摘要：目前将大语言模型（LLM）用于放射科报告结构化的实践，主要由云端专有闭源模型或需要高端 GPU 部署的大型开源模型所主导，这成为放射科临床工作流应用的关键瓶颈——该场景对数据隐私保护和离线运行有着严苛要求。为解决这一痛点，我们提出了 MedPAO-Assist，一个可在边缘端部署的 Agent 框架，能够直接根据放射科医生的口述录音与语音指令生成结构化放射学报告；得益于异步多线程执行管线，该系统在标准笔记本 CPU 上仅需 $\approx3.3$ GB 峰值 RAM，平均端到端延迟约为 31 秒。该系统目前支持四种主要影像模态——胸部 X 光、胸部 CT、腹部 CT 以及脑部 MRI。据我们所知，这是首批全管线仅基于 0.6B（6 亿）参数量级小语言模型（SLM）运行的 Agent 系统之一。该多 Agent 架构围绕“抽取-锚定-生成（Extract-Ground-Generate）”范式构建，负责执行概念抽取、SNOMED-CT 术语锚定以及结构化报告生成，随后衔接语音驱动的人机协同（human-in-the-loop）校正环节，共同构建出完全无需键盘输入、外部 API 调用或互联网连接的“全免触（no-touch）”报告工作流。我们对管线的各个阶段（概念抽取、本体锚定和概念分类）分别进行了评估，并提供了硬件性能分析结果，验证了该系统在纯 CPU 环境下实现实时临床部署的可行性。

Abstract: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 Brain MRI. To our knowledge, it is among the earliest agentic systems to run its entire pipeline with a Small Language Model (SLM) in the 0.6B-parameter range. This multi-agent architecture, built around an Extract-Ground-Generate paradigm, performs concept extraction, SNOMED-CT grounding, and structured report generation, followed by a speech-driven human-in-the-loop correction phase, together forming a fully "no-touch" reporting workflow that requires no keyboard input, external API calls, or internet connectivity. We evaluate each pipeline stage individually -- concept extraction, ontology grounding, and concept categorization and report hardware profiling results demonstrating the system's feasibility for real-time, CPU-only clinical deployment.