Chien Dinh Huynh
Artificial intelligence (AI) is reshaping pathology from a predominantly descriptive discipline into a quantitative and predictive science. Advances in digital pathology and machine learning enable extraction of clinically relevant information from routine histopathologic images, including diagnostic features, prognostic markers, molecular correlates, and indicators of therapeutic response. These developments position AI pathology as a central component of precision medicine. This review synthesizes recent advances in AI pathology, focusing on methodological frameworks, data requirements, and clinical applications. We outline model development using whole-slide images (WSIs), including weakly supervised and self-supervised approaches, and discuss integration with multimodal data such as genomics, transcriptomics, imaging, and clinical variables. Key barriers to clinical translation—including dataset bias, external validation, interpretability, regulatory considerations, and workflow integration—are critically examined. Emerging directions include multimodal AI systems, spatially resolved analysis, and digital twin models that link tissue morphology with molecular and functional data. Progress in this field will depend on a shift from technology-driven development to evidence-based implementation. With rigorous validation and responsible deployment, AI pathology has the potential to improve diagnostic accuracy, enhance clinical decision-making, and redefine the role of pathology in predictive medicine. • AI is transforming pathology into a quantitative, predictive, and integrative discipline. • Deep learning enables extraction of diagnostic, prognostic, and molecular insights from routine histopathology images. • AI pathology supports precision medicine across oncology, immunology, and infectious diseases. • Emerging methods include weakly supervised, self-supervised, and multimodal learning approaches. • Integration of pathology with genomic, clinical, and imaging data enhances predictive modeling. • Key challenges include dataset bias, validation, interpretability, and regulatory integration. • Future directions involve multimodal AI, spatial analysis, and digital twin frameworks. • Evidence-based implementation is essential for safe and effective clinical translation.