modeling of tumor growth. Tumor growth is a complex phenomenon, influenced by numerous biological, metabolic and environmental factors. Morphological magnetic resonance imaging (MRI) is widely used to detect and monitor tumors, thanks to its ability to provide detailed anatomical information. Complementing this, magnetic resonance spectroscopy (MRS) provides access to in vivo metabolic markers, such as lactate, offering a unique insight into biochemical activity within tumor tissue. The joint exploitation of these two modalities could significantly improve the prediction of tumor progression, but requires tools capable of integrating, modeling and interpreting this wealth of information. It is in this context that artificial intelligence (AI) approaches, particularly deep learning, offer considerable potential.

Activities

This post-doctoral project aims to develop a predictive model of tumor growth based on artificial intelligence (AI) approaches capable of integrating data from morphological MRI and MRS. The objective is to extract structural and metabolic biomarkers enabling precise spatio-temporal modeling of tumor evolution, for diagnosis, prognosis and personalized therapeutic follow-up.
Develop deep learning architectures capable of combining data from morphological MRI (T1, T2, FLAIR) and MRS (metabolic spectra) for better tumor characterization.
2. Design predictive models of tumor growth over time, integrating longitudinal MRI/MRS series, using spatio-temporal network approaches (Transformers, ConvLSTM, etc.).
3. Identify key metabolites (such as choline, lactate, NAA, etc.) correlated with tumor growth or aggressiveness, and integrate them into predictive models.
4. Ensure model transparency via explicability tools (SHAP, attention maps) and validate performance on annotated clinical databases.

Skills

The candidate will have the following skills :
- Experience in medical image processing, in particular MRI
- Knowledge of MR spectroscopy (or a strong desire to learn it quickly).
- Expertise of artificial intelligence techniques and deep learning
- Proficiency in Python, PyTorch/TensorFlow, and medical visualization tools (e.g., 3D Slicer, ITK-SNAP, MONAI...).
- Mastery of deep learning platforms Tensorflow/Pytorch/scikitlearn
- English: high level

Work Context

The selected candidate will work closely with multidisciplinary teams including radiologists, physicians and engineers to validate and evaluate the methods developed. The results obtained will be compared with clinical/preclinical references and interpreted in a medically relevant manner.

Constraints and risks

no constraints
no risk