Abstract
This study proposes MRI-CARDAIX, a novel multimodal multi-agent framework for integrated analysis of cine and Late Gadolinium Enhancement (LGE) MRI. MRI-CARDAIX employs a supervisor agent that coordinates mo-dality-specific cine MRI, LGE MRI, and evidence-retrieval agents for integrated cardiac assessment. All models achieved impressive segmentation accuracy for left ventricular delineation (median DSC ≥0.98), while nnU-Net v2 and U-Mamba demonstrated the most consistent performance across cardiac structures, with nnU-Net v2 achieving the highest median DSC for right ventricular segmen-tation (0.9517 [0.9169-0.9717]) and comparable myocardial segmentation per-formance (median DSC 0.957). For LGE MRI analysis, fibrosis localization achieved an F1-score of 92.98% using Swin-U-Mamba, with precision of 93.01% and recall of 92.76%. Quantitative myocardial fibrosis assessment demonstrated relative volume errors ranging from 5.56% to 11.13% across basal, mid-cavity, and apical regions. Clinical validation of AI-generated cardiac MRI reports using the proposed Total Clinical Reliability Score framework resulted in a mean score of 7.3/10, corresponding to a good level of clinical reliability. The supervisor agent achieved 92% routing accuracy (macro-F1 94%) with a mean la-tency of 1.8 s. Retrieval-augmented answers scored 0.88 faithfulness and 0.84 answer relevancy. MRI-CARDAIX provides an interpretable framework for in-tegrated cardiac MRI analysis and reporting. Source code will be made publicly available.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MedAgent_017.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=fYuXzPlvkT
BibTex
@InProceedings{AlWal_Towards_MICCAISAT2026,
author = { Al-Haidri, Walid AND Raza, Mukhlis AND Levchuk, Anatoliy AND Belousova, Kseniya AND Lukin, Maksim AND Gu, Yeong Hyeon AND Al-antari, Mugahed A. AND Brui, Ekaterina},
title = { { Towards Automated Cardiac MRI Assessment: A MultiAgent CAD Framework for Functional Analysis and Tissue Characterization } },
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}
}
