RadAgent: A tool-using AI agent for stepwise interpretation of chest computed tomography

TL;DR

RadAgent is an interpretable AI tool that generates chest CT reports through stepwise reasoning, providing transparent decision traces and improving accuracy, robustness, and faithfulness over previous models.

Contribution

It introduces RadAgent, a novel tool-using AI agent that produces interpretable, step-by-step CT reports with decision traces, enhancing transparency and reliability in medical imaging analysis.

Findings

RadAgent improves macro-F1 by 6.0 points and micro-F1 by 5.4 points over CT-Chat.

RadAgent's robustness under adversarial conditions increases by 24.7 points.

RadAgent achieves 37.0% faithfulness, a new capability in this domain.

Abstract

Vision-language models (VLM) have markedly advanced AI-driven interpretation and reporting of complex medical imaging, such as computed tomography (CT). Yet, existing methods largely relegate clinicians to passive observers of final outputs, offering no interpretable reasoning trace for them to inspect, validate, or refine. To address this, we introduce RadAgent, a tool-using AI agent that generates CT reports through a stepwise and interpretable process. Each resulting report is accompanied by a fully inspectable trace of intermediate decisions and tool interactions, allowing clinicians to examine how the reported findings are derived. In our experiments, we observe that RadAgent improves Chest CT report generation over its 3D VLM counterpart, CT-Chat, across three dimensions. Clinical accuracy improves by 6.0 points (36.4% relative) in macro-F1 and 5.4 points (19.6% relative) in micro-F1. Robustness under adversarial conditions improves by 24.7 points (41.9% relative). Furthermore, RadAgent achieves 37.0% in faithfulness, a new capability entirely absent in its 3D VLM counterpart. By structuring the interpretation of chest CT as an explicit, tool-augmented and iterative reasoning trace, RadAgent brings us closer toward transparent and reliable AI for radiology.