274 add multi organ segmentation bundle

multi_organ_segmentation/configs/evaluate.yaml

@@ -0,0 +1,64 @@
+---
+validate#postprocessing:
+  _target_: Compose
+  transforms:
+  - _target_: Activationsd
+    keys: pred
+    softmax: true
+  - _target_: Invertd
+    keys:
+    - pred
+    - label
+    transform: "@validate#preprocessing"
+    orig_keys: image
+    meta_key_postfix: meta_dict
+    nearest_interp:
+    - false
+    - true
+    to_tensor: true
+  - _target_: AsDiscreted
+    keys:
+    - pred
+    - label
+    argmax:
+    - true
+    - false
+    to_onehot: 8
+  - _target_: CopyItemsd
+    keys: "pred"
+    times: 1
+    names: "pred_save"
+  - _target_: AsDiscreted
+    keys:
+    - pred_save
+    argmax:
+    - true
+  - _target_: SaveImaged
+    keys: pred_save
+    meta_keys: pred_meta_dict
+    output_dir: "@output_dir"
+    resample: false
+    squeeze_end_dims: true
+validate#dataset:
+    _target_: Dataset
+    data: "@val_datalist"
+    transform: "@validate#preprocessing"
+validate#handlers:
+- _target_: CheckpointLoader
+  load_path: "$@ckpt_dir + '/model.pt'"
+  load_dict:
+    model: "@network"
+- _target_: StatsHandler
+  iteration_log: false
+- _target_: MetricsSaver
+  save_dir: "@output_dir"
+  metrics:
+  - val_mean_dice
+  - val_acc
+  metric_details:
+  - val_mean_dice
+  batch_transform: "$monai.handlers.from_engine(['image_meta_dict'])"
+  summary_ops: "*"
+evaluating:
+- "$setattr(torch.backends.cudnn, 'benchmark', True)"
+- "$@validate#evaluator.run()"

multi_organ_segmentation/configs/inference.yaml

@@ -0,0 +1,116 @@
+---
+imports:
+- "$import glob"
+- "$import os"
+input_channels: 1
+output_classes: 8
+arch_ckpt_path: "$@bundle_root + '/models/search_code_18590.pt'"
+arch_ckpt: "$torch.load(@arch_ckpt_path, map_location=torch.device('cuda'))"
+bundle_root: "/work/"
+output_dir: "$@bundle_root + '/eval'"
+dataset_dir: "/data/"
+data_list_file_path: "$@bundle_root + '/configs/dataset_0.json'"
+datalist: "$monai.data.load_decathlon_datalist(@data_list_file_path, data_list_key='testing',
+  base_dir=@dataset_dir)"
+device: "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')"
+dints_space:
+  _target_: monai.networks.nets.TopologyInstance
+  channel_mul: 1
+  num_blocks: 12
+  num_depths: 4
+  use_downsample: true
+  arch_code:
+  - "$@arch_ckpt['arch_code_a']"
+  - "$@arch_ckpt['arch_code_c']"
+  device: "$torch.device('cuda')"
+network_def:
+  _target_: monai.networks.nets.DiNTS
+  dints_space: "@dints_space"
+  in_channels: "@input_channels"
+  num_classes: "@output_classes"
+  use_downsample: true
+  node_a: "$torch.from_numpy(@arch_ckpt['node_a'])"
+network: "$@network_def.to(@device)"
+preprocessing:
+  _target_: Compose
+  transforms:
+  - _target_: LoadImaged
+    keys: image
+  - _target_: EnsureChannelFirstd
+    keys: image
+  - _target_: Orientationd
+    keys: image
+    axcodes: RAS
+  - _target_: Spacingd
+    keys: image
+    pixdim:
+    - 1
+    - 1
+    - 1
+    mode: bilinear
+  - _target_: ScaleIntensityRanged
+    keys: image
+    a_min: -500
+    a_max: 500
+    b_min: 0
+    b_max: 1
+    clip: true
+  - _target_: EnsureTyped
+    keys: image
+dataset:
+  _target_: Dataset
+  data: "@datalist"
+  transform: "@preprocessing"
+dataloader:
+  _target_: DataLoader
+  dataset: "@dataset"
+  batch_size: 1
+  shuffle: false
+  num_workers: 4
+inferer:
+  _target_: SlidingWindowInferer
+  roi_size:
+  - 96
+  - 96
+  - 96
+  sw_batch_size: 4
+  overlap: 0.625
+postprocessing:
+  _target_: Compose
+  transforms:
+  - _target_: Activationsd
+    keys: pred
+    softmax: true
+  - _target_: Invertd
+    keys: pred
+    transform: "@preprocessing"
+    orig_keys: image
+    meta_key_postfix: meta_dict
+    nearest_interp: false
+    to_tensor: true
+  - _target_: AsDiscreted
+    keys: pred
+    argmax: true
+  - _target_: SaveImaged
+    keys: pred
+    meta_keys: pred_meta_dict
+    output_dir: "@output_dir"
+handlers:
+- _target_: CheckpointLoader
+  load_path: "$@bundle_root + '/models/model.pt'"
+  load_dict:
+    model: "@network"
+- _target_: StatsHandler
+  iteration_log: false
+evaluator:
+  _target_: SupervisedEvaluator
+  device: "@device"
+  val_data_loader: "@dataloader"
+  network: "@network"
+  inferer: "@inferer"
+  postprocessing: "@postprocessing"
+  val_handlers: "@handlers"
+  amp: true
+evaluating:
+- "$setattr(torch.backends.cudnn, 'benchmark', True)"
+- "[email protected]()"

