Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model

The distinctions between tumor areas and surrounding tissues are often subtle. Additionally, the morphology and size of tumors can vary significantly among different patients. These factors pose considerable challenges for the precise segmentation of tumors and subsequent diagnosis. Therefore, resea...

Full description

Saved in:
Bibliographic Details
Main Authors: Shuli Xing, Zhenwei Lai, Junxiong Zhu, Wenwu He, Guojun Mao
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/11/5981
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850130855873216512
author Shuli Xing
Zhenwei Lai
Junxiong Zhu
Wenwu He
Guojun Mao
author_facet Shuli Xing
Zhenwei Lai
Junxiong Zhu
Wenwu He
Guojun Mao
author_sort Shuli Xing
collection DOAJ
description The distinctions between tumor areas and surrounding tissues are often subtle. Additionally, the morphology and size of tumors can vary significantly among different patients. These factors pose considerable challenges for the precise segmentation of tumors and subsequent diagnosis. Therefore, researchers are trying to develop an automated and accurate segmentation model. Currently, many segmentation models in deep learning rely on Convolutional Neural Network or Vision Transformer. However, Convolution-based models often fail to deliver precise segmentation results, while Transformer-based models often require more computational resources. To address these challenges, we propose a novel hybrid model named Local–Global UNet Transformer. In our model, we introduce: (1) a semantic-oriented masked attention to enhance the feature extraction capability of the decoder; and (2) network-in-network blocks to increase channel modeling complexity in the encoder while reducing the parameter consumption associated with residual blocks. We evaluate our model on two public brain tumor segmentation datasets, and the experimental results demonstrate that our model achieves the highest average Dice score on the BraTS2024-GLI dataset and ranks second on the BraTS2023-GLI dataset. In terms of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>H</mi><msub><mi>D</mi><mn>95</mn></msub></mrow></semantics></math></inline-formula>, our model attains the lowest values on both datasets. Furthermore, the ablation study proves the effectiveness of our model design.
format Article
id doaj-art-e79226962ca144eeadcedd3509cda0e9
institution OA Journals
issn 2076-3417
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-e79226962ca144eeadcedd3509cda0e92025-08-20T02:32:34ZengMDPI AGApplied Sciences2076-34172025-05-011511598110.3390/app15115981Semantic Segmentation of Brain Tumors Using a Local–Global Attention ModelShuli Xing0Zhenwei Lai1Junxiong Zhu2Wenwu He3Guojun Mao4College of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, ChinaCollege of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, ChinaCollege of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, ChinaCollege of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, ChinaCollege of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, ChinaThe distinctions between tumor areas and surrounding tissues are often subtle. Additionally, the morphology and size of tumors can vary significantly among different patients. These factors pose considerable challenges for the precise segmentation of tumors and subsequent diagnosis. Therefore, researchers are trying to develop an automated and accurate segmentation model. Currently, many segmentation models in deep learning rely on Convolutional Neural Network or Vision Transformer. However, Convolution-based models often fail to deliver precise segmentation results, while Transformer-based models often require more computational resources. To address these challenges, we propose a novel hybrid model named Local–Global UNet Transformer. In our model, we introduce: (1) a semantic-oriented masked attention to enhance the feature extraction capability of the decoder; and (2) network-in-network blocks to increase channel modeling complexity in the encoder while reducing the parameter consumption associated with residual blocks. We evaluate our model on two public brain tumor segmentation datasets, and the experimental results demonstrate that our model achieves the highest average Dice score on the BraTS2024-GLI dataset and ranks second on the BraTS2023-GLI dataset. In terms of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>H</mi><msub><mi>D</mi><mn>95</mn></msub></mrow></semantics></math></inline-formula>, our model attains the lowest values on both datasets. Furthermore, the ablation study proves the effectiveness of our model design.https://www.mdpi.com/2076-3417/15/11/5981braintumorsemantic segmentationhybrid modelchannel modeling
spellingShingle Shuli Xing
Zhenwei Lai
Junxiong Zhu
Wenwu He
Guojun Mao
Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
Applied Sciences
braintumor
semantic segmentation
hybrid model
channel modeling
title Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
title_full Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
title_fullStr Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
title_full_unstemmed Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
title_short Semantic Segmentation of Brain Tumors Using a Local–Global Attention Model
title_sort semantic segmentation of brain tumors using a local global attention model
topic braintumor
semantic segmentation
hybrid model
channel modeling
url https://www.mdpi.com/2076-3417/15/11/5981
work_keys_str_mv AT shulixing semanticsegmentationofbraintumorsusingalocalglobalattentionmodel
AT zhenweilai semanticsegmentationofbraintumorsusingalocalglobalattentionmodel
AT junxiongzhu semanticsegmentationofbraintumorsusingalocalglobalattentionmodel
AT wenwuhe semanticsegmentationofbraintumorsusingalocalglobalattentionmodel
AT guojunmao semanticsegmentationofbraintumorsusingalocalglobalattentionmodel