From: DeepHipp: accurate segmentation of hippocampus using 3D dense-block based on attention mechanism
Algorithm Description of DeepHipp |
---|
Input: Training data with augmentation: \(\chi\) |
1: Initialization: learning rate=0.01, bach_size=16, kernel size=3, randomly initialize W, b; |
2: for \(epoch=1\); \(epoch \le {epoch}_{max}\); \(epoch++ do\); |
3:Â Â Â Â Â Â Â Â Compute \(Dice\left( {y_{i},~\overset{\sim }{y_{i}}} \right)\); |
4:Â Â Â Â Â Â Â Â Compute \(Loss = 1 - Dice\left( {y_{i},~\overset{\sim }{y_{i}}} \right)\); |
5:Â Â Â Â Â Â Â Â Compute \(Minimum\left( {Loss} \right) ;\) |
6:    Update the model parameter W, b based on \(\chi\); |
7: end for; |
Output: The segmentation results from Model |