Pretraining Model

Labelprop Index / Labelprop / Pretraining Model

Auto-generated documentation for labelprop.Pretraining_model module.

LabelProp

Show source in Pretraining_model.py:12

Signature

class LabelProp(pl.LightningModule):
    def __init__(
        self,
        n_channels=1,
        n_classes=2,
        learning_rate=0.001,
        weight_decay=1e-08,
        way="up",
        shape=256,
        selected_slices=None,
        losses={},
        by_composition=False,
    ): ...

LabelProp().apply_deform

Show source in Pretraining_model.py:46

Apply deformation to x from flow field

Arguments

  • x Tensor - Image or mask to deform (BxCxHxW)
  • field Tensor - Deformation field (Bx2xHxW)

Returns

  • Tensor - Transformed image

Signature

def apply_deform(self, x, field, ismask=False): ...

LabelProp().apply_successive_transformations

Show source in Pretraining_model.py:77

Arguments

  • moving Tensor - Moving image (BxCxHxW)
  • flows [Tensor] - List of deformation fields (Bx2xHxW)

Returns

  • Tensor - Transformed image

Signature

def apply_successive_transformations(self, moving, flows): ...

LabelProp().automatic_optimization

Show source in Pretraining_model.py:14

Signature

@property
def automatic_optimization(self): ...

LabelProp().blend

Show source in Pretraining_model.py:136

Signature

def blend(self, x, y): ...

LabelProp().compose_deformation

Show source in Pretraining_model.py:66

Returns flow_k_j(flow_i_k(.)) flow

Arguments

flow_i_k flow_k_j

Returns

  • [Tensor] - Flow field flow_i_j = flow_k_j(flow_i_k(.))

Signature

def compose_deformation(self, flow_i_k, flow_k_j): ...

LabelProp().compose_list

Show source in Pretraining_model.py:59

Signature

def compose_list(self, flows): ...

LabelProp().compute_loss

Show source in Pretraining_model.py:116

Arguments

moved : Transformed anatomical image target : Target anatomical image moved_mask : Transformed mask target_mask : Target mask field : Deformation field

Signature

def compute_loss(
    self, moved=None, target=None, moved_mask=None, target_mask=None, field=None
): ...

LabelProp().configure_optimizers

Show source in Pretraining_model.py:262

Signature

def configure_optimizers(self): ...

LabelProp().forward

Show source in Pretraining_model.py:103

Arguments

  • moving Tensor - Moving image (BxCxHxW)
  • target [type] - Fixed image (BxCxHxW)
  • registration bool, optional - If False, also return non-integrated inverse flow field. Else return the integrated one. Defaults to False.

Returns

  • moved Tensor - Moved image
  • field Tensor - Deformation field from moving to target

Signature

def forward(self, moving, target, registration=True): ...

LabelProp().hardmax

Show source in Pretraining_model.py:265

Signature

def hardmax(self, Y, dim): ...

LabelProp().multi_class_dice

Show source in Pretraining_model.py:90

Arguments

  • pred_with_logits - Predicted mask with logits
  • target - Target mask

Signature

def multi_class_dice(self, pred_with_logits, target): ...

LabelProp().norm

Show source in Pretraining_model.py:17

Signature

def norm(self, x): ...

LabelProp().register_images

Show source in Pretraining_model.py:258

Signature

def register_images(self, moving, target, moving_mask): ...

LabelProp().training_step

Show source in Pretraining_model.py:142

Signature

def training_step(self, batch, batch_nb): ...