Abstract
Automated brain metastases segmentation must combine sensitivity to small lesions with reliable semantic separation of enhancing tumor (ET), non-enhancing tumor core (NETC), surrounding non-enhancing FLAIR hyperintensity (SNFH), and resection cavity (RC). We present a clinically guided multistage inference pipeline for BraTS 2026 Task 1. Two complementary nnU-Net v2 residual-encoder segmentation sources are fused with an internal OOF-selected recall-dominant weight of 0.7/0.3 and combined with auxiliary small-lesion detection and RC-specific evidence. Probability-aware re-fusion and connected-component post-processing filter low-confidence ET/NETC components, suppress RC predictions on baseline scans identified by the supplied longitudinal suffix, relabel uncertain RC according to local class probabilities and shell context, and consolidate RC at connected-region level. The ensemble weight and RC-consolidation operating point were selected by internal OOF proxy sweeps; semantic RC thresholds were expert-defined from qualitative failure-mode review and were not numerically optimized against validation labels or leaderboard scores. On the 179-case official validation set, the final pipeline achieved lesion-wise instance F1 of 0.7997 for ET, 0.4901 for RC, 0.7906 for TC, and 0.7548 for WT. The method provides an interpretable framework for targeted RC refinement after complementary evidence has been combined.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BraTS_METS_016.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=oKTZh7k2JI¬eId=oKTZh7k2JI
BibTex
@InProceedings{LinAle_Ensemble_MICCAISAT2026,
author = { Lindemann, Alessia AND Milanesi, Nadia AND Cappellini, Luca A. AND Savini, Giovanni AND Politi, Letterio S. AND Levi, Riccardo},
title = { { Ensemble nnU-Nets for pre- and post-operative Brain Metastases Segmentation in BraTS 2026 Task 1 } },
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}
}
