International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

An Advanced Dual Attention-based U-Net Using Breast Ultrasound Data for Image Segmentation

Authors :

Erfan Akbarnezhad Sany¹ Niloufar Asghari² Fatemeh Naserizadeh³ Seyyed Abed Hosseini⁴

1- Department of Computer Engineering, Ma.C., Islamic Azad University, Mashhad, Iran 2- Institute of Computer Science, University of Bonn, Bonn, Germany 3- Faculty of Electrical and Computer Engineering Malek Ashtar University of Technology Tehran, Iran 4- Department of Electrical Engineering, Ma.C., Islamic Azad University, Mashhad, Iran

Keywords :

Medical image segmentation،Breast cancer،Classification،Ultrasound

Abstract :

Breast cancer is a prevalent and life-threatening disease among women, requiring accurate diagnostic methods. This study proposes a deep learning-based approach for breast cancer detection using ultrasound images, combining multi-task learning and attention mechanisms to improve performance. The model simultaneously performs classification and segmentation, enabling precise tumor localization. It uses a VGG16-based encoder pre-trained on ImageNet, enhanced with convolutional and classification-based attention modules to refine feature representation and highlight diagnostic cues. Evaluated on a dataset of 780 annotated images from 600 patients, the model achieved a Dice score of 91.95% for segmentation, and classification metrics of 91.78% recall and 92.88% precision. These results demonstrate the effectiveness of integrating attention and multi-task learning for accurate and reliable breast cancer detection from ultrasound images.