In recent years, significant advances have been achieved in liver cancer management with the development of artificial intelligence (AI). AI-based pathological analysis can extract crucial information from whole slide images to assist clinicians in all aspects from diagnosis to prognosis and molecular profiling. However, AI techniques have a “black box” nature, which means that interpretability is of utmost importance because it is key to ensuring the reliability of the methods and building trust among clinicians for actual clinical implementation. In this paper, we provide an overview of current technical advancements in the AI-based pathological analysis of liver cancer, and delve into the strategies used in recent studies to unravel the “black box” of AI's decision-making process.

Advances in antibody engineering have led to the generation of more innovative antibody drugs, such as bispecific antibodies (bsAbs). Following the success associated with blinatumomab, bsAbs have attracted enormous interest in the field of cancer immunotherapy. By specifically targeting two different antigens, bsAbs reduce the distance between tumor and immune cells, thereby enhancing tumor killing directly. There are several mechanisms of action upon which bsAbs have been exploited. Accumulating experience on checkpoint-based therapy has promoted the clinical transformation of bsAbs targeting immunomodulatory checkpoints. Cadonilimab (PD-1 × CTLA-4) is the first approved bsAb targeting dual inhibitory checkpoints, which confirms the feasibility of bsAbs in immunotherapy. In this review we analyzed the mechanisms by which bsAbs targeting immunomodulatory checkpoints and their emerging applications in cancer immunotherapy.