An Improved Attention Gate for U-net Based Semantic Segmentation of Breast Cancer

Authors

• Md Bipul Hossain, University of South Alabama
• Mohamed Abouhashem Salama, University of South Alabama
• Madeline Potter, University of South Alabama
• Mohamed Shaban, University of South Alabama

Files

Description

Breast cancer is the most common cancer in the United States, where breast cells grow abnormally. Early detection of breast cancer via regular mammography or ultrasound screening may improve the efficacy of the targeted therapeutic treatments and increase the survival rates. However, automated lesion segmentation remains challenging because of the low contrast of images, the presence of speckle noise, and the high variability in lesion appearance. In this paper, we proposed an Improved Attention Gate (for the U Net, which incorporated a parallel realization of the spatial and channel attention schemes in order to exploit the significant regions and channels in the feature maps in contrast to the CBAM 6 where both spatial and channel attention schemes were cascaded.

Publication Date

3-2026

Department

Electrical & Computer Engineering

Disciplines

Electrical and Computer Engineering | Medicine and Health Sciences

Recommended Citation

Hossain, Md Bipul; Salama, Mohamed Abouhashem; Potter, Madeline; and Shaban, Mohamed, "An Improved Attention Gate for U-net Based Semantic Segmentation of Breast Cancer" (2026). Shelby Hall Graduate Research Forum Posters. 63.
https://jagworks.southalabama.edu/southalabama-shgrf-posters/63