Research Article
Low-Grade Glioma Segmentation Based on CNN with Fully Connected CRF
Table 2
Base line and improved CNN structure configurations.
| CNN configuration | Base line | Base line + deeper | Base line + more fc | Base line + deeper + more fc |
| Input | Conv13–64 | Conv3–64 | Conv3–64 | Conv3–64 | Conv3–64 | Conv3–64 | Conv3–64 | Conv3–64 | | Conv3–64 | | Conv3–64 | Max-pool | Conv3–128 | Conv3–128 | Conv3–128 | Conv3–128 | Conv3–128 | Conv3–128 | Conv3–128 | Conv3–128 | | Conv3–128 | | Conv3–128 | Max-pool | FC2–256 | FC-4096 | FC-256 | FC-4096 | FC-2 | Softmax | Output |
|
|
1Conv: followed by the size of convolutional kernels and the number of filter banks. 2FC: followed by the number of fully connected layers.
|