Early tumor detection is crucial for maximizing treatment effectiveness and the chances of recovery. Currently, medical doctors rely on the combination of non-specific (X-ray, MRI, CT) or invasive (PET) imaging modalities and the analysis of tissue samples removed through biopsy, which is a painful process. Among non-invasive and harmless imaging modalities, Diffusion Magnetic Resonance Imaging (dMRI), which is sensitive to the displacement of water protons, has demonstrated its potential for microstructure imaging. This method assumes that water molecules probe their environment and, therefore, the dMRI signal reflects the tissue architecture and properties. To decode the information embedded in the dMRI signals, biophysical modeling of diffusion is commonly used. This technique involves deriving the equation of the dMRI signal from the tissue properties.

Various models have been proposed, each tailored for a specific tissue or use case. This study focuses on lymph nodes, an organ essential for metastasis detection. An important assumption of most biophysical models is negligible water exchange between compartments, which might not be valid in unmyelinated tissue. This necessitates specific models that include water exchange between the intracellular and extracellular compartments.

This work begins with the implementation of Monte-Carlo Diffusion Simulations (MCDS) in numerical models of lymph nodes, which requires adapting the simulator to account for permeable membranes. Building on these MCDS, a biophysical model tailored to the specific properties of lymph node tissue is proposed. The benefits and limitations of this model, as well as the commonly used Pulsed Gradient Spin Echo (PGSE) sequence, are examined across various numerical substrates.

These simulations shows that long diffusion times are required to increase sensitivity to the permeability regime. However, these long diffusion times are not achievable using the PGSE sequence. In the second part of this work, the Stimulated Echo Acquisition Mode (STEAM) is used to reach the targeted diffusion times. The impacts of this sequence on the model estimates are discussed using lymph node data, and the results are compared to the best models reported in the literature.