181 use MONAI in pytorch medical segmentation program

docs/source/handlers.rst

@@ -20,7 +20,7 @@ Mean Dice metrics handler
 .. automodule:: monai.handlers.mean_dice
     :members:
 
-Metrics logger
+Metric logger
 ---------------
 .. automodule:: monai.handlers.metric_logger
     :members:

docs/source/index.rst

@@ -45,7 +45,7 @@ Technical highlights
    :maxdepth: 1
    :caption: APIs
 
-   transform_api
+   transforms
    losses
    networks
    metrics

docs/source/transform_api.rst → docs/source/transforms.rst

@@ -129,15 +129,21 @@ Vanilla Transforms
     :members:
     :special-members: __call__
 
-`IntensityNormalizer`
+`NormalizeIntensity`
 ~~~~~~~~~~~~~~~~~~~~~
-.. autoclass:: IntensityNormalizer
+.. autoclass:: NormalizeIntensity
     :members:
     :special-members: __call__
 
-`ImageEndPadder`
+`ScaleIntensityRange`
+~~~~~~~~~~~~~~~~~~~~~
+.. autoclass:: ScaleIntensityRange
+    :members:
+    :special-members: __call__
+
+`PadImageEnd`
 ~~~~~~~~~~~~~~~~
-.. autoclass:: ImageEndPadder
+.. autoclass:: PadImageEnd
     :members:
     :special-members: __call__
 
@@ -183,12 +189,30 @@ Vanilla Transforms
     :members:
     :special-members: __call__
 
+`Resample`
+~~~~~~~~~~~
+.. autoclass:: Resample
+    :members:
+    :special-members: __call__
+
 `RandAffine`
 ~~~~~~~~~~~~
 .. autoclass:: RandAffine
     :members:
     :special-members: __call__
 
+`RandDeformGrid`
+~~~~~~~~~~~~~~~~
+.. autoclass:: RandDeformGrid
+    :members:
+    :special-members: __call__
+
+`RandAffineGrid`
+~~~~~~~~~~~~~~~~
+.. autoclass:: RandAffineGrid
+    :members:
+    :special-members: __call__
+
 `Rand2DElastic`
 ~~~~~~~~~~~~~~~
 .. autoclass:: Rand2DElastic
@@ -262,6 +286,18 @@ Dictionary-based Composables
     :members:
     :special-members: __call__
 
