Utilization of Whole Slide Pathological Images and Tumor Clinical Data: A Hybrid Graph Convolutional Network Model for Hepatocellular Carcinoma Prognosis Prediction

Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide, ranking fourth in cancer-related mortality. Prognostic risk prediction for HCC patients can optimize treatment strategies, assess therapeutic efficacy, and ultimately improve post-operative survival rates. Pathological slides are considered the gold standard for cancer diagnosis and prognosis, playing a crucial role in prognostic risk stratification. However, in addition to pathological slides, clinical text information also contains significant prognostic value that cannot be ignored.To address this, we propose a text-driven multimodal fusion prognostic GCN. The core of the model is to integrate imaging and textual features by constructing a patient graph driven by text, overcoming the significant disparity between the two modalities. Validation results on the FAH-ZJUMS dataset show that our method improves performance by 21% compared to imaging-only prognostic models. These results demonstrate the great potential of this multimodal fusion approach in handling complex medical information.