Add more histopathology configs

CHANGELOG.md

@@ -34,7 +34,7 @@ jobs that run in AzureML.
 -([#603](https://github.com/microsoft/InnerEye-DeepLearning/pull/603)) Add histopathology module
 -([#614](https://github.com/microsoft/InnerEye-DeepLearning/pull/614)) Checkpoint downloading falls back to looking into AzureML if no checkpoints on disk
 -([#613](https://github.com/microsoft/InnerEye-DeepLearning/pull/613)) Add additional tests for histopathology datasets
-
+-([#616](https://github.com/microsoft/InnerEye-DeepLearning/pull/616)) Add more histopathology configs and tests
 
 ### Changed
 - ([#588](https://github.com/microsoft/InnerEye-DeepLearning/pull/588)) Replace SciPy with PIL.PngImagePlugin.PngImageFile to load png files.

InnerEye/ML/Histopathology/datamodules/base_module.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import pickle
 from enum import Enum
 from pathlib import Path

InnerEye/ML/Histopathology/datamodules/panda_module.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from typing import Tuple
 
 from InnerEye.ML.Histopathology.datamodules.base_module import TilesDataModule

InnerEye/ML/Histopathology/datamodules/tcga_crck_module.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from typing import Tuple, Any
 
 from InnerEye.ML.Histopathology.datamodules.base_module import TilesDataModule

InnerEye/ML/Histopathology/datasets/base_dataset.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Any, Dict, Optional, Union
 

InnerEye/ML/Histopathology/datasets/default_paths.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 PANDA_TILES_DATASET_ID = "PANDA_tiles"
 TCGA_CRCK_DATASET_ID = "TCGA-CRCk"
 TCGA_PRAD_DATASET_ID = "TCGA-PRAD"

InnerEye/ML/Histopathology/datasets/panda_dataset.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Any, Dict, Union, Optional
 

InnerEye/ML/Histopathology/datasets/panda_tiles_dataset.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Any, Callable, Optional, Tuple, Union
 

InnerEye/ML/Histopathology/datasets/tcga_crck_tiles_dataset.py

@@ -1,7 +1,13 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Any, Callable, Optional, Tuple, Union
 
 import pandas as pd
+
 from torchvision.datasets.vision import VisionDataset
 
 from InnerEye.ML.Histopathology.datasets.base_dataset import TilesDataset

InnerEye/ML/Histopathology/datasets/tcga_prad_dataset.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Any, Dict, Optional, Union
 

InnerEye/ML/Histopathology/models/deepmil.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 import pandas as pd
 import numpy as np

InnerEye/ML/Histopathology/models/encoders.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Callable, Optional, Sequence, Tuple
 

InnerEye/ML/Histopathology/models/transforms.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from pathlib import Path
 from typing import Mapping, Sequence, Union
 

InnerEye/ML/Histopathology/preprocessing/create_tiles_dataset.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import functools
 import os
 import logging

InnerEye/ML/Histopathology/preprocessing/tiling.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 # These tiling implementations are adapted from PANDA Kaggle solutions, for example:
 # https://github.com/kentaroy47/Kaggle-PANDA-1st-place-solution/blob/master/src/data_process/a00_save_tiles.py
 from typing import Any, Optional, Tuple
@@ -25,7 +30,7 @@ def pad_for_tiling_2d(array: np.ndarray, tile_size: int, channels_first: Optiona
         original array to obtain indices for the padded array.
     """
     height, width = array.shape[1:] if channels_first else array.shape[:-1]
-    padding_h = get_1d_padding(height, tile_size) 
+    padding_h = get_1d_padding(height, tile_size)
     padding_w = get_1d_padding(width, tile_size)
     padding = [padding_h, padding_w]
     channels_axis = 0 if channels_first else 2

InnerEye/ML/Histopathology/scripts/aggregate_metrics_crossvalidation.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 """
 Script to find mean and standard deviation of desired metrics from cross validation child runs.
 """

