Lupus Nephritis is a chronic inflammatory kidney disease that arises in the context of systemic lupus erythematosus (SLE). It is characterized by the deposition of immune complexes in the kidney, localized inflammation, and ultimately kidney failure. In the systemic model of lupus pathogenesis, antigen presenting cells activate auto-reactive CD4+ T cells, which in turn provide help to auto-reactive B cells that differentiate into plasma cells that produce auto-antibodies. Many therapeutic approaches (such as B cell depletion) have been attempted to disrupt this pathway. However, despite the rational basis of these approaches, none have demonstrated robust performance in clinical trials. This suggests that our understanding of how the adaptive immune system contributes to disease burden is incomplete. This work aims to fill this gap by investigating inflammation that is localized in the kidney, using computer vision to identify cells in tissue, and spatial analysis to define how these cells are organized. We first related cellular features of inflammation with progression to renal failure using the biopsies of LuN patients for whom we have at least 2 years of clinical follow-up data. Five classes of cells were identified in these biopsies: CD4+ T cells, CD4- T cells, B cells, myeloid dendritic cells, and plasmacytoid dendritic cells. Two striking results emerged from this work—first, we observed that dense cellular neighborhoods of CD4- T cells are associated with progression to renal failure. Second, dense regions of B cells were found in a subset of patients who had preserved renal function. T cell and B cell phenotypes were further interrogated using a highly-multiplexed dataset from a separate cohort of LuN patients. The richness of markers in this second dataset allowed for investigation into the spatial distribution of more specific cell phenotypes. We showed that “CD4- T cells” are not exclusively CD8+ but are rather a diverse compartment that include CD3+CD4-CD8- (double negative) T cells, roughly 50% of which might be gamma-delta T cells. In addition, we found that regulatory T cells were relatively rare, while T follicular helper cells were abundant and frequently found in large cellular neighborhoods with B cells. Finally, we found that dense cellular neighborhoods often exist in the context of larger inflammation, while some smaller neighborhoods are isolated within the tissue. Analyzing both image datasets together allowed us to identify features of in situ inflammation that associate with patient outcomes in LuN and define metrics that can be used to evaluate inflammatory structures in tissue. Finally, we developed methods for improving cellular segmentation in inflamed tissue. We investigated the generalizability of cellular segmentation algorithms to other diseases states and demonstrated techniques by which a segmentation algorithm trained for lupus nephritis can be applied to triple negative breast cancer. We also evaluated the effects of sample preparation and staining panel choice on the performance of segmentation algorithms. These insights will aid the development of robust quantitative pipelines for understanding human tissue inflammation.