NIAgent: Multi-Agent AI System for Autonomous Neuroimaging Analysis
A new multi-agent system named NIAgent has been developed for fully autonomous neuroimaging analysis. In contrast to traditional flat tool-calling agents, NIAgent employs a code-centric execution approach, allowing specialized agents to work together in creating and refining executable workflows. The objective of this system is to connect standardized yet rigid workflows, such as fMRIPrep, with the flexible reasoning capabilities of human researchers. This facilitates closed-loop adaptation, which can automatically adjust parameters and address pipeline failures. Such advancements could greatly enhance the development of clinical biomarkers by minimizing the need for manual trial-and-error processes.
Key facts
- NIAgent is a multi-agent system for autonomous neuroimaging analysis.
- It uses a code-centric execution paradigm with specialist agents.
- It addresses limitations of static workflows like fMRIPrep.
- The system enables closed-loop adaptation and automated troubleshooting.
- It aims to scale clinical biomarker development.
- The approach reduces manual trial-and-error for domain experts.
- The system was introduced in a paper on arXiv (2605.09366).
- The paper is categorized as a new type of announcement.
Entities
Institutions
- arXiv