complete the model package

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+models/model.ts filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,103 @@
+---
+tags:
+- monai
+- medical
+library_name: monai
+license: unknown
+---
+# Model Overview
+A pre-trained model for volumetric (3D) segmentation of brain tumor subregions from multimodal MRIs based on BraTS 2018 data. The whole pipeline is modified from [clara_pt_brain_mri_segmentation](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/med/models/clara_pt_brain_mri_segmentation).
+
+## Workflow
+
+The model is trained to segment 3 nested subregions of primary brain tumors (gliomas): the "enhancing tumor" (ET), the "tumor core" (TC), the "whole tumor" (WT) based on 4 aligned input MRI scans (T1c, T1, T2, FLAIR).
+- The ET is described by areas that show hyper intensity in T1c when compared to T1, but also when compared to "healthy" white matter in T1c.
+- The TC describes the bulk of the tumor, which is what is typically resected. The TC entails the ET, as well as the necrotic (fluid-filled) and the non-enhancing (solid) parts of the tumor.
+-  The WT describes the complete extent of the disease, as it entails the TC and the peritumoral edema (ED), which is typically depicted by hyper-intense signal in FLAIR.
+
+## Data
+
+The training data is from the [Multimodal Brain Tumor Segmentation Challenge (BraTS) 2018](https://www.med.upenn.edu/sbia/brats2018/data.html).
+
+- Target: 3 tumor subregions
+- Task: Segmentation
+- Modality: MRI
+- Size: 285 3D volumes (4 channels each)
+
+The provided labelled data was partitioned, based on our own split, into training (200 studies), validation (42 studies) and testing (43 studies) datasets.
+
+Please run `scripts/prepare_datalist.py` to produce the data list. The command is like:
+
+```
+python scripts/prepare_datalist.py --path your-brats18-dataset-path
+```
+
+## Training configuration
+
+This model utilized a similar approach described in 3D MRI brain tumor segmentation
+using autoencoder regularization, which was a winning method in BraTS2018 [1]. The training was performed with the following:
+
+- GPU: At least 16GB of GPU memory.
+- Actual Model Input: 224 x 224 x 144
+- AMP: True
+- Optimizer: Adam
+- Learning Rate: 1e-4
+- Loss: DiceLoss
+
+## Input
+
+Input: 4 channel MRI (4 aligned MRIs T1c, T1, T2, FLAIR at 1x1x1 mm)
+
+1. Normalizing to unit std with zero mean
+2. Randomly cropping to (224, 224, 144)
+3. Randomly spatial flipping
+4. Randomly scaling and shifting intensity of the volume
+
+## Output
+
+Output: 3 channels
+- Label 0: TC tumor subregion
+- Label 1: WT tumor subregion
+- Label 2: ET tumor subregion
+
+## Model Performance
+
+The achieved Dice scores on the validation data are:
+- Tumor core (TC): 0.8559
+- Whole tumor (WT): 0.9026
+- Enhancing tumor (ET): 0.7905
+- Average: 0.8518
+
+## commands example
+
+Execute training:
+
+```
+python -m monai.bundle run training --meta_file configs/metadata.json --config_file configs/train.json --logging_file configs/logging.conf
+```
+
+Override the `train` config to execute multi-GPU training:
+
+```
+torchrun --standalone --nnodes=1 --nproc_per_node=8 -m monai.bundle run training --meta_file configs/metadata.json --config_file "['configs/train.json','configs/multi_gpu_train.json']" --logging_file configs/logging.conf
+```
+
+Override the `train` config to execute evaluation with the trained model:
+
+```
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file "['configs/train.json','configs/evaluate.json']" --logging_file configs/logging.conf
+```
+
+Execute inference:
+
+```
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file configs/inference.json --logging_file configs/logging.conf
+```
+
+# Disclaimer
+
+This is an example, not to be used for diagnostic purposes.
+
+# References
+
+[1] Myronenko, Andriy. "3D MRI brain tumor segmentation using autoencoder regularization." International MICCAI Brainlesion Workshop. Springer, Cham, 2018. https://arxiv.org/abs/1810.11654.

