Related Experiment Video
Updated: May 10, 2025
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
DSIT UNet a dual stream iterative transformer based UNet architecture for segmenting brain tumors from FLAIR MRI
Shakib Al Hasan1,2, S M Mahim1,2, Md Emamul Hossen1,2
1Department of Biomedical Engineering, Islamic University, Kushtia, 7003, Bangladesh.
Abstract:
Brain tumor segmentation remains challenging in medical imaging with conventional therapies and rehabilitation owing to the complex morphology and heterogeneous nature of tumors. Although convolutional neural networks (CNNs) have advanced medical image segmentation, they struggle with long-range dependencies because of their limited receptive fields. We propose Dual-Stream Iterative Transformer UNet (DSIT-UNet), a novel framework that combines Iterative Transformer (IT) modules with a dual-stream encoder-decoder architecture. Our model incorporates a transformed spatial-hybrid attention optimization (TSHAO) module to enhance multiscale feature interactions and balance local details with the global context. We evaluated DSIT-UNet using three benchmark datasets: The Cancer Imaging Archive (TCIA) from The Cancer Genome Atlas (TCGA), BraTS2020, and BraTS2021. On TCIA, our model achieved a Mean Intersection over Union of 95.21%, mean Dice Coefficient of 96.23%, precision of 95.91%, and recall of 96.55%. BraTS2020 attained a Mean IoU of 95.88%, mDice of 96.32%, precision of 96.21%, and recall of 96.44%, surpassing the performance of the existing methods. The superior results of DSIT-UNet demonstrate its effectiveness in capturing tumor boundaries and improving segmentation robustness through hierarchical attention mechanisms and multiscale feature extraction. This architecture advances automated brain tumor segmentation, with potential applications in clinical neuroimaging and future extensions to 3D volumetric segmentation.
Insights
A new Dual-Stream Iterative Transformer UNet (DSIT-UNet) improves brain tumor segmentation accuracy in medical images. This advanced model effectively captures tumor boundaries, outperforming existing methods for better clinical neuroimaging applications.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Brain tumor segmentation is crucial for treatment planning but challenging due to tumor complexity.
- Conventional Convolutional Neural Networks (CNNs) struggle with long-range dependencies in medical image segmentation.
Purpose of the Study:
- To introduce a novel framework, Dual-Stream Iterative Transformer UNet (DSIT-UNet), for enhanced automated brain tumor segmentation.
- To address the limitations of existing CNNs in capturing long-range dependencies and multiscale features.
Main Methods:
- Developed DSIT-UNet, integrating Iterative Transformer (IT) modules within a dual-stream encoder-decoder architecture.
- Incorporated a transformed spatial-hybrid attention optimization (TSHAO) module for improved multiscale feature interaction.
- Validated the model on TCIA (TCGA), BraTS2020, and BraTS2021 benchmark datasets.
Main Results:
- DSIT-UNet achieved high performance on benchmark datasets, with Mean Intersection over Union (IoU) up to 95.21% and Dice Coefficient up to 96.23%.
- The model demonstrated superior segmentation accuracy compared to existing methods on BraTS2020.
- Achieved high precision (95.91%) and recall (96.55%) on TCIA, indicating robust tumor boundary detection.
Conclusions:
- DSIT-UNet effectively captures tumor boundaries and enhances segmentation robustness through hierarchical attention and multiscale feature extraction.
- The proposed architecture represents a significant advancement in automated brain tumor segmentation.
- Potential applications include clinical neuroimaging and future extensions to 3D segmentation.
