[AI in Cancer Pathology-Present Developments and Future Directions for Treatment Optimization]

Recent advances in artificial intelligence (AI) technologies have substantially expanded the role of pathological image analysis beyond improvements in diagnostic accuracy and efficiency. These developments now enable the integration of histopathological images with non-image data, prediction of treatment response and prognosis, and reduction of the workload for medical professionals. In this review, we provide an overview of representative analytical methods for pathological images, as well as approaches for predicting biomarkers and therapeutic responsiveness directly from histological images. We summarize key studies across major cancer types, pan-cancer investigations, and examples that have been successfully implemented in clinical practice. Additionally, we introduce emerging frameworks such as quantitative analysis of cellular components and tumor microenvironments, pathology foundation models, mRNA expression-based treatment response prediction, integration with spatial transcriptomics data, and applications in clinical trial design. Despite this progress, several challenges remain, such as limited availability of large-scale, high-quality datasets, domain shift across institutions, lack of model interpretability, potential biases, and significant barriers to clinical implementation and regulatory approval. Nevertheless, future developments are expected to enable the simultaneous estimation of multiple biomarkers from a single pathological image, potentially eliminating the need for additional tests. Ultimately, such advances may facilitate rapid, patient-specific drug selection and contribute to more efficient and personalized cancer treatment.