Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Attention-Enhanced Hybrid Feature Aggregation Network for 3D Brain Tumor Segmentation

Ziya Ata Yazıcı, İlkay Öksüz, Hazım Kemal Ekenel

TL;DR

This work tackles automated 3D glioblastoma segmentation from multi-modal MRI by introducing GLIMS, a U-Net–style network that blends depth-wise multi-scale convolutions with Swin Transformer blocks and attention-guided skip connections to capture both local and global context. The architecture employs patch-based training, deep supervision, and MONAI-based on-the-fly augmentations, achieving robust performance on the BraTS 2023 dataset across whole tumor, tumor core, and enhancing tumor regions. Quantitatively, GLIMS achieves competitive lesion-wise Dice scores, outperforming nnU-Net baselines in cross-validation and delivering strong validation and test results after targeted post-processing and ensembling. The study emphasizes practical considerations for clinical deployment, including computational efficiency and the potential benefits of synthetic data to improve generalization to diverse data distributions.

Abstract

Glioblastoma is a highly aggressive and malignant brain tumor type that requires early diagnosis and prompt intervention. Due to its heterogeneity in appearance, developing automated detection approaches is challenging. To address this challenge, Artificial Intelligence (AI)-driven approaches in healthcare have generated interest in efficiently diagnosing and evaluating brain tumors. The Brain Tumor Segmentation Challenge (BraTS) is a platform for developing and assessing automated techniques for tumor analysis using high-quality, clinically acquired MRI data. In our approach, we utilized a multi-scale, attention-guided and hybrid U-Net-shaped model -- GLIMS -- to perform 3D brain tumor segmentation in three regions: Enhancing Tumor (ET), Tumor Core (TC), and Whole Tumor (WT). The multi-scale feature extraction provides better contextual feature aggregation in high resolutions and the Swin Transformer blocks improve the global feature extraction at deeper levels of the model. The segmentation mask generation in the decoder branch is guided by the attention-refined features gathered from the encoder branch to enhance the important attributes. Moreover, hierarchical supervision is used to train the model efficiently. Our model's performance on the validation set resulted in 92.19, 87.75, and 83.18 Dice Scores and 89.09, 84.67, and 82.15 Lesion-wise Dice Scores in WT, TC, and ET, respectively. The code is publicly available at https://github.com/yaziciz/GLIMS.

Attention-Enhanced Hybrid Feature Aggregation Network for 3D Brain Tumor Segmentation

TL;DR

This work tackles automated 3D glioblastoma segmentation from multi-modal MRI by introducing GLIMS, a U-Net–style network that blends depth-wise multi-scale convolutions with Swin Transformer blocks and attention-guided skip connections to capture both local and global context. The architecture employs patch-based training, deep supervision, and MONAI-based on-the-fly augmentations, achieving robust performance on the BraTS 2023 dataset across whole tumor, tumor core, and enhancing tumor regions. Quantitatively, GLIMS achieves competitive lesion-wise Dice scores, outperforming nnU-Net baselines in cross-validation and delivering strong validation and test results after targeted post-processing and ensembling. The study emphasizes practical considerations for clinical deployment, including computational efficiency and the potential benefits of synthetic data to improve generalization to diverse data distributions.

Abstract

Glioblastoma is a highly aggressive and malignant brain tumor type that requires early diagnosis and prompt intervention. Due to its heterogeneity in appearance, developing automated detection approaches is challenging. To address this challenge, Artificial Intelligence (AI)-driven approaches in healthcare have generated interest in efficiently diagnosing and evaluating brain tumors. The Brain Tumor Segmentation Challenge (BraTS) is a platform for developing and assessing automated techniques for tumor analysis using high-quality, clinically acquired MRI data. In our approach, we utilized a multi-scale, attention-guided and hybrid U-Net-shaped model -- GLIMS -- to perform 3D brain tumor segmentation in three regions: Enhancing Tumor (ET), Tumor Core (TC), and Whole Tumor (WT). The multi-scale feature extraction provides better contextual feature aggregation in high resolutions and the Swin Transformer blocks improve the global feature extraction at deeper levels of the model. The segmentation mask generation in the decoder branch is guided by the attention-refined features gathered from the encoder branch to enhance the important attributes. Moreover, hierarchical supervision is used to train the model efficiently. Our model's performance on the validation set resulted in 92.19, 87.75, and 83.18 Dice Scores and 89.09, 84.67, and 82.15 Lesion-wise Dice Scores in WT, TC, and ET, respectively. The code is publicly available at https://github.com/yaziciz/GLIMS.
Paper Structure (11 sections, 5 equations, 4 figures, 3 tables)

This paper contains 11 sections, 5 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Dataset
  3. Methods
  4. Model Overview
  5. Data Pre-processing
  6. Evaluation Metrics & Loss Function
  7. Deep Supervision
  8. Post-processing
  9. Implementation Details
  10. Results
  11. Discussion and Conclusions

Figures (4)

  • Figure 1: A sample MRI scan displayed in four modalities -- T1, T1c, T2, FLAIR -- and the corresponding segmentation mask, left to right. NCR is represented by green, ET by red, and ED by yellow.
  • Figure 2: The proposed architecture of 3D segmentation model, GLIMS. Each color represents a unique module.
  • Figure 3: The proposed DACB and CSAB modules from left to right, respectively.
  • Figure 4: The prediction result of Case ID: 208 in the validation set. Left: The T2 image of the slice. Middle: The segmented output. Right: 3D rendered visualization of the tumor. The yellow, red, and green colors represent ED, ET, and NCR regions.