Related Experiment Video
Updated: Jan 14, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.4K
Enhancing brain tumor segmentation using attention based convolutional UNet on MRI images.
Mohammad Abrar1, Abdu Salam2, Faizan Ullah3
1Faculty of Computer Studies, Arab Open University, 122, Muscat, P.O. Box 1596, Oman. abrar.m@aou.edu.om.
Scientific Reports
|October 21, 2025
Summary
Attention-based Convolutional U-Net (ACU-Net) improves brain tumor segmentation accuracy on MRI scans. This AI model enhances precision and reliability for clinical applications, outperforming traditional methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Biology
Background:
- Accurate brain tumor segmentation is critical for diagnosis and treatment planning.
- Current automated methods struggle with complex tumor shapes, while manual segmentation is time-consuming and subjective.
- Existing deep learning models like U-Nets and CNNs have limitations in capturing intricate tumor boundaries.
Purpose of the Study:
- To propose and evaluate an Attention-based Convolutional U-Net (ACU-Net) model for enhanced brain tumor segmentation on MRI data.
- To improve the precision and dependability of tumor edge delineation using attention mechanisms within a U-Net architecture.
- To quantitatively assess the performance of ACU-Net against baseline models using standard segmentation metrics.
Main Methods:
- Utilized the BraTS 2018 MRI dataset for brain tumor segmentation.
- Preprocessed data through normalization, spatial resolution adjustment, and augmentation.
- Developed the ACU-Net model incorporating attention gates, trained using dice and cross-entropy loss functions.
- Compared ACU-Net performance with U-Nets and Convolutional Neural Networks (CNNs) using precision, recall, Dice Similarity Coefficient (DSC), and Intersection over Union (IoU).
Main Results:
- ACU-Net achieved high Dice Similarity Coefficients: 94.04% for Whole Tumor (WT), 98.63% for Tumor Core (TC), and 98.77% for Enhancing Tumor (ET).
- The proposed ACU-Net model demonstrated superior performance compared to baseline U-Net and CNN models.
- Attention mechanisms significantly improved the accuracy and robustness of medical image segmentation for brain tumors.
Conclusions:
- The ACU-Net model offers a reliable and effective tool for precise brain tumor segmentation in clinical settings.
- Attention mechanisms are crucial for enhancing the performance of deep learning models in medical image analysis.
- The study highlights the potential of ACU-Net to overcome limitations of existing segmentation methods, improving diagnostic and therapeutic planning.
