Abstract
Small brain metastases are frequently missed by whole-volume semantic segmentation, while increasing sensitivity can increase false pos- itives predictions. We present METS-PRO R2, a lesion-aware refinement of a high-resolution residual-encoder nnU-Net for BraTS MICCAI 2026 Task 1. An equal-mass enhancing-tumour proposal field gates a strictly non-negative residual added only to the enhancing-tumour logit. Across 1,295 out-of-fold scans, lesion-count micro F1 increased from 0.7056 to 0.7171 for enhancing tumour, from 0.7186 to 0.7298 for tumour core, and from 0.7044 to 0.7148 for whole tumour. The improvement persisted on a patient-independent subset of 552 scans, with R2 scores of 0.7215, 0.7344, and 0.7217, respectively. A matched continuation control using the same baseline checkpoints and 60 additional epochs showed only minimal gains, supporting the contribution of the proposed refinement. On the hidden challenge set, the final five-fold ensemble achieved all-instance F1 scores of 0.7517, 0.7660, and 0.7437 for enhancing tumour, tumour core, and whole tumour, respectively. Resection-cavity performance remained weak. A conservative post-hoc tiny-lesion hybrid was rejected after official out-of- fold evaluation because microscopic small-lesion gains were offset by more regressions in large-lesion and boundary metrics. The results support targeted residual refinement as an practical strategy for improving lesion detection while preserving the base segmentation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BraTS_METS_033.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/BraTS_METS_033_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=t3gkVUYwE3¬eId=GPUT8Hytvs
BibTex
@InProceedings{SinMan_METSPRO_MICCAISAT2026,
author = { Singla, Manthan AND George, Ajo Babu AND Ranjan, Uma},
title = { { METS-PRO R2: Lesion-Aware Positive Residual Refinement for Brain Metastasis 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}
}