InnerEye/ML/Histopathology/scripts/azure/azure_tiles_creation.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 """
 This script is an example of how to use the submit_to_azure_if_needed function from the hi-ml package to run the
 main pre-processing function that creates tiles from slides in the PANDA dataset. The advantage of using this script

InnerEye/ML/Histopathology/scripts/mount_azure_dataset.py

@@ -2,6 +2,7 @@
 #  Copyright (c) Microsoft Corporation. All rights reserved.
 #  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
 #  ------------------------------------------------------------------------------------------
+
 from health_azure import DatasetConfig
 from health_azure.utils import get_workspace
 

InnerEye/ML/Histopathology/utils/analysis_plot_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import numpy as np
 from typing import List, Any
 
@@ -85,7 +90,7 @@ def plot_box_whisker(data_list: List[Any], column_names: List[str], show_outlier
 def plot_histogram(data: List[Any], title: str = "") -> None:
     """
     Plot a histogram given some data
-    :param data: data to be plotted 
+    :param data: data to be plotted
     :param title: plot title string
     """
     plt.figure()

InnerEye/ML/Histopathology/utils/download_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import os
 from pathlib import Path
 

InnerEye/ML/Histopathology/utils/layer_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from typing import Callable, Tuple
 
 from torch import as_tensor, device, nn, prod, rand
@@ -25,7 +30,7 @@ def setup_feature_extractor(pretrained_model: nn.Module,
 def load_weights_to_model(weights_url: str, model: nn.Module) -> nn.Module:
     """
     Load weights to the histoSSL model from the given URL
-    https://github.com/ozanciga/self-supervised-histopathology 
+    https://github.com/ozanciga/self-supervised-histopathology
     """
     map_location = device('cpu')
     state = load_state_dict_from_url(weights_url, map_location=map_location)

InnerEye/ML/Histopathology/utils/metrics_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from typing import Tuple, List, Any, Dict
 import torch
 import matplotlib.pyplot as plt

InnerEye/ML/Histopathology/utils/naming.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 from enum import Enum
 
 class ResultsKey(str, Enum):

InnerEye/ML/Histopathology/utils/tcga_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import pandas as pd
 
 

InnerEye/ML/Histopathology/utils/viz_utils.py

@@ -1,3 +1,8 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
 import math
 import matplotlib.pyplot as plt
 

