Abstract The integration of artificial intelligence (AI) with digital pathology provides unprecedented opportunities to decode the spatial complexity of tumor-immune interactions. We developed an AI-driven spatial pathology framework that combines multiplex fluorescence in situ hybridization (mFISH) and deep-learning-based digital histopathology to classify cells and identify tumor-infiltrating lymphocyte (TIL) patches in bladder tumors. Formalin-fixed paraffin-embedded samples were analyzed using RNAscope mFISH targeting three arsenic-responsive genes - NKIRAS2, AKTIP, and HLA-DQA1. Whole-slide images were co-registered with H92% accuracy in distinguishing tumor, stromal, and lymphocytic populations. Quantitative mapping revealed that low-grade tumors contained densely packed TIL clusters with a mean inter-TIL distance of 30 μm, reflecting robust immune infiltration. In contrast, high-grade tumors exhibited a 3-fold reduction in TIL patch density and increased spatial separation between lymphocytes and tumor boundaries, consistent with immune exclusion. Spatial co-expression of NKIRAS2-AKTIP was enriched in immune-cold regions, while HLA-DQA1 expression correlated with higher TIL density and active immune interfaces. Pathway enrichment analysis of TIL-proximal regions highlighted upregulation of inflammatory and oxidative stress pathways, whereas immune-depleted zones showed activation of DNA repair and cell proliferation signatures. Gradient-based quantification of TIL patch density and co-expression topology stratified tumors into immune-inflamed, immune-intermediate, and immune-cold phenotypes that correlated with histologic grade and exposure signatures. This integrative approach bridges molecular multiplex imaging with spatial computational pathology, establishing a scalable framework for digital immune landscape mapping. Our findings highlight how environmental exposures reshape tumor-immune architecture and demonstrate the potential of AI-guided pathology for precision immunoprofiling and exposure-related cancer risk assessment. Citation Format: Sonalika Singhal, Anushka Dikshit, Samarth Singhal, Kevin L. Gardner. AI-driven spatial cell classification and tumor-infiltrating lymphocyte (TIL) patch analysis reveal spatial immunoarchitecture in arsenic-associated bladder cancer abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4001.
Singhal et al. (Fri,) studied this question.