📄 中文摘要
临床影像解读是一个多步骤且以工具为中心的过程，临床医生需要迭代地结合视觉证据与患者背景，量化发现，并通过一系列专业程序来完善决策。尽管基于大型语言模型的代理有望协调这些异构医疗工具，但现有系统在部署后将工具集和调用策略视为静态。这种设计在现实世界中的领域变化、任务间的转变以及不断演变的诊断需求下显得脆弱，预定义的工具链往往会降级并需要昂贵的手动重新设计。提出的MACRO是一种自我进化的、经验增强的医疗代理，旨在从静态工具组合转向经验驱动的工具发现。
通过经验驱动的自我技能发现进化医疗影像代理
📄 English Summary
Evolving Medical Imaging Agents via Experience-driven Self-skill Discovery
Clinical image interpretation is a multi-step, tool-centric process where clinicians iteratively combine visual evidence with patient context, quantify findings, and refine decisions through specialized procedures. While LLM-based agents have the potential to orchestrate heterogeneous medical tools, existing systems treat tool sets and invocation strategies as static post-deployment. This design is fragile under real-world domain shifts, task variations, and evolving diagnostic needs, often leading to degraded predefined tool chains that require costly manual redesign. The proposed MACRO is a self-evolving, experience-augmented medical agent that transitions from static tool composition to experience-driven tool discovery.