configs/evaluate.json

@@ -0,0 +1,92 @@
+{
+    "validate#postprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "Activationsd",
+                "keys": "pred",
+                "sigmoid": true
+            },
+            {
+                "_target_": "Invertd",
+                "keys": "pred",
+                "transform": "@validate#preprocessing",
+                "orig_keys": "image",
+                "meta_keys": "pred_meta_dict",
+                "nearest_interp": false,
+                "to_tensor": true,
+                "device": "@validate#evaluator#device"
+            },
+            {
+                "_target_": "AsDiscreted",
+                "keys": "pred",
+                "threshold": 0.5
+            },
+            {
+                "_target_": "SplitChanneld",
+                "keys": [
+                    "pred",
+                    "label"
+                ],
+                "output_postfixes": [
+                    "tc",
+                    "wt",
+                    "et"
+                ]
+            },
+            {
+                "_target_": "CopyItemsd",
+                "keys": "pred",
+                "names": "pred_combined",
+                "times": 1
+            },
+            {
+                "_target_": "Lambdad",
+                "keys": "pred_combined",
+                "func": "$lambda x: torch.where(x[[2]] > 0, 4, torch.where(x[[0]] > 0, 1, torch.where(x[[1]] > 0, 2, 0)))"
+            },
+            {
+                "_target_": "SaveImaged",
+                "keys": "pred_combined",
+                "meta_keys": "pred_meta_dict",
+                "output_dir": "@output_dir",
+                "output_postfix": "seg",
+                "output_dtype": "uint8",
+                "resample": false,
+                "squeeze_end_dims": true
+            }
+        ]
+    },
+    "validate#handlers": [
+        {
+            "_target_": "CheckpointLoader",
+            "load_path": "$@ckpt_dir + '/model.pt'",
+            "load_dict": {
+                "model": "@network"
+            }
+        },
+        {
+            "_target_": "StatsHandler",
+            "iteration_log": false
+        },
+        {
+            "_target_": "MetricsSaver",
+            "save_dir": "@output_dir",
+            "metrics": [
+                "val_mean_dice",
+                "val_mean_dice_tc",
+                "val_mean_dice_wt",
+                "val_mean_dice_et"
+            ],
+            "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()"
+    ]
+}

configs/inference.json

@@ -0,0 +1,131 @@
+{
+    "imports": [
+        "$import glob",
+        "$import os"
+    ],
+    "bundle_root": "/workspace/brats_mri_segmentation",
+    "ckpt_dir": "$@bundle_root + '/models'",
+    "output_dir": "$@bundle_root + '/eval'",
+    "data_list_file_path": "$@bundle_root + '/configs/datalist.json'",
+    "data_file_base_dir": "/workspace/data/medical/brats2018challenge",
+    "test_datalist": "$monai.data.load_decathlon_datalist(@data_list_file_path, data_list_key='testing', base_dir=@data_file_base_dir)",
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
+    "amp": true,
+    "network_def": {
+        "_target_": "SegResNet",
+        "blocks_down": [
+            1,
+            2,
+            2,
+            4
+        ],
+        "blocks_up": [
+            1,
+            1,
+            1
+        ],
+        "init_filters": 16,
+        "in_channels": 4,
+        "out_channels": 3,
+        "dropout_prob": 0.2
+    },
+    "network": "$@network_def.to(@device)",
+    "preprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": "image"
+            },
+            {
+                "_target_": "NormalizeIntensityd",
+                "keys": "image",
+                "nonzero": true,
+                "channel_wise": true
+            },
+            {
+                "_target_": "ToTensord",
+                "keys": "image"
+            }
+        ]
+    },
+    "dataset": {
+        "_target_": "Dataset",
+        "data": "@test_datalist",
+        "transform": "@preprocessing"
+    },
+    "dataloader": {
+        "_target_": "DataLoader",
+        "dataset": "@dataset",
+        "batch_size": 1,
+        "shuffle": true,
+        "num_workers": 4
+    },
+    "inferer": {
+        "_target_": "SlidingWindowInferer",
+        "roi_size": [
+            240,
+            240,
+            160
+        ],
+        "sw_batch_size": 1,
+        "overlap": 0.5
+    },
+    "postprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "Activationsd",
+                "keys": "pred",
+                "sigmoid": true
+            },
+            {
+                "_target_": "Invertd",
+                "keys": "pred",
+                "transform": "@preprocessing",
+                "orig_keys": "image",
+                "meta_keys": "pred_meta_dict",
+                "nearest_interp": false,
+                "to_tensor": true
+            },
+            {
+                "_target_": "AsDiscreted",
+                "keys": "pred",
+                "threshold": 0.5
+            },
+            {
+                "_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)",
+        "$@evaluator.run()"
+    ]
+}

