Name
Title | ||
|---|---|---|
U-Shaped Densely Connected Convolutional Networks For Automatic 3d Cardiovascular Mr Segmentation |
Abstract | ||
|---|---|---|
Amounts of experiments have verified the U-Net and DenseNet have strong power in visual object recognition, such as classification, regression, localization and so on. We here present an ingenious U-shaped densely connected convolutional networks that absorb the main advantages of U-Net and DenseNet. As a consequence, our proposed network has four outstanding advantages. First, this is a U-shaped framework on the whole, which allows the network to propagate context information to high resolution layers, and also a fully convolutional network, hence alleviate the network training. Second, it avoids learning redundant feature maps by adding DenseBlock before most convolutions in the network, thus the fewer parameters are needed to get a better outcome. Third, it achieves stable performance and excellent output even with different initial configuration and parameters. Fourth, the network obtains impressive performance with small cardiovascular MR dataset, which is of crucial importance for medical image processing. We evaluate our proposed architecture on the HVSMR2016 dataset, and achieve accurate cardiovascular MR segmentaion results, indicating the effectiveness of the proposed network in cardiovascular MR segmentaion. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1109/ROBIO.2018.8664897 | 2018 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND BIOMIMETICS (ROBIO) |
Field | DocType | Citations |
Pattern recognition,Segmentation,Convolution,Image processing,Image segmentation,Control engineering,Artificial intelligence,Engineering,Cognitive neuroscience of visual object recognition | Conference | 0 |
PageRank | References | Authors |
0.34 | 0 | 5 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Chongyang Ran | 1 | 0 | 0.68 |
| Ping Liu | 2 | 17 | 10.30 |
| Yinling Qian | 3 | 9 | 4.90 |
| He Yucheng | 4 | 3 | 5.82 |
| Qiong Wang | 5 | 30 | 15.18 |