[论文] MedPAO-Assist: Edge-Deployable, Modality-Aware Agentic AI for Radiolog...

论文概要

研究领域: ML 作者: Shrish Shrinath Vaidya, Minmini Selvam, Ganapathy Krishnamurthi 发布时间: 2026-10-22 arXiv: 2610.26913

中文摘要

大语言模型在放射学报告结构化方面的应用主要被专有云端模型或需要高端GPU部署的大型开源模型所主导，这对于要求数据隐私和离线操作的放射科工作流来说是临床采用的关键瓶颈。为此，我们提出MedPAO-Assist，一个可边缘部署的智能体框架，可直接从放射科医生的口述和语音指令生成结构化放射学报告，仅需约3.3 GB峰值内存，并在标准笔记本CPU上通过异步多线程执行管线实现约31秒的平均端到端延迟。该系统目前支持胸部X光、胸部CT、腹部CT和脑部MRI四种主要影像模态。据我们所知，它是最早完整管线运行在0.6B参数级小型语言模型（SLM）上的智能体系统之一。这一多智能体架构围绕『提取-锚定-生成』范式构建，执行概念提取、SNOMED-CT本体锚定和结构化报告生成，随后进入语音驱动的人机协作校正阶段，共同构成完全『无触碰』的报告工作流，无需键盘输入、外部API调用或互联网连接。我们对管线各阶段单独评估了概念提取、本体锚定和概念分类，并报告了硬件性能分析结果，证明了该系统在实时、纯CPU临床部署中的可行性。

原文摘要

The adoption of Large Language Models for radiology report structuring has been largely dominated by either proprietary cloud-based models or large open models requiring high-end GPUs for deployment, which is a critical bottleneck for clinical adoption in radiology workflows that demand data privacy and offline operation. To address this, we present MedPAO-Assist, an edge-deployable agentic framework that generates structured radiology reports directly from radiologist dictation and voice commands, requiring only \(\approx3.3\) GB peak RAM and achieving an average end-to-end latency of \(\approx31\) seconds on a standard laptop CPU through an asynchronous, multi-threaded execution pipeline. The system currently supports four major imaging modalities -- Chest X-ray, Chest CT, Abdomen CT, and Br...

#论文 #arXiv #ML #小凯