InnerEye/ML/configs/histo_configs/classification/BaseMIL.py

@@ -0,0 +1,112 @@
+#  ------------------------------------------------------------------------------------------
+#  Copyright (c) Microsoft Corporation. All rights reserved.
+#  Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
+#  ------------------------------------------------------------------------------------------
+
+"""BaseMIL is an abstract container defining basic functionality for running MIL experiments.
+It is responsible for instantiating the encoder and full DeepMIL model. Subclasses should define
+their datamodules and configure experiment-specific parameters.
+"""
+import os
+from pathlib import Path
+from typing import Type
+
+import param
+from torch import nn
+from torchvision.models.resnet import resnet18
+
+from health_azure.utils import CheckpointDownloader, get_workspace
+from health_ml.networks.layers.attention_layers import AttentionLayer, GatedAttentionLayer
+from InnerEye.Common import fixed_paths
+from InnerEye.ML.lightning_container import LightningContainer
+from InnerEye.ML.Histopathology.datamodules.base_module import CacheMode, TilesDataModule
+from InnerEye.ML.Histopathology.models.deepmil import DeepMILModule
+from InnerEye.ML.Histopathology.models.encoders import (HistoSSLEncoder, IdentityEncoder,
+                                                        ImageNetEncoder, ImageNetSimCLREncoder,
+                                                        InnerEyeSSLEncoder, TileEncoder)
+
+
+class BaseMIL(LightningContainer):
+    # Model parameters:
+    pooling_type: str = param.String(doc="Name of the pooling layer class to use.")
+    # l_rate, weight_decay, adam_betas are already declared in OptimizerParams superclass
+
+    # Encoder parameters:
+    encoder_type: str = param.String(doc="Name of the encoder class to use.")
+    tile_size: int = param.Integer(224, bounds=(1, None), doc="Tile width/height, in pixels.")
+    n_channels: int = param.Integer(3, bounds=(1, None), doc="Number of channels in the tile.")
+
+    # Data module parameters:
+    batch_size: int = param.Integer(16, bounds=(1, None), doc="Number of slides to load per batch.")
+    max_bag_size: int = param.Integer(1000, bounds=(0, None),
+                                      doc="Upper bound on number of tiles in each loaded bag. "
+                                          "If 0 (default), will return all samples in each bag. "
+                                          "If > 0, bags larger than `max_bag_size` will yield "
+                                          "random subsets of instances.")
+    cache_mode: CacheMode = param.ClassSelector(default=CacheMode.MEMORY, class_=CacheMode,
+                                                doc="The type of caching to perform: "
+                                                    "'memory' (default), 'disk', or 'none'.")
+    save_precache: bool = param.Boolean(True, doc="Whether to pre-cache the entire transformed "
+                                                  "dataset upfront and save it to disk.")
+    # local_dataset (used as data module root_path) is declared in DatasetParams superclass
+
+    @property
+    def cache_dir(self) -> Path:
+        raise NotImplementedError
+
+    def setup(self) -> None:
+        if self.encoder_type == InnerEyeSSLEncoder.__name__:
+            self.downloader = CheckpointDownloader(
+                aml_workspace=get_workspace(),
+                run_id="updated_transforms:updated_transforms_1636471522_5473e3ff",
+                checkpoint_filename="best_checkpoint.ckpt",
+                download_dir='outputs/'
+            )
+            os.chdir(fixed_paths.repository_root_directory())
+            self.downloader.download_checkpoint_if_necessary()
+
+        self.encoder = self.get_encoder()
+        self.encoder.cuda()
+        self.encoder.eval()
+
+    def get_encoder(self) -> TileEncoder:
+        if self.encoder_type == ImageNetEncoder.__name__:
+            return ImageNetEncoder(feature_extraction_model=resnet18,
+                                   tile_size=self.tile_size, n_channels=self.n_channels)
+
+        elif self.encoder_type == ImageNetSimCLREncoder.__name__:
+            return ImageNetSimCLREncoder(tile_size=self.tile_size, n_channels=self.n_channels)
+
+        elif self.encoder_type == HistoSSLEncoder.__name__:
+            return HistoSSLEncoder(tile_size=self.tile_size, n_channels=self.n_channels)
+
+        elif self.encoder_type == InnerEyeSSLEncoder.__name__:
+            return InnerEyeSSLEncoder(pl_checkpoint_path=self.downloader.local_checkpoint_path,
+                                      tile_size=self.tile_size, n_channels=self.n_channels)
+
+        else:
+            raise ValueError(f"Unsupported encoder type: {self.encoder_type}")
+
+    def get_pooling_layer(self) -> Type[nn.Module]:
+        if self.pooling_type == AttentionLayer.__name__:
+            return AttentionLayer
+        elif self.pooling_type == GatedAttentionLayer.__name__:
+            return GatedAttentionLayer
+        else:
+            raise ValueError(f"Unsupported pooling type: {self.pooling_type}")
+
+    def create_model(self) -> DeepMILModule:
+        self.data_module = self.get_data_module()
+        # Encoding is done in the datamodule, so here we provide instead a dummy
+        # no-op IdentityEncoder to be used inside the model
+        return DeepMILModule(encoder=IdentityEncoder(input_dim=(self.encoder.num_encoding,)),
+                             label_column=self.data_module.train_dataset.LABEL_COLUMN,
+                             n_classes=self.data_module.train_dataset.N_CLASSES,
+                             pooling_layer=self.get_pooling_layer(),
+                             class_weights=self.data_module.class_weights,
+                             l_rate=self.l_rate,
+                             weight_decay=self.weight_decay,
+                             adam_betas=self.adam_betas)
+
+    def get_data_module(self) -> TilesDataModule:
+        raise NotImplementedError