ResSAXU-Net for multimodal brain tumor segmentation from brain MRI
Abstract Glioma, the most common brain tumour, carries the highest risk of death. Successful treatment planning and the accurate diagnosis of glioma depend heavily on magnetic resonance imaging (MRI). Classification of brain tumours from MR data should be automated for rigorous pathologic diagnosis...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-09539-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849332772538155008 |
|---|---|
| author | Zheyuan Xiong |
| author_facet | Zheyuan Xiong |
| author_sort | Zheyuan Xiong |
| collection | DOAJ |
| description | Abstract Glioma, the most common brain tumour, carries the highest risk of death. Successful treatment planning and the accurate diagnosis of glioma depend heavily on magnetic resonance imaging (MRI). Classification of brain tumours from MR data should be automated for rigorous pathologic diagnosis and ongoing monitoring. Because of glioma’s aggressive potential and diverse characteristics, standardised and accurate classification methods classifying tumours within the bladder are essential. Recent studies of U-Net separation of brain tumours have revealed challenges related to inadequate down-sampling feature extraction and loss of information from up-sampling. It is important to address these problems to enhance the accuracy of U-Net in classifying brain tumours. Deep residual network and squeeze-excitation network U-Net model. The enhanced version of Ressaxu-Net presented in this work has two new features: Ressaxu-Net improves feature information extraction and solves brain tumour classification problems by using deep residual networks to reduce network damage. By reducing information loss, the squeeze-excitation network enables the network to prioritise the most essential feature maps. This method improves the classification accuracy of small brain tumours, thereby solving problems associated with poor performance. Combining dice loss and cross-entropy loss, the fusion loss function is introduced to deal with issues such as network convergence and data imbalance, and then Ressaxu-Net performance was simulated using the Brats2018 and Brats2019 datasets study, examining how the model performs in brain tumour classification. According to the experimental results, Ressaxu-Net obtained dice similarity coefficients of 0.9597,0.9618 and 0.9595 for the total tumour, intratumoral, and elevated tumours, respectively, and 8.10%, 15.88%, and 17.33% showed improvement. This suggests that Ressaxu-Net is competitively effective in accurately classifying multiple brain tumours. |
| format | Article |
| id | doaj-art-67e75a9133a04ab7a4aafb12c83c18b6 |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-67e75a9133a04ab7a4aafb12c83c18b62025-08-20T03:46:07ZengNature PortfolioScientific Reports2045-23222025-07-0115111510.1038/s41598-025-09539-1ResSAXU-Net for multimodal brain tumor segmentation from brain MRIZheyuan Xiong0School of Computer Science, Jiangxi University of Chinese MedicineAbstract Glioma, the most common brain tumour, carries the highest risk of death. Successful treatment planning and the accurate diagnosis of glioma depend heavily on magnetic resonance imaging (MRI). Classification of brain tumours from MR data should be automated for rigorous pathologic diagnosis and ongoing monitoring. Because of glioma’s aggressive potential and diverse characteristics, standardised and accurate classification methods classifying tumours within the bladder are essential. Recent studies of U-Net separation of brain tumours have revealed challenges related to inadequate down-sampling feature extraction and loss of information from up-sampling. It is important to address these problems to enhance the accuracy of U-Net in classifying brain tumours. Deep residual network and squeeze-excitation network U-Net model. The enhanced version of Ressaxu-Net presented in this work has two new features: Ressaxu-Net improves feature information extraction and solves brain tumour classification problems by using deep residual networks to reduce network damage. By reducing information loss, the squeeze-excitation network enables the network to prioritise the most essential feature maps. This method improves the classification accuracy of small brain tumours, thereby solving problems associated with poor performance. Combining dice loss and cross-entropy loss, the fusion loss function is introduced to deal with issues such as network convergence and data imbalance, and then Ressaxu-Net performance was simulated using the Brats2018 and Brats2019 datasets study, examining how the model performs in brain tumour classification. According to the experimental results, Ressaxu-Net obtained dice similarity coefficients of 0.9597,0.9618 and 0.9595 for the total tumour, intratumoral, and elevated tumours, respectively, and 8.10%, 15.88%, and 17.33% showed improvement. This suggests that Ressaxu-Net is competitively effective in accurately classifying multiple brain tumours.https://doi.org/10.1038/s41598-025-09539-1GliomaBrain tumourRessaxu-NetClassificationBrats2018 and Brats2019 |
| spellingShingle | Zheyuan Xiong ResSAXU-Net for multimodal brain tumor segmentation from brain MRI Scientific Reports Glioma Brain tumour Ressaxu-Net Classification Brats2018 and Brats2019 |
| title | ResSAXU-Net for multimodal brain tumor segmentation from brain MRI |
| title_full | ResSAXU-Net for multimodal brain tumor segmentation from brain MRI |
| title_fullStr | ResSAXU-Net for multimodal brain tumor segmentation from brain MRI |
| title_full_unstemmed | ResSAXU-Net for multimodal brain tumor segmentation from brain MRI |
| title_short | ResSAXU-Net for multimodal brain tumor segmentation from brain MRI |
| title_sort | ressaxu net for multimodal brain tumor segmentation from brain mri |
| topic | Glioma Brain tumour Ressaxu-Net Classification Brats2018 and Brats2019 |
| url | https://doi.org/10.1038/s41598-025-09539-1 |
| work_keys_str_mv | AT zheyuanxiong ressaxunetformultimodalbraintumorsegmentationfrombrainmri |