multi_organ_segmentation/configs/logging.conf

@@ -0,0 +1,21 @@
+[loggers]
+keys=root
+
+[handlers]
+keys=consoleHandler
+
+[formatters]
+keys=fullFormatter
+
+[logger_root]
+level=INFO
+handlers=consoleHandler
+
+[handler_consoleHandler]
+class=StreamHandler
+level=INFO
+formatter=fullFormatter
+args=(sys.stdout,)
+
+[formatter_fullFormatter]
+format=%(asctime)s - %(name)s - %(levelname)s - %(message)s

multi_organ_segmentation/configs/metadata.json

@@ -0,0 +1,90 @@
+{
+    "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
+    "version": "0.0.1",
+    "changelog": {
+        "0.0.1": "initialize the model package structure"
+    },
+    "monai_version": "1.0.1",
+    "pytorch_version": "1.13.0",
+    "numpy_version": "1.21.2",
+    "optional_packages_version": {
+        "fire": "0.4.0",
+        "nibabel": "4.0.1",
+        "pytorch-ignite": "0.4.9"
+    },
+    "task": "Multi-organ segmentation in abdominal CT",
+    "description": "DiNTS architectures for volumetric (3D) segmentation of the abdominal from CT image",
+    "authors": "Chen Shen, Holger R. Roth, Kazunari Misawa, Kensaku Mori",
+    "copyright": "",
+    "data_source": "Aichi Cancer Center, Japan",
+    "data_type": "nibabel",
+    "image_classes": "single channel data, intensity scaled to [0, 1]",
+    "label_classes": "eight channels data, 1 is artery, 2 is portal vein, 3 is liver, 4 is spleen, 5 is stomach, 6 is gallbladder, 7 is pancreas, 0 is everything else",
+    "pred_classes": "8 channels OneHot data, 1 is artery, 2 is portal vein, 3 is liver, 4 is spleen, 5 is stomach, 6 is gallbladder, 7 is pancreas, 0 is background",
+    "eval_metrics": {
+        "mean_dice": 0.88
+    },
+    "references": [
+        "He, Y., Yang, D., Roth, H., Zhao, C. and Xu, D., 2021. Dints: Differentiable neural network topology search for 3d medical image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 5841-5850).",
+        "Roth, H., Shen C, Oda H., Sugino T., Oda M., Hayashi Y., Misawa K., Mori K., 2018. A multi-scale pyramid of 3D fully convolutional networks for abdominal multi-organ segmentation. International conference on medical image computing and computer-assisted intervention",
+        "Shen, C., Roth, H. R., Nath, V., Hayashi, Y., Oda, M., Misawa, K., Mori, K., 2022. Effective hyperparameter optimization with proxy data for multi-organ segmentation. In Medical Imaging 2022: Image Processing (Vol. 12032, pp. 200-206)"
+    ],
+    "network_data_format": {
+        "inputs": {
+            "image": {
+                "type": "image",
+                "format": "hounsfield",
+                "modality": "CT",
+                "num_channels": 1,
+                "spatial_shape": [
+                    96,
+                    96,
+                    96
+                ],
+                "dtype": "float64",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "image"
+                }
+            }
+        },
+        "outputs": {
+            "pred": {
+                "type": "image",
+                "format": "segmentation",
+                "num_channels": 8,
+                "spatial_shape": [
+                    96,
+                    96,
+                    96
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1,
+                    2,
+                    3,
+                    4,
+                    5,
+                    6,
+                    7
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "background",
+                    "1": "artery",
+                    "2": "portal vein",
+                    "3": "liver",
+                    "4": "spleen",
+                    "5": "stomach",
+                    "6": "gallbladder",
+                    "7": "pancreas"
+                }
+            }
+        }
+    }
+}