+`NormalizeIntensityd`
+~~~~~~~~~~~~~~~~~~~~~
+.. autoclass:: NormalizeIntensityd
+    :members:
+    :special-members: __call__
+
+`ScaleIntensityRanged`
+~~~~~~~~~~~~~~~~~~~~~~
+.. autoclass:: ScaleIntensityRanged
+    :members:
+    :special-members: __call__
+
 `RandRotate90d`
 ~~~~~~~~~~~~~~~
 .. autoclass:: RandRotate90d

docs/source/utils.rst

@@ -14,7 +14,9 @@ Sliding window inference
 Module utils
 ------------
 .. automodule:: monai.utils.module
+  :members:
 
 Aliases
 -------
 .. automodule:: monai.utils.aliases
+  :members:

examples/classification_3d/densenet_training_array.py

@@ -149,7 +149,7 @@ def run_validation(engine):
 
 # create a training data loader
 train_ds = NiftiDataset(image_files=images[:10], labels=labels[:10], transform=train_transforms)
-train_loader = DataLoader(train_ds, batch_size=2, num_workers=2, pin_memory=torch.cuda.is_available())
+train_loader = DataLoader(train_ds, batch_size=2, shuffle=True, num_workers=2, pin_memory=torch.cuda.is_available())
 
 train_epochs = 30
 state = trainer.run(train_loader, train_epochs)

examples/classification_3d/densenet_training_dict.py

@@ -158,7 +158,7 @@ def run_validation(engine):
 
 # create a training data loader
 train_ds = monai.data.Dataset(data=train_files, transform=train_transforms)
-train_loader = DataLoader(train_ds, batch_size=2, num_workers=4, pin_memory=torch.cuda.is_available())
+train_loader = DataLoader(train_ds, batch_size=2, shuffle=True, num_workers=4, pin_memory=torch.cuda.is_available())
 
 train_epochs = 30
 state = trainer.run(train_loader, train_epochs)

examples/segmentation_3d/unet_training_array.py

@@ -82,12 +82,11 @@
 
 # create a training data loader
 train_ds = NiftiDataset(images[:20], segs[:20], transform=train_imtrans, seg_transform=train_segtrans)
-train_loader = DataLoader(train_ds, batch_size=5, num_workers=8, pin_memory=torch.cuda.is_available())
+train_loader = DataLoader(train_ds, batch_size=5, shuffle=True, num_workers=8, pin_memory=torch.cuda.is_available())
 # create a validation data loader
 val_ds = NiftiDataset(images[-20:], segs[-20:], transform=val_imtrans, seg_transform=val_segtrans)
 val_loader = DataLoader(val_ds, batch_size=5, num_workers=8, pin_memory=torch.cuda.is_available())
 
-
 # Create UNet, DiceLoss and Adam optimizer
 net = monai.networks.nets.UNet(
     dimensions=3,
@@ -102,7 +101,6 @@
 opt = torch.optim.Adam(net.parameters(), lr)
 device = torch.device("cuda:0")
 
-
 # ignite trainer expects batch=(img, seg) and returns output=loss at every iteration,
 # user can add output_transform to return other values, like: y_pred, y, etc.
 trainer = create_supervised_trainer(net, opt, loss, device, False)
@@ -123,7 +121,6 @@
 train_tensorboard_stats_handler = TensorBoardStatsHandler()
 train_tensorboard_stats_handler.attach(trainer)
 
-
 validation_every_n_epochs = 1
 # Set parameters for validation
 metric_name = 'Mean_Dice'
@@ -168,6 +165,5 @@ def run_validation(engine):
 )
 evaluator.add_event_handler(event_name=Events.EPOCH_COMPLETED, handler=val_tensorboard_image_handler)
 
-
 train_epochs = 30
 state = trainer.run(train_loader, train_epochs)

examples/segmentation_3d/unet_training_dict.py

@@ -78,18 +78,16 @@
 check_data = monai.utils.misc.first(check_loader)
 print(check_data['img'].shape, check_data['seg'].shape)
 
-
 # create a training data loader
 train_ds = monai.data.Dataset(data=train_files, transform=train_transforms)
 # use batch_size=2 to load images and use RandCropByPosNegLabeld to generate 2 x 4 images for network training
-train_loader = DataLoader(train_ds, batch_size=2, num_workers=4, collate_fn=list_data_collate,
-                          pin_memory=torch.cuda.is_available())
+train_loader = DataLoader(train_ds, batch_size=2, shuffle=True, num_workers=4,
+                          collate_fn=list_data_collate, pin_memory=torch.cuda.is_available())
 # create a validation data loader
 val_ds = monai.data.Dataset(data=val_files, transform=val_transforms)
 val_loader = DataLoader(val_ds, batch_size=5, num_workers=8, collate_fn=list_data_collate,
                         pin_memory=torch.cuda.is_available())
 
-
 # Create UNet, DiceLoss and Adam optimizer
 net = monai.networks.nets.UNet(
     dimensions=3,
@@ -104,7 +102,6 @@
 opt = torch.optim.Adam(net.parameters(), lr)
 device = torch.device("cuda:0")
 
-
 # ignite trainer expects batch=(img, seg) and returns output=loss at every iteration,
 # user can add output_transform to return other values, like: y_pred, y, etc.
 def prepare_batch(batch, device=None, non_blocking=False):
@@ -129,7 +126,6 @@ def prepare_batch(batch, device=None, non_blocking=False):
 train_tensorboard_stats_handler = TensorBoardStatsHandler()
 train_tensorboard_stats_handler.attach(trainer)
 
-
 validation_every_n_iters = 5
 # Set parameters for validation
 metric_name = 'Mean_Dice'
@@ -175,6 +171,5 @@ def run_validation(engine):
 evaluator.add_event_handler(
     event_name=Events.ITERATION_COMPLETED(every=2), handler=val_tensorboard_image_handler)
 
-
 train_epochs = 5
 state = trainer.run(train_loader, train_epochs)

examples/unet_segmentation_3d.ipynb

@@ -496,7 +496,7 @@
     "logging.basicConfig(stream=sys.stdout, level=logging.INFO)\n",
     "\n",
     "train_ds = NiftiDataset(images[:20], segs[:20], transform=imtrans, seg_transform=segtrans)\n",
-    "train_loader = DataLoader(train_ds, batch_size=5, num_workers=8, pin_memory=torch.cuda.is_available())\n",
+    "train_loader = DataLoader(train_ds, batch_size=5, shuffle=True, num_workers=8, pin_memory=torch.cuda.is_available())\n",
     "\n",
     "train_epochs = 30\n",
     "state = trainer.run(train_loader, train_epochs)"
@@ -571,7 +571,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.4"
+   "version": "3.6.9"
   }
  },
