A new multimodal fusion (MMF) system integrating artificial intelligence and clinical data has shown promising results in predicting outcomes for patients with unresectable hepatocellular carcinoma (HCC), the most common form of liver cancer.
Developed using retrospective data from 859 patients across multiple centres, the system combines CT scan-based deep learning features with clinical information to estimate overall survival (OS) and progression-free survival (PFS) in patients receiving immune checkpoint inhibitors (ICIs).
The MMF system significantly outperformed existing benchmarks. It achieved a concordance index (C-index) of 0.74 for OS and 0.69 for PFS in external validation, marking improvements of nearly 30% over traditional radiomics and mRECIST assessment tools. It also surpassed clinical benchmarks and standalone deep learning models.
Importantly, the system maintained strong performance across diverse patient subgroups and demonstrated interpretability, using activation maps and correlations with radiomic features to enhance clinical understanding. Gene expression analysis further revealed enrichment of the PI3K/Akt pathway in patients identified by the model as likely responders, offering insight into underlying biological mechanisms.
Researchers say the MMF system could become a valuable tool for personalising ICI treatment, helping clinicians better identify patients who are most likely to benefit from immunotherapy. By integrating imaging and clinical data in a single predictive framework, the technology represents a step forward in precision oncology for advanced liver cancer.
Reference
Xu J et al. A multimodal fusion system predicting survival benefits of immune checkpoint inhibitors in unresectable hepatocellular carcinoma. NPJ Precis Oncol. 2025;9(1):185
