Table 6 Effect of the proposed modules with the pretrained networks on the intracranial blood vessel dataset
From: DNL-Net: deformed non-local neural network for blood vessel segmentation
Method | Dice coefficient (%) | Mean IoU (%) |
|---|---|---|
ResNet50 + DNL + MFF | 94.72 | 90.65 |
ResNet50 + DNL + MFF + ASPP | 94.30 | 90.32 |
ResNet50 + DNL + MFF + RSEP | 93.83 | 88.99 |
DNL-Net | 96.63 | 92.93 |