Low-Grade Glioma Segmentation Based on CNN with Fully Connected CRF

Algorithm 1: Proposed segmentation method.

Input: image , mean value , max value , standard deviation , parameters of settled multi-way CNN network and parameters of fully connected CRF model ()

Output: tumor segmentation result

1: Normalize using and and get

2: Enter into the settled multi-way CNN, and get the output from the softmax layers

3: Using as original unary potential

4: for  do

5: Compute fully connected CRF with and unary potential using parameters

6: Correct the segmentation results using several morphology methods

Algorithm 1: Proposed segmentation method.