Integração da patologia digital: da digitalização de lâminas histológicas à genômica de alta precisão

Abstract

The integration of digital pathology, Artificial Intelligence (AI), and genomic analysis represents a new paradigm in modern anatomic pathology, particularly in resource-limited settings. This study aimed to implement and validate a digital pathology workflow—from whole slide imaging (WSI) of histological and cytological specimens to next-generation sequencing (NGS)—with a focus on non-small cell lung cancer (NSCLC). Two complementary axes were explored: (1) implementation of digital pathology in a reference laboratory in Northeastern Brazil, including diagnostic validation of 384 HE and 99 cervical cytology slides, which demonstrated high diagnostic concordance (κ = 0.96); an(2) application of AI-based algorithms for automated quantification of cellular and tissue composition in 47 lung cancer cases and correlation of histological features, molecular quality (DNA/RNA), variant allele frequency (VAF), and genomic data obtained via NGS. A significant correlation was observed between DNA concentration and the number of tumor cells (ρ = 0.59; p< 0.001), as well as between fibroblast density and DNA concentration (ρ = 0.57; p< 0.001). VAF was positively associated with DNA yield (ρ = 0.41; p = 0.020), but not with specific cellular composition. Analysis of the tumor microenvironment revealed higher lymphocytic and granulocytic infiltration in tumors with elevated PD-L1 expression (TPS ≥ 50%), suggesting distinct immunological profiles. These findings validate the use of AI for objective tumor cellularity estimation and underscore the importance of histological selection to optimize yield and accuracy in molecular testing. This integrated approach—feasible even in low-resource settings—may contribute to the broader implementation of precision oncology in developing countries.