Deeper U-Net with Pyramid Feature Maps for Breast Tumor Segmentation

Background: Breast cancer continues to be one of the primary causes of cancer-related deaths among women globally. Accurate segmentation of breast tumors plays a vital role in early diagnosis, effective treatment planning, and reducing the mortality rate. Methods: This paper proposed a U-Net-based architecture (DU-Net-PFM) that incorporates pyramid feature maps in the encoder and increased layers that improve the overall accuracy of breast tumor segmentation. Results: The DU-Net-PFM achieved a Dice 0.925, Jaccard 0.864, sensitivity 0.911, and specificity 0.992 on the BUSIS dataset, outperforming U-Net and several state-of-the-art models, demonstrating its robustness and reliability for breast tumor segmentation. Qualitative results showed improved delineation of tumor boundaries. Conclusion: The DU-Net-PFM model demonstrates its effectiveness in improving segmentation accuracy, offering a promising tool for clinical applications in breast cancer diagnosis and treatment. Future work includes exploring its potential across other imaging modalities.