SURGENT: A Surgical Multi-Agent Assistance System Across the Perioperative Workflow

SURGENT is a multi-agent AI system designed to assist surgical teams throughout the perioperative workflow by combining large language models with specialized reasoning, memory management, and clinical knowledge retrieval. The system addresses critical limitations of standard LLMs—including token constraints and poor context retention—and demonstrates superior performance across five surgical tasks compared to existing medical AI frameworks.

SURGENT represents a significant advancement in applying AI to high-stakes medical environments where accuracy, traceability, and comprehensive context are non-negotiable. The system addresses a genuine technical gap: while general-purpose LLMs excel at language tasks, they struggle with the demands of surgical care, which requires synthesizing extensive patient histories, maintaining consistent reasoning across complex workflows, and providing auditable decision-making that clinicians can trust. The architecture's innovation lies in its dual-memory design, separating long-term patient records from short-term working summaries, enabling the system to maintain coherence across the entire perioperative journey—from preoperative planning through post-operative rehabilitation.

This development reflects a broader maturation trend in medical AI where researchers move beyond simple chatbots toward domain-specific systems engineered for clinical workflows. The emphasis on local deployment using DeepSeek as a backbone model addresses growing concerns about data privacy and regulatory compliance in healthcare, where patient information cannot be transmitted to centralized cloud services. This approach aligns with emerging healthcare IT requirements around data sovereignty.

For the healthcare and medical AI sectors, SURGENT signals that specialized surgical AI systems can outperform general-purpose models when properly architected. The five-task evaluation framework—case analysis, surgical planning, safety monitoring, risk assessment, and rehabilitation—provides a replicable validation model for other medical specialties. Hospitals and surgical centers will likely view such systems as infrastructure investments rather than experimental tools.

The next critical phase involves real-world clinical validation and regulatory pathways. Developers should monitor adoption barriers including hospital IT integration, liability frameworks, and physician acceptance rates.

• →SURGENT combines tree-of-thought planning with multi-agent collaboration to overcome LLM limitations in surgical decision support.
• →The system's dual-memory architecture manages both long-term patient histories and short-term working contexts for coherent perioperative reasoning.
• →Local deployment using DeepSeek enables privacy-preserving surgical AI without cloud-based data transmission.
• →Experimental results across five surgical tasks show SURGENT outperforms baseline LLMs and existing medical multi-agent systems.
• →The framework provides an auditable, traceable reasoning process critical for clinical adoption and liability management.