What question did this study set out to answer?

The aim is to enhance breast tumor segmentation in medical imaging through a novel U-Net architecture with attention mechanisms.

March 18, 2026Cognitive ComputationOpen Access

Hybrid Loss-Driven Attention Residual U-Net with Multiscale Cascaded CBAM for Breast Cancer Segmentation

View Full Paper

Ask AI

Bookmark

Authors

Demonstrates improved tumor segmentation in breast cancer using advanced AI techniques and combined loss functions.

The aim is to enhance breast tumor segmentation in medical imaging through a novel U-Net architecture with attention mechanisms.
Proposed a U-Net architecture with multi-cascade convolutional block attention module (MC-CBAM).
Integrated channel and spatial attention for improved feature refinement.
Used a combined loss function of Tversky loss and Dice-based loss.
Trained the model with specific hyperparameters and assessed it on ultrasound and histopathology datasets.
Achieved a Dice score of 80.54% and Jaccard of 67.43% on the Triple Negative Breast Cancer histopathology dataset.
On the UDIAT ultrasound dataset, the model reached a Dice score of 87.23% and Jaccard of 77.36%.
For the BUSI dataset, achieved Dice score of 83.96% and Jaccard of 73.25%.
Demonstrated robustness across different imaging modalities.

Zafar et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark