EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

MedAgentOps: A Distributed Multi-Agent Collaborative Reasoning Infrastructure Based on Memory-Augmented Graph Reasoning Network for Emergency Department Flow Forecasting and Resource Optimization

Hao Hu, Yan Zeng, Weijia Li

Fundamental Scientific Reports in Multidisciplinary Areas · 25.06.2026 · 2 (02) · 146-158

DOI: 10.66238/fsrma101 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Fundamental Scientific Reports in Multidisciplinary Areas, online erschienen am 25.06.2026. Verfasst von Hao Hu und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Emergency departments (EDs) in modern hospitals face persistent challenges including unpredictable patient influx, resource scarcity, and inefficient operational coordination, which significantly degrade service quality and response efficiency. To address these issues, this study proposes MedAgentOps, a distributed multi-agent collaborative reasoning infrastructure built upon a Memory-Augmented Graph Reasoning Network (MAGR-Net) for emergency flow forecasting and resource optimization. The proposed framework integrates four key components: Dynamic Temporal Graph Fusion (DTGF) for modeling heterogeneous temporal dependencies in emergency-related factors, a Shared Medical Memory (SMM) module for consolidating short-term operational states and long-term historical knowledge, an Agent Graph Reasoning Module (AGRM) for adaptive coordination among forecasting, scheduling, and resource allocation agents, and a Multi-Objective Proximal Policy Optimization (MOPPO) strategy for optimizing operational efficiency under multiple constraints. These modules jointly enable end-to-end reasoning from patient arrival prediction to real-time resource scheduling in a unified multi-agent system. Experimental results on a real-world emergency department dataset show that the proposed framework achieves a Mean Absolute Percentage Error (MAPE) of 8.73%, outperforming all baseline models significantly. In addition, MedAgentOps reduces average patient waiting time by approximately 17.9% compared with the strongest baseline (GNN-RL) while improving resource utilization to 91.3%. These results confirm the effectiveness of integrating multi-agent reasoning with memory-augmented graph learning for intelligent healthcare optimization.

Zitieren

APA

Hu, H., Zeng, Y., & Li, W. (2026). MedAgentOps: A Distributed Multi-Agent Collaborative Reasoning Infrastructure Based on Memory-Augmented Graph Reasoning Network for Emergency Department Flow Forecasting and Resource Optimization. Fundamental Scientific Reports in Multidisciplinary Areas, 2(02), 146-158. https://doi.org/10.66238/fsrma101

Vancouver

Hu H, Zeng Y, Li W. MedAgentOps: A Distributed Multi-Agent Collaborative Reasoning Infrastructure Based on Memory-Augmented Graph Reasoning Network for Emergency Department Flow Forecasting and Resource Optimization. Fundamental Scientific Reports in Multidisciplinary Areas. 2026;2(02):146-158. doi:10.66238/fsrma101

BibTeX
@article{Hu2026,
  author = {Hu, Hao and Zeng, Yan and Li, Weijia},
  title = {MedAgentOps: A Distributed Multi-Agent Collaborative Reasoning Infrastructure Based on Memory-Augmented Graph Reasoning Network for Emergency Department Flow Forecasting and Resource Optimization},
  journal = {Fundamental Scientific Reports in Multidisciplinary Areas},
  year = {2026},
  volume = {2},
  number = {02},
  pages = {146--158},
  doi = {10.66238/fsrma101}
}
Kostenlos starten →