Asymmetric encoder-decoder U-net with hybrid attention and residual dilated convolution for brain tumor segmentation
Abstract
InroductionThe accurate and automatic segmentation of brain tumors from magnetic resonance imaging is regarded as essential for clinical diagnosis and surgical planning. Nevertheless, the concurrent extraction of both local fine details and global contextual cues remains challenging, primarily due to ambiguous tumor margins and marked disparities in tumor dimensions.MethodsTo address these issues, a novel HCSA-RMSDC U-Net is designed in this work, which is characterized by an asymmetric architecture with a computationally heavy encoder and a lightweight decoder. Specifically, a hybrid channel–spatial attention (HCSA) module is incorporated into the encoder, wherein axial decomposed convolution is utilized to capture directional spatial information, thereby enhancing spatial feature representation for tumor segmentation. Concurrently, a residual multi-scale dilated convolution (RMSDC) module is deployed in the bottleneck layer to capture semantic information across varying scales without increasing the spatial resolution.ResultsThe proposed model was evaluated on the publicly available BRISC 2025 and Figshare datasets, achieving Dice coefficients of 87.39% and 82.84%, respectivelyDiscussionThese findings support the complementary roles of the HCSA and RMSDC mechanisms in enhancing feature representation and multi-scale contextual modeling. Overall, the proposed framework provides a fully automated approach for brain tumor segmentation.
Related
- PECSA Net: Pupil Segmentation Using Attention Integrated Encoder-based Multi-Level Feature Extractor for Analysis of Ocular Dynamics
- AMF-U-Net: an adaptive multimodal fusion residual attention 3D U-Net for boundary-aware brain tumour segmentation
- A Hybrid Residual-Swin Transformer Design with Attention for Prostate Cancer Segmentation
- EPRA U-Net: an efficient pyramid residual attention framework for accurate infarct segmentation in diffusion-weighted MRI.
- Vehicle lane change prediction with explainable soft mask attention mechanism and heterogeneous information encoder
BibTeX
@article{cao2026asymmetric,
title = {{Asymmetric encoder-decoder U-net with hybrid attention and residual dilated convolution for brain tumor segmentation}},
author = {Xiaoshuai Cao and Xige Song and Jincan Zhang and Wenna Chen and Ganqin Du},
journal = {Frontiers in Neuroscience},
year = {2026},
month = oct,
doi = {10.3389/fnins.2026.1999080},
url = {https://www.frontiersin.org/articles/10.3389/fnins.2026.1999080},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Asymmetric encoder-decoder U-net with hybrid attention and residual dilated convolution for brain tumor segmentation
AU - Xiaoshuai Cao
AU - Xige Song
AU - Jincan Zhang
AU - Wenna Chen
AU - Ganqin Du
PY - 2026
DA - 2026/10/28
JO - Frontiers in Neuroscience
DO - 10.3389/fnins.2026.1999080
UR - https://www.frontiersin.org/articles/10.3389/fnins.2026.1999080
AB - InroductionThe accurate and automatic segmentation of brain tumors from magnetic resonance imaging is regarded as essential for clinical diagnosis and surgical planning. Nevertheless, the concurrent extraction of both local fine details and global contextual cues remains challenging, primarily due to ambiguous tumor margins and marked disparities in tumor dimensions.MethodsTo address these issues, a novel HCSA-RMSDC U-Net is designed in this work, which is characterized by an asymmetric architecture with a computationally heavy encoder and a lightweight decoder. Specifically, a hybrid channel–spatial attention (HCSA) module is incorporated into the encoder, wherein axial decomposed convolution is utilized to capture directional spatial information, thereby enhancing spatial feature representation for tumor segmentation. Concurrently, a residual multi-scale dilated convolution (RMSDC) module is deployed in the bottleneck layer to capture semantic information across varying scales without increasing the spatial resolution.ResultsThe proposed model was evaluated on the publicly available BRISC 2025 and Figshare datasets, achieving Dice coefficients of 87.39% and 82.84%, respectivelyDiscussionThese findings support the complementary roles of the HCSA and RMSDC mechanisms in enhancing feature representation and multi-scale contextual modeling. Overall, the proposed framework provides a fully automated approach for brain tumor segmentation.
SN - 1662-453X
SN - 1662-4548
ER -