跳转至

文章背景与核心概要

卒中（中风）具有极高的致残率和致死率，院前的快速准确评估对于争取黄金救治时间至关重要。尽管基于FAST（面部、手臂、言语、时间）的临床筛查标准已被广泛采用，但在家庭或社区等非临床场景中，普通用户往往难以准确描述症状并遵循复杂的评估流程。

为了解决这一痛点，本文介绍了 StrokeGuard——一个旨在规范并提高院前卒中评估准确性的创新型多智能体系统。该系统通过引入“双通道智能体机制”，将正式的临床症状评估与流程支持（如实时反馈和纠错）进行解耦。借助多智能体协作、状态机控制以及阶段局部回退恢复机制，StrokeGuard 具备极高的流程容错能力。基于 MATES-9 量表的实际用户评估表明，与传统的纸质表单相比，该系统使用户体验得分实现了 23.8% 的相对提升。

StrokeGuard: A Multi-Agent Guided System for Prehospital Stroke Assessment

Authors: Wentao Yang, Zhenye Xu, Ruoyi Li, Musen Zhang, Yao Guo
Date: August 25, 2026
Subject: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
Cite as: arXiv:2608.24555

Summary

StrokeGuard 是一个创新的多智能体系统，旨在规范和提高院前卒中评估的准确性。虽然基于临床的 FAST（面部、手臂、言语、时间）筛查是标准做法，但在家庭或社区环境中的非临床用户往往难以提供准确的症状描述，也难以应对复杂的程序要求。

StrokeGuard is an innovative multi-agent system designed to standardize and improve the accuracy of prehospital stroke assessments. While clinical FAST-based (Face, Arm, Speech, Time) screenings are standard, non-clinical users in home or community settings often struggle with inaccurate symptom descriptions and complex procedural requirements.

StrokeGuard 通过利用双通道智能体机制（将正式的临床评估与程序支持（如实时反馈和纠错）分离开来）解决了这些空白。通过采用多智能体协作、状态机控制和阶段局部回退恢复，该系统确保了极高的程序容错能力。使用 MATES-9 量表的用户评估表明，StrokeGuard 显着优于传统的纸质表单，使用户体验得分实现了 23.8% 的相对增长。

StrokeGuard addresses these gaps by utilizing a dual-channel agent mechanism that separates formal clinical assessment from procedural support (such as real-time feedback and error correction). By employing multi-agent collaboration, state-machine control, and stage-local fallback recovery, the system ensures high procedural fault tolerance. User evaluations using the MATES-9 scale demonstrated that StrokeGuard significantly outperforms traditional paper-based forms, achieving a 23.8% relative increase in user experience scores.

Key Features

  • 双通道交互： 将症状评估（面瘫、手臂无力、言语障碍）与程序指导（分步提示和纠错）解耦。 > Dual-Channel Interaction: Decouples the assessment of symptoms (facial palsy, arm weakness, speech impairment) from the procedural guidance (step-by-step prompts and error correction).
  • 智能控制： 利用状态机控制和阶段局部回退机制，即使在用户遇到困难时也能保持连续性。 > Intelligent Control: Utilizes state-machine control and stage-local fallback mechanisms to maintain continuity even when users encounter difficulties.
  • 自动报告： 通过受限的预训练视频评估模块，将特定阶段的评分与结构化报告生成相结合。 > Automated Reporting: Integrates stage-specific scoring via constrained pretrained video assessment modules with structured report generation.
  • 经过验证的功效： 在模拟院前场景中，与标准方法相比，该系统在 MATES-9 量表上的用户表现和体验提高了 10.83 分。 > Proven Efficacy: In simulated prehospital scenarios, the system improved user performance and experience by 10.83 points on the MATES-9 scale compared to standard methods.

Accessing the Paper

Metadata & Citations