From Pixels to Stroma: AI-Driven Spatial Profiling of Cancer-Associated Fibroblasts on H&E and Its Implications for Immunotherapy

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Immune checkpoint blockade (ICB) has transformed cancer therapy, but clinical responses remain heterogeneous across tumor types and patient populations. Cancer-associated fibroblasts (CAFs) are key stromal components of the tumor microenvironment and can contribute to immunotherapy resistance through immune exclusion, extracellular matrix remodeling, chemokine signaling, and interactions with suppressive immune cells. Although CAF-directed strategies are under active investigation, their clinical translation is limited by marked CAF heterogeneity and the lack of scalable biomarkers for patient stratification. Computational pathology based on hematoxylin and eosin (H&E) whole-slide images (WSIs) provides a potential approach for extracting stromal and spatial features from routine histology, although digitized WSIs and the infrastructure required for large-scale AI analysis are not universally available. In this review, we synthesize current evidence on CAF classification, CAF-mediated immunotherapy resistance, H&E-based computational pathology, and emerging histology-based biomarker models. We further propose a conceptual roadmap for developing CAF-aware H&E spatial signatures with potential relevance to future immunotherapy stratification. Current evidence supports the biological rationale and computational feasibility of this approach, whereas its clinical utility remains to be established through rigorous external validation and prospective clinical evaluation.