GraphLSurv: A scalable survival prediction network with adaptive and sparse structure learning for histopathological whole-slide images

Pei Liu, Luping Ji, Feng Ye, Bo Fu

Abstract: Highlights•A scalable graph-based network is developed to predict survival outcomes for WSIs.•Adaptive and sparse structures can be learned to better capture patch correlations.•A new GCN-HMP module is proposed to aggregate the patch features in hybrid graphs.•Our network could perform better than previous models on two large WSI datasets.•Adaptive and sparse patch structures could be more suitable for characterizing WSIs.