Related Experiment Video
Updated: May 6, 2026
Assessment of Dendritic Arborization in the Dentate Gyrus of the Hippocampal Region in Mice
Published on: March 31, 2015
HIPPA-DGCN: Survival analysis on whole slide images via hyper-image patches and position-aware dense graph
Huhan Xie1, Tong Xu1, Huaishui Yang2
1College of Health Science and Environmental Engineering, Shenzhen Technology University, Shenzhen 518118, China; School of Applied Technology, Shenzhen University, Shenzhen 518060, China.
Background And Objective:
Whole Slide Images (WSIs) are the gold standard for cancer diagnosis and prognosis. However, the enormous scale of WSIs presents significant challenges for effective information aggregation in survival analysis, limiting prognostic accuracy and clinical applicability. This study aims to develop a computationally efficient framework for accurate and interpretable survival prediction from WSIs.
Methods:
We propose HIPPA-DGCN, a novel framework that utilizes hyper-image patch clustering for feature distillation and a position-aware dense graph convolutional network for global context modeling. This architecture is designed to efficiently integrate pathological features with their spatial relationships.
Results:
Extensive evaluation on seven public datasets (five from TCGA and two from CPTAC) demonstrates state-of-the-art performance. Our method achieves superior prognostic accuracy while drastically improving computational efficiency, reducing model parameters to 0.73M and computational cost to 0.08 GFLOPs per WSI. The model generates interpretable attention maps that highlight histopathological regions with significant prognostic relevance.
Conclusions:
HIPPA-DGCN provides an accurate, efficient, and interpretable solution for WSI-based survival analysis. Its lightweight architecture and robust performance make it particularly suitable for clinical deployment in resource-constrained environments, potentially enhancing cancer prognosis workflows.
