Data‐sharing strategies in medical consortium based on master‐slave multichain and federated learning

Abstract

In order to encourage participants to actively join the data sharing and to meet the distributed structure and privacy requirement in the medical consortium, the data‐sharing strategy based on the master‐slave multichain is presented in this paper. According to the different computing resources and the responsibility of participants, the adaptive Proof of Liveness and Quality consensus and hierarchical federated learning algorithm for master‐slave multichain are proposed. Meanwhile, by quantifying the utility function and the optimization constraint of participants, this paper designs the cooperative incentive mechanism of medical consortium in multi‐leader Stackelberg game to solve the optimal decision and pricing selection of the master‐slave multichain. The simulation experiments show that the proposed methods can decrease the training loss and improve the parameter accuracy by MedMINST datasets, as well as reach the optimal equilibrium in selection and pricing strategy in the system, guaranteeing the fairness of profit distribution for participants in master‐slave multichain.