The tumor-centric neighborhoods which are defined by their proximity to their nearest immune cell, indicated that dispersed regions had the closest distances to the TME, and were significantly more likely to neighbor an immune-active phenotype compared with a suppressive/exhausted phenotype (Fisher exact test P = .047) (top). The mantle/crust regions were moderately further from the TME interface and immunosuppressive TME phenotypes were significantly more likely to neighbor within this district compared with an activate phenotype (Fisher exact P = .027). The tumor core neighborhood represented an immune desert that contained increased CXCR3 expression on CD4 subphenotypes (bottom). The tumor core proportions did not have an association with COO, or any DLBCL subtype, whereas the mantle (I), and dispersed tumor (E) neighborhoods were significantly positively associated with DLBCL subtypes C2 (P = .029), C3 (P = .046), and C4 (P = .025), respectively.