TY - GEN
T1 - BLSegMamba
T2 - Brain TumorS Lighthouse Cluster of Challenges, and the Automated Identification of Moderate-Severe Traumatic Brain Injury Lesions Challenge, BraTS 2025 and AIMS-TBI 2025, held in Conjunction International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2025
AU - Zhu, Yueyue
AU - Bai, Xiaoyu
AU - Jiang, Haotian
AU - Chen, Geng
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026
Y1 - 2026
N2 - Moderate-to-Severe Traumatic Brain Injury (msTBI) often leads to complex and highly heterogeneous structural damage in the brain. Lesions may be focal or diffuse and can involve multiple tissue types, including gray matter, white matter, and cerebrospinal fluid. They also exhibit considerable variability in size, shape, spatial distribution, and hemispheric symmetry. This high degree of heterogeneity greatly increases the difficulty of automatic segmentation based on unimodal T1-weighted MRI. To address this challenge, we propose a customized optimization of the SegMamba architecture. The resulting optimized version, Brain Lesion SegMamba (BLSegMamba), retains the core structural components of SegMamba while integrating a more robust data augmentation strategy and a loss function specifically designed for the segmentation of msTBI lesions. On the final test dataset of the AIMS-TBI Challenge, our BLSegMamba achieves the top overall ranking after weighted aggregation of all evaluation metrics. Our code is publicly available at https://github.com/YueyueZhu/BLSegMamba.
AB - Moderate-to-Severe Traumatic Brain Injury (msTBI) often leads to complex and highly heterogeneous structural damage in the brain. Lesions may be focal or diffuse and can involve multiple tissue types, including gray matter, white matter, and cerebrospinal fluid. They also exhibit considerable variability in size, shape, spatial distribution, and hemispheric symmetry. This high degree of heterogeneity greatly increases the difficulty of automatic segmentation based on unimodal T1-weighted MRI. To address this challenge, we propose a customized optimization of the SegMamba architecture. The resulting optimized version, Brain Lesion SegMamba (BLSegMamba), retains the core structural components of SegMamba while integrating a more robust data augmentation strategy and a loss function specifically designed for the segmentation of msTBI lesions. On the final test dataset of the AIMS-TBI Challenge, our BLSegMamba achieves the top overall ranking after weighted aggregation of all evaluation metrics. Our code is publicly available at https://github.com/YueyueZhu/BLSegMamba.
KW - Mamba
KW - Segmentation
KW - U-net
KW - msTBI
UR - https://www.scopus.com/pages/publications/105038042384
U2 - 10.1007/978-3-032-16370-7_27
DO - 10.1007/978-3-032-16370-7_27
M3 - 会议稿件
AN - SCOPUS:105038042384
SN - 9783032163691
T3 - Lecture Notes in Computer Science
SP - 301
EP - 310
BT - Segmentation, Classification, and Synthesis for Brain Tumors and Traumatic Brain Injuries - MICCAI 2025 Challenges
A2 - Bakas, Spyridon
A2 - Dennis, Emily
A2 - Astaraki, Mehdi
A2 - Baid, Ujjwal
A2 - Conte, Gian Marco
A2 - Foltyn-Dumitru, Martha
A2 - Jiang, Zhifan
A2 - Linguraru, Marius George
A2 - Labella, Dominic
A2 - Metz, Marie-Christin
A2 - Anazodo, Udunna
A2 - de Verdier, Maria Correia
A2 - Kofler, Florian
A2 - Li, Hongwei Bran
A2 - Maleki, Nazanin
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 23 September 2025 through 27 September 2025
ER -