The plateau environment poses significant challenges to the physiology and operational capability of military personnel, which is also a key contributor to non-combat casualties. Traditional evaluation methods using single-modal indicators lack dynamic monitoring capabilities and fail to capture complex dynamic stress reactions. This paper reviews the research progress in multimodal data-driven evaluation models for plateau adaptability by analyzing the pathological mechanisms and influencing factors of plateau adaptability and outlining the limitations of current standards and single-modal machine learning. The construction of multimodal data-driven models is explored, focusing on both data architectures that combine physiological, behavioral and environmental modalities for military scenarios and cross-modal alignment and fusion techniques. Furthermore, military applications such as dynamic early warning of non-combat casualties, personalized acclimatization and intelligent military health support are described. Finally, future developments in multimodal data-driven evaluation models for plateau adaptability are predicted in hopes of contributing to the combat effectiveness of troops on the plateau.

Multi-source and multi-modal data in the medical area include structured data, texts, images, and continuous vital sign monitoring data generated by multiple devices of the Internet of Things, known as the Internet of medical things (IoMT) data. The IoMT system integrates a multitude of sensors, medical devices, and intelligent equipment in hospitals, leveraging perceptual and communication technologies, is popularized increasingly. The devices communicate through diverse protocols, and the absence of standardized IoMT interfaces presents a realistic dilemma in integrating IoMT data for holistic clinical analysis. Additionally, the scarcity of computing resources poses a constraint for the extensive training of models and the execution of complex reasoning processes, particularly in high-stakes settings such as intensive care unit (ICU). To address these challenges, we introduce a novel framework designed to facilitate semantic interoperability across multiple devices and to transform multi-source and multi-modal data into a unified data structure. Furthermore, we propose an innovative cloud-edge collaborative system, which could conduct intelligent computing in resource-constrained environments. Our approach was rigorously tested across various metrics, including system response time, data transmission latency, and overall system accuracy. The outcomes demonstrate clear advantages, and offer promising prospects for the future of medical data integration and analysis.