The growing ability of large language models and vision language models to jointly interpret and reason over images and text is reshaping medical agents from task-specific predictors to autonomous systems capable of perceiving, reasoning, planning, remembering, and acting in clinical environments. This work departs from the capability-first perspective of existing literature and focuses on clinical deployment, exploring the tasks, contamination-resistant benchmarks, and interactive training environments required for medical agents to be trusted in practice.
Medical agents are formalized as sequential decision-making systems under partial observability, accompanied by a three-level autonomy taxonomy: assisted, cooperative, and fully autonomous operation. The field is organized along a unified scaling spine that includes framework scaling, capability scaling, and environment scaling. Within this framework, clinical environment scaling, which integrates tools, data, and clinical gyms, is identified as the most actionable yet underexplored direction for agents operating in PACS, EHR, and FHIR ecosystems.
Clinical self-evolution, where agents improve through interaction with their environments rather than just parameter scaling, is positioned as a key research frontier, drawing insights from self-improving agents, agent gyms, and test-time compute scaling. Applications across radiology, pathology, ophthalmology, and hospital workflows are examined, alongside deployment challenges such as hallucination, cascading failures, and fairness. By consolidating over 300 references, with particular emphasis on advances from 2025 to 2026, this work provides a roadmap towards trustworthy, self-improving medical imaging systems for real clinical practice.
Blogger's Review: This research offers a fresh perspective on the autonomy of medical agents, outlining the necessary conditions for achieving self-evolution in real clinical environments and emphasizing the importance of environmental scaling. This direction will drive innovation in medical technology, making agents more reliable and effective in practical applications.