Advances in Artificial Intelligence-Based Liver-Related Semantic Segmentation Techniques and Applications Using CT Imaging.

[BACKGROUND AND AIMS] Artificial intelligence (AI)-assisted semantic segmentation of liver computed tomography (CT) images has important clinical value in disease assessment, surgical planning, treatment evaluation, and longitudinal monitoring. This review aims to summarize the current clinical applications and recent technical advances in AI-based liver-related semantic segmentation on CT.

[METHODS] This narrative review synthesizes recent studies on liver organ, tumor, and vascular segmentation, focusing on both clinical applications across different hepatic diseases and technical developments in model architectures, data processing, information fusion, and strategies for improving robustness and generalizability.

[RESULTS] AI-based segmentation models, particularly those built on U-Net and hybrid attention/Transformer frameworks, have enabled automated analysis for clinically relevant tasks such as future liver remnant estimation, graft volumetry, chronic liver disease assessment, tumor burden evaluation, prognosis prediction, and vascular-intervention planning. Despite strong performance in liver organ segmentation, challenges remain in small tumor and fine vascular segmentation, as well as in external validation, deployability, and workflow integration.

[CONCLUSIONS] AI-based liver CT segmentation shows strong potential to support precision hepatobiliary imaging and to improve existing clinical workflows. Further progress will depend on improving robustness, reducing computational burden, enhancing performance in fine-structure segmentation, and facilitating real-world clinical deployment.