A fuzzy dematel-based delegated proof-of-stake consensus mechanism for medical model fusion on blockchain

Abstract

To ensure consensus regarding the contribution of distributed medical institutions to data models and the transformation of their application value, this paper proposes a fuzzy DEMATEL-based delegated proof-of-stake consensus mechanism for medical model fusion on blockchain. By utilizing transparent, verifiable consensus methods and monitorable on-chain distributed service logic, this framework determines the value-added performance and value-added application of distributed models. Considering that traditional consensus mechanisms are designed primarily for static, deterministic numerical data, they fall short in terms of accommodating consensus for dynamic, interval-based models. To address this limitation, we propose an enhancement to the DPOS consensus mechanism by using fuzzy DEMATEL. This approach enables contribution measurement and consensus for distributed models on the basis of interval-based model characteristics, thereby improving the interpretability of contribution assessments in medical institutions. Since the current lack of application paradigms for data models in distributed environments limits the value conversion of models at the application layer, we propose the construction of a distributed application logic using blockchain and smart contracts. By leveraging smart contracts to protect data privacy and model ownership, this approach enables the standardized and service-oriented transformation of application values. Finally, we conducted an experimental case study using a real medical image diagnostic model to verify and evaluate the feasibility and efficiency of the proposed framework, and a prototype system is established to demonstrate the distributed model consensus and service requirements when collaborating with companies in real-life scenarios. Four sets of experiments were conducted to ensure the feasibility and efficiency of both the distributed consensus and the distributed service process. The results indicate that the proposed consensus mechanism achieves distributed consensus with a latency of approximately 0.2853 s. While the proposed distributed service framework has disadvantages in terms of the throughput and average latency, the differences are minimal—only 0.3937 requests per second and 0.4060 s, respectively, compared with on-chain business creation. Additionally, compared with on-chain business creation, the framework increases CPU and memory utilization by just 15.8902% and 2.4697%, respectively.