Name
Abstract | ||
|---|---|---|
•A new curvilinear structure segmentation network is proposed based on dual self-attention modules, which can deal with both 2D and 3D imaging modalities in an unified manner.•Two self-attention mechanisms are employed in the channel and spatial spaces to generate attention-aware expressive features. They can enhance the network to capture long-range dependencies and make an effective use of the multi-channel space for feature representation and normalization, enabling the network to classify the curvilinear structure from background more effectively.•Experimental results on nine datasets (six 2D datasets and three 3D datasets) demonstrate that our proposed method achieves on the whole state-of-the-art performances in detecting curvilinear structures from different biomedical imaging modalities both quantitatively and qualitatively. |
Year | DOI | Venue |
|---|---|---|
2021 | 10.1016/j.media.2020.101874 | Medical Image Analysis |
Keywords | DocType | Volume |
Curvilinear structure,Blood vessel,Nerve fiber,Segmentation,Attention mechanism,Deep neural network | Journal | 67 |
ISSN | Citations | PageRank |
1361-8415 | 5 | 0.45 |
References | Authors | |
12 | 12 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Lei Mou | 1 | 16 | 2.64 |
| Yitian Zhao | 2 | 246 | 33.15 |
| Huazhu Fu | 3 | 1235 | 65.07 |
| Yonghuai Liu | 4 | 675 | 61.65 |
| Jun Cheng | 5 | 214 | 20.65 |
| Yalin Zheng | 6 | 264 | 34.69 |
| Pan Su | 7 | 8 | 0.84 |
| Jianlong Yang | 8 | 18 | 4.01 |
| Li Chen | 9 | 332 | 31.94 |
| Alejandro F. Frangi | 10 | 4333 | 309.21 |
| Masahiro Akiba | 11 | 12 | 2.04 |
| Jiang Liu | 12 | 335 | 34.30 |