Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Research on Multi-task Segmentation Method for Breast Mammography Based on Improved U-Net
1Faculty of Applied Sciences, Macao Polytechnic University, China.
2School of Artificial Intelligence, Dongguan City University, China.
ABSTRACT
Breast cancer is a malignant tumor that poses a serious health threat to women. Early screening and intervention are effective methods for preventing and controlling breast cancer. In order to provide powerful auxiliary tools for doctors' diagnosis and treatment, this paper proposes a multi-task segmentation algorithm for breast cancer mammography based on deep learning. To address the problem of lesion under-detection, a Convolutional Block Attention Module (CBAM) is introduced into the U-Net backbone network to extract features of small targets in both channel and spatial dimensions. To improve the training effectiveness and generalization ability of the model, deepsupervision learning is incorporated to enhance the model's recognition capability. Experimental results demonstrate that compared to other mainstream algorithms, this approach achieves better segmentation performance, significantly improving upon the original algorithm, and is more suitable for segmenting breast cancer lesions.
Keywords: Breast cancer, Deep learning, Attention mechanism, Deep supervision learning.
Download PDF
1Faculty of Applied Sciences, Macao Polytechnic University, China.
2School of Artificial Intelligence, Dongguan City University, China.
ABSTRACT
Breast cancer is a malignant tumor that poses a serious health threat to women. Early screening and intervention are effective methods for preventing and controlling breast cancer. In order to provide powerful auxiliary tools for doctors' diagnosis and treatment, this paper proposes a multi-task segmentation algorithm for breast cancer mammography based on deep learning. To address the problem of lesion under-detection, a Convolutional Block Attention Module (CBAM) is introduced into the U-Net backbone network to extract features of small targets in both channel and spatial dimensions. To improve the training effectiveness and generalization ability of the model, deepsupervision learning is incorporated to enhance the model's recognition capability. Experimental results demonstrate that compared to other mainstream algorithms, this approach achieves better segmentation performance, significantly improving upon the original algorithm, and is more suitable for segmenting breast cancer lesions.
Keywords: Breast cancer, Deep learning, Attention mechanism, Deep supervision learning.
Download PDF