Brain metastases segmentation improved with enhanced neural network and post-processing
Pre- and Post-Treatment Brain Metastases Segmentation Using nnU-Net with Post-Processing for BraTS 2026
Computer Vision and Pattern Recognition
Summary
Segmenting brain metastases before and after treatment is hard because the lesions look very different in size and appearance. The authors built a system using a specialized neural network called nnU-Net with added rule-based steps that help detect lesions more accurately. They carefully tested their method to make sure improvements were real and not just luck on the leaderboard scores. This work also explains why some kinds of post-processing help more than others and reports many unsuccessful experiments to guide future research.
What this means in practice
- •For medical imaging teams: Segment pre- and post-treatment brain metastases from MRI scans with improved accuracy using a neural network and post-processing pipeline tuned for lesion detection.
- •For radiology software developers: Incorporate robust, ensemble-based neural network models with targeted post-processing to enhance detection and volume estimation of metastatic brain tumors in clinical tools.
Authors
Haobin Liu, Xin Wang
Abstract
Brain metastases exhibit high inter-lesion variability in size, enhancement pattern, and post-treatment appearance, making volumetric segmentation of both pre- and post-treatment cases the central challenge of the BraTS 2026 Task 1 (Brain Metastases). We build a pragmatic pipeline on a 5-fold nnU-Net ResEnc-L ensemble, in which each fold is trained independently for 1,000 epochs with the standard Dice + cross-entropy loss on 1,296 four-modality training cases. This ensemble is followed by a rule-based post-processing cascade tuned for the lesion-wise Dice similarity coefficient (LW-DSC), a detection-oriented metric that behaves very differently from the traditional global Dice. The final pipeline reaches an LW-DSC of 0.733 / 0.751 / 0.713 / 0.549 on the enhancing tumour (ET), tumour core (TC), whole tumour (WT), and resection cavity (RC) sub-regions on the official validation leaderboard. Rather than trusting these leaderboard gains, we audit every post-processing stage with a five-fold out-of-fold (OOF) analysis with no model-training leakage over all 1,296 training cases, scored with the official BraTS evaluation code (BraTS_evaluation): it confirms two stages as robust, per-fold-consistent improvements while the third improves only the leaderboard and does not reproduce out-of-fold. We further provide a mechanistic analysis of the LW-DSC metric that explains why recall-recovering post-processing carries low risk whereas component deletion does not, and we report thirteen negative results spanning loss engineering, alternative backbones, and inference-time settings, several of which run counter to widely held intuitions. Source code is released under Apache-2.0 at https://github.com/hornbeamliu/brats2026-met.