Abstract

Secure data management and effective data sharing have become paramount inthe rapidly evolving healthcare landscape, especially with the growingintegration of the Internet of Medical Things (IoMT). The rise of generativeartificial intelligence has further elevated Multi-modal Large Language Models(MLLMs) as essential tools for managing and optimizing healthcare data in IoMT.MLLMs can support multi-modal inputs and generate diverse types of content byleveraging large-scale training on vast amounts of multi-modal data. However,critical challenges persist in developing medical MLLMs, including security andfreshness issues of healthcare data, affecting the output quality of MLLMs. Tothis end, in this paper, we propose a hybrid Retrieval-Augmented Generation(RAG)-empowered medical MLLM framework for healthcare data management. Thisframework leverages a hierarchical cross-chain architecture to facilitatesecure data training. Moreover, it enhances the output quality of MLLMs throughhybrid RAG, which employs multi-modal metrics to filter various unimodal RAGresults and incorporates these retrieval results as additional inputs to MLLMs.Additionally, we employ age of information to indirectly evaluate the datafreshness impact of MLLMs and utilize contract theory to incentivize healthcaredata holders to share their fresh data, mitigating information asymmetry duringdata sharing. Finally, we utilize a generative diffusion model-based deepreinforcement learning algorithm to identify the optimal contract for efficientdata sharing. Numerical results demonstrate the effectiveness of the proposedschemes, which achieve secure and efficient healthcare data management.