Abstract
Moderate-to-Severe Traumatic Brain Injury (msTBI) is a major health concern arising from external forces that cause structural brain damage and can result in life-threatening conditions. Accurate lesion segmentation and detection remain highly challenging due to the heterogeneity of msTBI, which limits the effectiveness of conventional neuroimaging methods. The AIMS-TBI Challenge 2026 was established to advance automated approaches for msTBI lesion detection and segmentation, leveraging a large multi-site dataset of T1-weighted MRIs. This dataset comprised both lesion-present and lesion-free cases, highlighting the importance of accurate lesion detection alongside segmentation. In this work, we present our submission to the challenge, building on the nnU-Net framework and exploring the domain randomization strategy. We expanded the training data with synthetic images generated with additional small lesions, and reframed the task as a multiclass segmentation problem, distinguishing small from large lesions as separate classes. On the evaluation test set, our method achieved promising performances, with an overall Dice score of 0.557 across all images containing lesions. A simpler approach relying solely on real data was used for the detection task, reaching a balanced accuracy of 0.849 in the test set.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/AIMS_TBI_008.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/profile?id=~Rachika_E._Hamadache1
BibTex
@InProceedings{HamRac_nnUNet_MICCAISAT2026,
author = { Hamadache, Rachika E. AND Casamitjana, Adrià AND Cabezas, Mariano AND Oliver, Arnau AND Lladó, Xavier},
title = { { nnU-Net and Domain Randomization for Automated Segmentation of Moderate-to-Severe Traumatic Brain Injury } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17254},
month = {pending},
page = {pending}
}
