niftynet.network.deepmedic module¶
-
class
DeepMedic(num_classes, w_initializer=None, w_regularizer=None, b_initializer=None, b_regularizer=None, acti_func='prelu', name='DeepMedic')[source]¶ Bases:
niftynet.network.base_net.BaseNet### Description reimplementation of DeepMedic:
Kamnitsas et al., “Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation”, MedIA ‘17### Building blocks [CONV] - 3x3x3 convolutional layer [denseCONV] - 1x1x1 convolutional layer
### Diagram INPUT –> CROP ——-> [CONV]x8 ——> [SUM] —-> [denseCONV]x3 –> OUTPUT
DOWNSAMPLE —> [CONV]x8 —> UPSAMPLE
### Constraints: - The downsampling factor (d_factor) should be odd - Label size = [(image_size / d_factor) - 16]* d_factor - Image size should be divisible by d_factor
# Examples: - Appropriate configuration for training: image spatial window size = 57, label spatial window size = 9, d_ factor = 3 - Appropriate configuration for inference: image spatial window size = 105, label spatial window size = 57, d_ factor = 3
-
__init__(num_classes, w_initializer=None, w_regularizer=None, b_initializer=None, b_regularizer=None, acti_func='prelu', name='DeepMedic')[source]¶ Parameters: - num_classes – int, number of channels of output
- w_initializer – weight initialisation for network
- w_regularizer – weight regularisation for network
- b_initializer – bias initialisation for network
- b_regularizer – bias regularisation for network
- acti_func – activation function to use
- name – layer name
-
