The goal of this PhD was to improve kidney fibrosis assessment by diffusion-weighted imaging (DWI). A respiratory implementation of the RESOLVE sequence improved significantly the image quality by reducing susceptibility effects and enhancing cortico-medullary Apparent Diffusion Coefficient (ADC) differentiation. In light of that fact, RESOLVE was tested in well-controlled models of fibrosis to separate pathological from healthy kidneys. These results were validated afterward in a cohort of kidney allograft patients. ∆ADC from RESOLVE had a stronger correlation to fibrosis than ∆ADC from classical single-shot Echo-Planar DWI. Most importantly, a negative ∆ADC was measured for all patients harboring more than 40% fibrosis, opening the possibility for a diagnostic threshold. In a second part, a motion compensation method was provided to remove signal dropout and the overestimation of ADC. By addressing and minimizing the main sources of the initial poor DW image quality, it was possible to improve renal fibrosis assessment.