Acute leukemia has a high mortality rate of approximately 50%, and current methods are not effective in predicting disease progression and relapse. To improve our understanding of hematopoiesis and develop new markers for predicting disease relapse in deadly bone marrow (BM) diseases like acute myeloid leuke-mia, it is important to uncover the cellular composition of the BM stroma using quantitative and spatial in situ information. Recent research has shown that diseased BM stroma can induce a pre-leukemic state, myelodysplastic syndrome, and then lead to leukemia progression. However, the lack of specific markers and technical challenges have prevented the thorough characterization of human BM stroma. The aim of this thesis is to define a set of morphological features and specific markers for characterizing the non-hematopoietic bone marrow stroma in a quantitative manner. This can be used to evaluate the quality of the stroma and complement Minimal Residual Disease (MRD) evaluation to predict disease progression and relapse in acute myeloid leukemia patients. Computational pathology, cell surface stromal markers, histo-immunochemistry, multiplexing immunostaining, 3D imaging, and microfluidics were used as methods to tackle the previous hypothesis. We first validated a new computational pathology tool, MarrowQuant 2.0, for quantifying the hematopoietic, vascular, adipocytic, and trabecular bone fractions within human BM. A large panel of human stromal cell surface markers was tested and compared to assess in vivo BM stromal heterogeneity on BM trephine biopsies and on hematon units as an ex-vivo model of human BM. A defined panel of BM stroma markers was generated and tested on control hBM samples. The resulting plug-in (MarrowQuant 2.0) and stromal panel were then tested on longitudinal BM trephine biopsies from pa-tients undergoing chemotherapy for acute leukemia to determine disease evolution. A specific stromal compartment of interest was the bone marrow adipocytes. I was also involved in the development of a tool to isolate intact bone marrow adipocytes, which are challenging to study as a pure population. I was able to highlight the BM stromal population and develop tools and models to track its evolution and correlate the quantification outcome with disease evolution in the context of hematological malignancies. Overall, this personalized health-oriented project has important implications for computational quantita-tive pathology, hematopathology, hematological malignancies, and precision medicine. It highlights the importance of understanding the BM stroma for predicting disease relapse in acute myeloid leukemia pa-tients and developing targeted therapies.