To better understand the role of stromal cells in cancer progression, there is a growing interest in 3D patient-derived models for their translational and throughput advantages. In parallel, the emergence of multiplexed spatial biology platforms is advancing translational oncology forward by providing a more global understanding of the tumor microenvironment (TME). Despite the enthusiasm for these new technologies, there are no studies showing the applicability and significance of using multiplexed spatial biology to characterize 3D patient-derived models. Here, we modeled spatial heterogeneity of lung adenocarcinomas (LUAD) by assembling EGFR mutated patient-derived organoids (PDOs) with fibroblasts harvested from the tumor edge (tumor-adjacent fibroblasts; TAFs) or tumor core (tumor core fibroblasts; TCFs).  These mixed cellular organoids are referred to here as tumor-stroma assembloids. Using these assembloids, we generated an extensive dataset with CODEX imaging technology and used a quantitative computational approach to dissect crosstalk between tumor cells and the associated stromal elements. Our results show that the assembloids undergo drastic spatial reorganization following erlotinib treatment even though subpopulation fractions and the degree of cell heterogeneity remain unchanged. This study provides new insights on tumor-stroma crosstalk in the LUAD and motivates the use 3D patient-derived models coupled with multiplex imaging to answer complex questions related to the TME.