Graph Attention-based Fusion of Pathology Images and Gene Expression for Prediction of Cancer Survival
We propose a simple yet scalable framework for integrating histopathology images and genomic data through graph attention-based fusion. Histological images are modeled using attention-based graph neural networks, enabling the extraction of spatially and morphologically relevant features. The architecture is flexible and can accommodate the integration of multiple genomic modalities with histopathological information, supporting enhanced diagnostic accuracy, prognostic stratification, and therapeutic response prediction.