Lymphoma comprises a highly heterogeneous group of neoplasms arising from diverse immune cell precursors. Subtypes differ markedly in their localization, cell of origin, molecular profiles and clinical outcomes. Understanding the origins of the disease is of great scientific interest since lymphoma affects millions of people globally, climbing to the top 10 most frequent cancers. In this light, we employed spatial transcriptomics using the 10X Genomics Visium platform on tissue samples from a cohort of 107 patients. Unlike bulk RNA-sequencing, spatial transcriptomics retains the tissue architecture, enabling high-resolution mapping of gene expression within the tumor microenvironment (TME). A major challenge in lymphoma is accurately delineating tumor cells from the surrounding TME, as malignant cells themselves originate from B and T lymphocytes, making classification a complex process. To address this, we applied a holistic, pre-trained cell deconvolution method, integrating gene expression and spatial context, and validated its performance against a single-cell RNA-seq reference. Our analysis revealed conserved cell-type frequencies across lymphoma subtypes as well as cell-type niches, which highlighted the role of macrophages in the lymphoma TME. This approach provides a granular perspective on the lymphoma ecosystem and lays the groundwork for identifying spatially informed biomarkers and therapeutic targets.