Abstract
The complexity of glioma identification and characterization from multi-sequence MRI poses a significant burden on clinical radiology workflows. While unimodal AI systems have shown promise, they often struggle with the nuanced, multi-step reasoning required for comprehensive clinical reporting. We propose and validate a novel multi-agent architecture designed to automate the end-to-end generation of radiology reports for glioma patients. In this retrospective study, a multi-agent framework was developed comprising several specialized autonomous agents that cover several sub-tasks such as, multi-sequence MRI alignment; abnormality segmentation, diagnosis and grading of Glioma, report editing, revising and quality assurance. The whole process is coordinated with an experienced orchestrator with expert human-in-the-loop. The system was validated using a real-world dataset of 63 adult glioma patients. Performance was measured against radiologist-verified reference reports using an LLM-as-a-Judge protocol that scores five clinical criteria on a 1–10 scale. The multi-agent system achieved high diagnostic fidelity. The deployment of a multi-agent architecture significantly enhances the accuracy and clinical utility of automated radiology reporting in glioma. By decomposing complex interpretive tasks into specialized, collaborative units, this framework provides a robust pathway for integrating generative AI into high-stakes clinical oncology environments.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MedAgent_007.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=MhxUIsL6ci
BibTex
@InProceedings{ElAhm_AMultiAgent_MICCAISAT2026,
author = { El Boardy, Ahmed Tamer AND Elshaer, Ziad AND Ahmed, Sameh AND Elshehabi, Ibrahim AND Raslan, Ahmed Salah Elsayed AND Mabrok, Mohamed AND Khoriba, Ghada AND Rashed, Essam A.},
title = { { A Multi-Agent Framework for Automated MRI Reporting in Glioma: A Real-World Clinical Validation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17263},
month = {pending},
page = {pending}
}