Research Article
Joint Dense Residual and Recurrent Attention Network for DCE-MRI Breast Tumor Segmentation
Table 2
Comparison of experimental results of the different prediction models on the test set. All values in this table are averages from 32 test cases.
| Method | Jaccard | DSC | SEN | SPE | PPV | HD |
| U-net [8] | 0.6348 | 0.7557 | 0.8018 | 0.9989 | 0.7688 | 33.2948 | Attention U-Net [22] | 0.6386 | 0.7594 | 0.8013 | 0.9990 | 0.7787 | 19.5235 | Residual U-Net [20] | 0.6322 | 0.7555 | 0.8220 | 0.9988 | 0.7526 | 25.1890 | Dense U-Net [21] | 0.6529 | 0.7667 | 0.8031 | 0.9991 | 0.7781 | 20.9375 | V-net [25] | 0.6360 | 0.7587 | 0.8184 | 0.9989 | 0.7559 | 43.3051 | Present model | 0.6649 | 0.7827 | 0.8296 | 0.9990 | 0.7928 | 17.0818 |
|
|
The best results for the segmentation metrics in the comparison experiment are shown in bold.
|