Sürdürülebilir Kalkınma Amaçları

• 2

• BildiriAçık Erişim2026/r/adakent/10060
  Explainable Attention-Guided Swin Transformer Networks for Brain Tumor Segmentation from 3D MRI

  Ranjbarzadeh, Ramin; Keles, Ayse; Anari, Shokofeh; Zarbakhsh, Payam; Bendechache, Malika

  This research presents a two-stage cascaded 3D segmentation approach designed for accurate brain tumor delineation in multimodal MRI data. The design incorporates a Swin Transformer backbone alongside Atrous Spatial Pyramid Pooling (ASPP) and Squeeze-and-Excitation (SE) blocks during the coarse segmentation phase to extract comprehensive, multi-scale contextual data. The second stage enhances the preliminary output with a class-wise attention decoder, which highlights tumor subregions—specifically, edema, necrotic core, and enhancing tumor—by masking the original input with the coarse predictions. The model employs a weighted Tversky loss to address the substantial class imbalance present in tumor segmentation tasks. Extensive experiments were executed on the BraTS 2020 dataset, incorporating preprocessing processes such as volume cropping and scaling to 128 × 128 × 128, with evaluation completed across training, validation, and test sets divided in a 70%–15%–15% ratio. The model exhibited exceptional accuracy in identifying tumor regions, with a Dice score of 0.99 for healthy tissue and above 0.5 for tumor subregions. The investigation of ROC curves and confusion matrices further substantiated the reliability of the forecasts. To improve model transparency, layer-wise Grad-CAM heatmaps were produced, illustrating the transition of attention from general background in first levels to concentrated tumor localization in subsequent layers. The proposed method attains competitive segmentation performance while providing a more interpretable and therapeutically pertinent answer. The results underscore the efficacy of integrating coarse-to-fine cascaded architectures with transformer-based encoders and attention-driven refinement for brain tumor segmentation.

  • 
  İncele
• MakaleAçık Erişim2026/r/adakent/10057
  Explainable deep learning framework for brain tumor segmentation using vision transformer and conditional random fields

  Safarpour, Homayoun; Sadeghi, Soroush; Zarbakhsh, Payam; Kamsari, Mohammadreza; Kia, Marjan; Ranjbarzadeh, Ramin

  Brain tumor segmentation is critical for accurate identification and treatment planning. However, delineating intricate tumor areas like Edema, Tumor Core, and Enhancing Tumor remains challenging owing to diversity in dimensions, morphology, and delineation. This introduces a novel deep learning model integrating a pretrained Vision Transformer (ViT) for global context representation, Atrous Spatial Pyramid Pooling (ASPP) in both encoder and merging stages, and ReLU6. Additionally, attention mechanisms and Conditional Random Fields (CRF) were incorporated for refining feature emphasis and segmentation boundaries. The model was trained and tested on the BraTS 2020 dataset and compared with state-of-the-art methods. The suggested model obtained the highest Dice coefficients across all tumor regions, with improvements of up to 4% over nnU-Net and SwinBTS in the Enhancing Tumor and Edema region. Grad-CAM heatmaps illustrated the model’s proficiency in precisely identifying tumor areas, hence enhancing the interpretability of predictions. The incorporation of multi-scale feature extraction, attention mechanism, and global context representation markedly improved segmentation accuracy and interpretability. The findings underscore the proposed model’s capacity to enhance clinical processes and establish a standard for brain tumor segmentation tasks. Future endeavors involve testing on varied datasets and investigating lightweight structures for real-time applications.

  • 
  İncele