A Heterogeneous Bipartite Graph Framework for Donor-Recipient Matching in Kidney Transplantation

This paper presents a novel graph-based framework for donor-recipient matching for organ transplantation, specifically targeting kidney transplantation. We pursue donor-recipient matching as a feature set matching problem using a graph-based approach. To represent kidney transplant clinical and operational considerations, we developed a kidney transplant graph by embedding kidney transplantation knowledge and relevant datasets to capture the complex relationships between the heterogeneous donor and recipient feature sets to provide a system for optimal donor and recipient matches. Our approach includes multi-dimensional graph edge types representing clinical considerations to provide comprehensive and optimal pairs. We propose the kidney transplant graph with a graph neural network to optimize donor-recipient compatibility predictions, enhancing decision-making in transplantation.