Abstract
Accurate brain tumor segmentation from multi-center multimodal magnetic resonance imaging (MRI) remains challenging due to variations in imaging protocols, scanners, and modality distributions across institutions. Although deep learning approaches have achieved promising performance, effectively exploiting complementary information from heterogeneous MRI modalities and maintaining reliable feature propagation under distribution variations remain challenging. To address these issues, we propose MCS-UNet, a multimodal collaborative framework for robust multi-center brain tumor segmentation. The proposed framework integrates adaptive modality collaboration, hierarchical context modeling, and region-aware feature refinement into a unified encoder-decoder architecture. The Modality Collaboration Module (MCM) learns modality-specific spatial responses and feature-level modality contributions to achieve adaptive multimodal fusion. The Hierarchical Context Recalibration (HCR) module enhances tumor representation by combining multi-scale contextual aggregation with adaptive feature recalibration. The Guided Region Screening (GRS) module refines skip connection features through encoder-decoder guidance and modality-aware information, improving the reliability of feature transmission. The Hierarchical Region Decoding Head (HRD) further introduces region-aware supervision to enhance structural consistency among tumor subregions. Validation results on the BraTS 2026 Task 3 validation set demonstrate that MCS-UNet shows stable segmentation performance under heterogeneous multi-center MRI conditions. The source code is publicly available at \url{https://github.com/zhang-yxin/models} to facilitate reproducibility and future research.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BraTS_GoAT_006.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{WanHon_MCSUNet_MICCAISAT2026,
author = { Wang, Hongjuan AND Zhang, Yixin AND Li, Chengwei AND Feng, Yan AND Qu, Qingchao AND An, Xinjun},
title = { { MCS-UNet: A Multimodal Collaborative Framework for Robust Multi-center MRI Brain Tumor Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17253},
month = {pending},
page = {pending}
}