World Scientific Annual Review of Cancer Immunology

The application of three-dimensional (3D) spheroids or organoids in drug testing and precision medicine has led to significant advancements in how cancer and other diseases are treated. Not only can these 3D structures mimic the architecture and structure of tumors, but organoids formed from primary patient samples are able to recapitulate the molecular and functional characteristics of the original patient tumors. These clinically and physiologically relevant organoids can therefore be used to address questions related to drug efficacy and resistance and can even be used to predict patient-specific drug responses. However, despite such evident advantages, the lack of a patient-specific tumor microenvironment (TME), or even the basic TME that includes sufficient immune cells and other cell types, limits the potential of these organoids in immunotherapy drug testing. As such, co-culture models of patient-derived organoids with immune cells have since been developed to explore cancer-immune interactions, monotherapy or combinatorial immunotherapy drug testing, and variable patient drug responses. Moreover, when coupled with artificial intelligence-driven platforms, these organoid models can be more efficiently utilized to identify better therapeutic options and improve health outcomes through precision medicine. This review aims to highlight the use of organoids and the broader implementations of such organoid models in functional precision medicine, particularly in the context of immunotherapy.

In this paper, we explore the transformative role of artificial intelligence (AI) in the analysis and transformation of biomedical images in the field of cancer immunotherapy. This review highlights the potential of AI to be deployed in the clinic as part of value-added healthcare, streamlining image analysis, facilitating the prediction of immunotherapy response, and integrating multi-modal data to offer a holistic view of patient conditions. We discuss the challenges facing clinical implementation, including data privacy, model interpretability, and the need for standardized reporting, which must be addressed to ensure patient safety.

Recent advances in clinical and research pathology are fueled by new histomorphological and molecular techniques, moving beyond traditional methods like Hematoxylin and Eosin (H&E) and immunohistochemical (IHC), which offer limited information and have high observer variability. Advanced methods like spatial omics provide detailed molecular mapping but are costly and complex. Artificial intelligence (AI) has emerged as a promising solution to these challenges by digitally predicting cellular and molecular biomarkers from H&E-stained images. This review highlights the development of a new field, H&E 2.0, which focuses on current and emerging AI advancements leveraging H&E stained images. We discuss the current capabilities of H&E 2.0 in predicting bulk biomarker quantity at the sample level, generating virtual staining at cellular resolution, and inferring in situ gene expression information from H&E slides across various tumor types. Additionally, we offer insights into the current challenges and potential future directions in this evolving field.