TY - GEN
T1 - Mssa-net
T2 - 18th IEEE International Symposium on Biomedical Imaging, ISBI 2021
AU - Xu, Meng
AU - Huang, Kuan
AU - Chen, Qiuxiao
AU - Qi, Xiaojun
N1 - Publisher Copyright: © 2021 IEEE.
PY - 2021/4/13
Y1 - 2021/4/13
N2 - Ultrasound imaging is one of the most commonly used diagnostic tools to detect and classify abnormalities of the women breast. Automatic ultrasound image segmentation provides radiologists a second opinion to increase diagnosis accuracy. Deep neural networks have recently been employed to achieve better image segmentation results than conventional approaches. In this paper, we propose a novel deep learning architecture, a Multi-Scale Self-Attention Network (MSSA-Net), which can be trained on small datasets to explore relationships between pixels to achieve better segmentation accuracy. Our MSSA-Net integrates rich local features and global contextual information at different scales and applies self-attention to multi-scale feature maps. We evaluate the proposed MSSA-Net on three public breast ultrasound datasets and compare its performance with six state-of-the-art deep neural network-based approaches in terms of five metrics. MSSA-Net achieves best overall segmentation results and improves the second best approach by 1.21% for Jaccard Index (JI) and 0.94% for Dice's Coefficient (DSC).
AB - Ultrasound imaging is one of the most commonly used diagnostic tools to detect and classify abnormalities of the women breast. Automatic ultrasound image segmentation provides radiologists a second opinion to increase diagnosis accuracy. Deep neural networks have recently been employed to achieve better image segmentation results than conventional approaches. In this paper, we propose a novel deep learning architecture, a Multi-Scale Self-Attention Network (MSSA-Net), which can be trained on small datasets to explore relationships between pixels to achieve better segmentation accuracy. Our MSSA-Net integrates rich local features and global contextual information at different scales and applies self-attention to multi-scale feature maps. We evaluate the proposed MSSA-Net on three public breast ultrasound datasets and compare its performance with six state-of-the-art deep neural network-based approaches in terms of five metrics. MSSA-Net achieves best overall segmentation results and improves the second best approach by 1.21% for Jaccard Index (JI) and 0.94% for Dice's Coefficient (DSC).
KW - Breast ultrasound image segmentation
KW - MSSA-Net
KW - Multi-scale self attention
UR - https://www.scopus.com/pages/publications/85107199069
U2 - 10.1109/ISBI48211.2021.9433899
DO - 10.1109/ISBI48211.2021.9433899
M3 - Conference contribution
T3 - Proceedings - International Symposium on Biomedical Imaging
SP - 827
EP - 831
BT - 2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021
PB - IEEE Computer Society
Y2 - 13 April 2021 through 16 April 2021
ER -