Related Experiment Video
Updated: Jun 13, 2026
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
ABR-UNet3D: Aspect-Aware Boundary-Resilient Attention for Robust Cardiac MRI Segmentation
Serdar Akyel1, Zeki Cetinkaya2, Fatih Topaloglu3
1Afyonkarahisar State Hospital, Ministry of Health, 03030 Afyonkarahisar, Türkiye.
Insights
A new Aspect-Aware Boundary-Resilient UNet3D (ABR-UNet3D) model improves cardiac MRI segmentation, especially in challenging right ventricle (RV) and myocardium (MYO) regions, by enhancing boundary detection.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Biomedical Engineering
Background:
- Cardiac MRI segmentation faces challenges due to low contrast and indistinct boundaries in myocardium (MYO) and right ventricle (RV).
- Existing segmentation methods struggle with anatomical variability, impacting reliability.
- Robust, boundary-aware approaches are crucial for accurate cardiac image analysis.
Purpose of the Study:
- To introduce a novel Aspect-Aware Boundary-Resilient UNet3D (ABR-UNet3D) architecture for improved cardiac MRI segmentation.
- To enhance the model's ability to capture contextual information and refine boundaries.
- To evaluate the proposed method's performance and robustness on a standard dataset.
Main Methods:
- Developed an Aspect-Aware Complementary Attention (AAC) module combining multi-planar context and complementary gating.
- Implemented the ABR-UNet3D architecture for cardiac MRI segmentation.
- Evaluated performance using Dice Similarity Coefficient (DSC), Intersection over Union (IoU), and boundary metrics (HD95, ASD, Surface Dice) on the ACDC dataset.
- Conducted five-fold cross-validation and ablation studies to assess robustness and component contributions.
Main Results:
- Achieved a mean DSC of 0.9603 in single-run experiments and 0.952 ± 0.009 in cross-validation on the ACDC dataset.
- Demonstrated consistent performance in challenging RV and MYO segmentation.
- Boundary metrics showed improved surface agreement and reduced errors compared to baseline models.
- Ablation studies confirmed the synergistic benefit of AAC module components.
Conclusions:
- The ABR-UNet3D architecture offers a stable and competitive framework for cardiac MRI segmentation.
- Jointly modeling context and boundary refinement enhances segmentation reliability in difficult regions.
- The proposed method achieves consistent and improved performance over existing approaches for cardiac MRI analysis.
Abstract:
Background: Cardiac magnetic resonance (MRI) images often exhibit low contrast, anatomical variability, and indistinct boundaries, particularly in the myocardium (MYO) and right ventricle (RV). These challenges can reduce the reliability of both manual and automated segmentation, highlighting the need for more robust and boundary-aware approaches. Methods: In this study, an Aspect-Aware Boundary-Resilient UNet3D (ABR-UNet3D) architecture is proposed for cardiac MRI segmentation. The model incorporates an Aspect-Aware Complementary Attention (AAC) module that combines multi-planar contextual information with a complementary gating mechanism to enhance boundary representation. The method was evaluated on the ACDC dataset under consistent training conditions. In addition to Dice Similarity Coefficient (DSC) and Intersection over Union (IoU), boundary-based metrics, including the 95th percentile Hausdorff Distance (HD95), Average Surface Distance (ASD), and Surface Dice, were employed. Furthermore, a five-fold cross-validation protocol and detailed ablation studies were conducted to assess robustness and analyze the contribution of individual AAC components. Results: The proposed method achieved a mean DSC of 0.9603 in single-run experiments on the ACDC dataset and showed consistent performance in anatomically challenging regions, particularly for RV and MYO segmentation. In addition, five-fold cross-validation experiments resulted in an average DSC of 0.952 ± 0.009 and IoU of 0.908 ± 0.012, indicating stable performance across different data splits within the evaluated dataset. Boundary-based metrics also showed improved surface agreement and lower boundary errors compared with the evaluated baseline models. Ablation studies further indicated that the combined use of multi-planar contextual information and complementary gating contributes more effectively to segmentation performance than the individual components used separately. Conclusions: The results suggest that the proposed ABR-UNet3D architecture provides a stable and competitive segmentation framework for cardiac MRI images within the scope of the ACDC dataset. By jointly modeling contextual information and boundary refinement, the method improves segmentation reliability in challenging regions while maintaining competitive and consistent performance with respect to existing approaches.