  "nbformat": 4,

monai/transforms/composables.py

@@ -24,7 +24,8 @@
 from monai.transforms.transforms import (LoadNifti, AsChannelFirst, Orientation,
                                          AddChannel, Spacing, Rotate90, SpatialCrop,
                                          RandAffine, Rand2DElastic, Rand3DElastic,
-                                         Rescale, Resize, Flip, Rotate, Zoom)
+                                         Rescale, Resize, Flip, Rotate, Zoom,
+                                         NormalizeIntensity, ScaleIntensityRange)
 from monai.utils.misc import ensure_tuple
 from monai.transforms.utils import generate_pos_neg_label_crop_centers, create_grid
 from monai.utils.aliases import alias
@@ -117,7 +118,7 @@ class LoadNiftid(MapTransform):
     dictionary-based wrapper of LoadNifti, must load image and metadata together.
     """
 
-    def __init__(self, keys, as_closest_canonical=False, dtype=None, meta_key_format='{}.{}', overwriting_keys=False):
+    def __init__(self, keys, as_closest_canonical=False, dtype=np.float32, meta_key_format='{}.{}', overwriting_keys=False):
         """
         Args:
             keys (hashable items): keys of the corresponding items to be transformed.
@@ -352,6 +353,57 @@ def __call__(self, data):
         return d
 
 
+@export
+@alias('NormalizeIntensityD', 'NormalizeIntensityDict')
+class NormalizeIntensityd(MapTransform):
+    """
+    dictionary-based wrapper of NormalizeIntensity.
+
+    Args:
+        keys (hashable items): keys of the corresponding items to be transformed.
+            See also: monai.transform.composables.MapTransform
+        subtrahend (ndarray): the amount to subtract by (usually the mean)
+        divisor (ndarray): the amount to divide by (usually the standard deviation)
+    """
+
+    def __init__(self, keys, subtrahend=None, divisor=None):
+        MapTransform.__init__(self, keys)
+        self.normalizer = NormalizeIntensity(subtrahend, divisor)
+
+    def __call__(self, data):
+        d = dict(data)
+        for key in self.keys:
+            d[key] = self.normalizer(d[key])
+        return d
+
+
+@export
+@alias('ScaleIntensityRangeD', 'ScaleIntensityRangeDict')
+class ScaleIntensityRanged(MapTransform):
+    """
+    dictionary-based wrapper of ScaleIntensityRange.
+
+    Args:
+        keys (hashable items): keys of the corresponding items to be transformed.
+            See also: monai.transform.composables.MapTransform
+        a_min (int or float): intensity original range min.
+        a_max (int or float): intensity original range max.
+        b_min (int or float): intensity target range min.
+        b_max (int or float): intensity target range max.
+        clip (bool): whether to perform clip after scaling.
+    """
+
+    def __init__(self, keys, a_min, a_max, b_min, b_max, clip=False):
+        MapTransform.__init__(self, keys)
+        self.scaler = ScaleIntensityRange(a_min, a_max, b_min, b_max, clip)
+
+    def __call__(self, data):
+        d = dict(data)
+        for key in self.keys:
+            d[key] = self.scaler(d[key])
+        return d
+
+
 @export
 @alias('RandCropByPosNegLabelD', 'RandCropByPosNegLabelDict')
 class RandCropByPosNegLabeld(Randomizable, MapTransform):
@@ -369,9 +421,13 @@ class RandCropByPosNegLabeld(Randomizable, MapTransform):
         neg (int, float): used to calculate the ratio ``pos / (pos + neg)`` for the probability to pick a
           foreground voxel as a center rather than a background voxel.
         num_samples (int): number of samples (crop regions) to take in each list.
+        image_key (str): if image_key is not None, use ``label == 0 & image > image_threshold`` to select
+            the negative sample(background) center. so the crop center will only exist on valid image area.
+        image_threshold (int or float): if enabled image_key, use ``image > image_threshold`` to determine
+            the valid image content area.
     """
 
-    def __init__(self, keys, label_key, size, pos=1, neg=1, num_samples=1):
+    def __init__(self, keys, label_key, size, pos=1, neg=1, num_samples=1, image_key=None, image_threshold=0):
         MapTransform.__init__(self, keys)
         assert isinstance(label_key, str), 'label_key must be a string.'
         assert isinstance(size, (list, tuple)), 'size must be list or tuple.'
@@ -385,15 +441,19 @@ def __init__(self, keys, label_key, size, pos=1, neg=1, num_samples=1):
         self.size = size
         self.pos_ratio = float(pos) / (float(pos) + float(neg))
         self.num_samples = num_samples
+        self.image_key = image_key
+        self.image_threshold = image_threshold
         self.centers = None
 
-    def randomize(self, label):
-        self.centers = generate_pos_neg_label_crop_centers(label, self.size, self.num_samples, self.pos_ratio, self.R)
+    def randomize(self, label, image):
+        self.centers = generate_pos_neg_label_crop_centers(label, self.size, self.num_samples, self.pos_ratio,
+                                                           image, self.image_threshold, self.R)
 
     def __call__(self, data):
         d = dict(data)
         label = d[self.label_key]
-        self.randomize(label)
+        image = d[self.image_key] if self.image_key else None
+        self.randomize(label, image)
         results = [dict() for _ in range(self.num_samples)]
         for key in data.keys():
             if key in self.keys:

monai/transforms/compose.py

@@ -71,8 +71,6 @@ def set_random_state(self, seed=None, state=None):
 
         Returns:
             a Randomizable instance.
-        Note:
-            thread safety
         """
         if seed is not None:
             _seed = id(seed) if not isinstance(seed, int) else seed