multi_organ_segmentation/configs/multi_gpu_train.yaml

@@ -0,0 +1,52 @@
+---
+device: "$torch.device(f'cuda:{dist.get_rank()}')"
+network:
+  _target_: torch.nn.parallel.DistributedDataParallel
+  module: "$@network_def.to(@device)"
+  find_unused_parameters: true
+  device_ids:
+  - "@device"
+optimizer#lr: "$0.025*dist.get_world_size()"
+lr_scheduler#step_size: "$80*dist.get_world_size()"
+train#handlers:
+  - _target_: LrScheduleHandler
+    lr_scheduler: "@lr_scheduler"
+    print_lr: true
+  - _target_: ValidationHandler
+    validator: "@validate#evaluator"
+    epoch_level: true
+    interval: "$10*dist.get_world_size()"
+  - _target_: StatsHandler
+    tag_name: train_loss
+    output_transform: "$monai.handlers.from_engine(['loss'], first=True)"
+  - _target_: TensorBoardStatsHandler
+    log_dir: "@output_dir"
+    tag_name: train_loss
+    output_transform: "$monai.handlers.from_engine(['loss'], first=True)"
+train#trainer#max_epochs: "$400*dist.get_world_size()"
+train#trainer#train_handlers: "$@train#handlers[: -2 if dist.get_rank() > 0 else None]"
+validate#evaluator#val_handlers: "$None if dist.get_rank() > 0 else @validate#handlers"
+training:
+- "$import torch.distributed as dist"
+- "$dist.init_process_group(backend='nccl')"
+- "$torch.cuda.set_device(@device)"
+- "$monai.utils.set_determinism(seed=123)"
+- "$setattr(torch.backends.cudnn, 'benchmark', True)"
+- "$@train#trainer.run()"
+- "$dist.destroy_process_group()"
+train_data_partition: "$monai.data.partition_dataset(data=@train_datalist, num_partitions=dist.get_world_size(),
+  shuffle=True, even_divisible=True,)[dist.get_rank()]"
+train#dataset:
+  _target_: CacheDataset
+  data: "@train_data_partition"
+  transform: "@train#preprocessing"
+  cache_rate: 0.3
+  num_workers: 4
+val_data_partition: "$monai.data.partition_dataset(data=@val_datalist, num_partitions=dist.get_world_size(),
+  shuffle=False, even_divisible=False,)[dist.get_rank()]"
+validate#dataset:
+  _target_: CacheDataset
+  data: "@val_data_partition"
+  transform: "@validate#preprocessing"
+  cache_rate: 0.3
+  num_workers: 4