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

configs/metadata.json

@@ -0,0 +1,79 @@
+{
+    "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
+    "version": "0.1.0",
+    "changelog": {
+        "0.1.0": "complete the model package"
+    },
+    "monai_version": "0.9.0",
+    "pytorch_version": "1.12.0",
+    "numpy_version": "1.21.2",
+    "optional_packages_version": {
+        "nibabel": "3.2.1",
+        "pytorch-ignite": "0.4.8"
+    },
+    "task": "Multimodal Brain Tumor segmentation",
+    "description": "A pre-trained model for volumetric (3D) segmentation of brain tumor subregions from multimodal MRIs based on BraTS 2018 data",
+    "authors": "MONAI team",
+    "copyright": "Copyright (c) MONAI Consortium",
+    "data_source": "https://www.med.upenn.edu/sbia/brats2018/data.html",
+    "data_type": "nibabel",
+    "image_classes": "4 channel data, T1c, T1, T2, FLAIR at 1x1x1 mm",
+    "label_classes": "3 channel data, channel 0 for Tumor core, channel 1 for Whole tumor, channel 2 for Enhancing tumor",
+    "pred_classes": "3 channels data, same as label_classes",
+    "eval_metrics": {
+        "val_mean_dice": 0.8518,
+        "val_mean_dice_tc": 0.8559,
+        "val_mean_dice_wt": 0.9026,
+        "val_mean_dice_et": 0.7905
+    },
+    "intended_use": "This is an example, not to be used for diagnostic purposes",
+    "references": [
+        "Myronenko, Andriy. '3D MRI brain tumor segmentation using autoencoder regularization.' International MICCAI Brainlesion Workshop. Springer, Cham, 2018. https://arxiv.org/abs/1810.11654"
+    ],
+    "network_data_format": {
+        "inputs": {
+            "image": {
+                "type": "image",
+                "format": "magnitude",
+                "modality": "MR",
+                "num_channels": 4,
+                "spatial_shape": [
+                    "8*n",
+                    "8*n",
+                    "8*n"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "image"
+                }
+            }
+        },
+        "outputs": {
+            "pred": {
+                "type": "image",
+                "format": "segmentation",
+                "num_channels": 3,
+                "spatial_shape": [
+                    "8*n",
+                    "8*n",
+                    "8*n"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "background",
+                    "1": "spleen"
+                }
+            }
+        }
+    }
+}

configs/multi_gpu_train.json

@@ -0,0 +1,36 @@
+{
+    "device": "$torch.device(f'cuda:{dist.get_rank()}')",
+    "network": {
+        "_target_": "torch.nn.parallel.DistributedDataParallel",
+        "module": "$@network_def.to(@device)",
+        "device_ids": [
+            "@device"
+        ]
+    },
+    "train#sampler": {
+        "_target_": "DistributedSampler",
+        "dataset": "@train#dataset",
+        "even_divisible": true,
+        "shuffle": true
+    },
+    "train#dataloader#sampler": "@train#sampler",
+    "train#dataloader#shuffle": false,
+    "train#trainer#train_handlers": "$@train#handlers[: -2 if dist.get_rank() > 0 else None]",
+    "validate#sampler": {
+        "_target_": "DistributedSampler",
+        "dataset": "@validate#dataset",
+        "even_divisible": false,
+        "shuffle": false
+    },
+    "validate#dataloader#sampler": "@validate#sampler",
+    "validate#evaluator#val_handlers": "$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()"
+    ]
+}

configs/train.json

@@ -0,0 +1,335 @@
+{
+    "imports": [
+        "$import glob",
+        "$import os"
+    ],
+    "bundle_root": "/workspace/brats_mri_segmentation",
+    "ckpt_dir": "$@bundle_root + '/models'",
+    "output_dir": "$@bundle_root + '/eval'",
+    "data_list_file_path": "$@bundle_root + '/configs/datalist.json'",
+    "data_file_base_dir": "/workspace/data/medical/brats2018challenge",
+    "train_datalist": "$monai.data.load_decathlon_datalist(@data_list_file_path, data_list_key='training', base_dir=@data_file_base_dir)",
+    "val_datalist": "$monai.data.load_decathlon_datalist(@data_list_file_path, data_list_key='validation', base_dir=@data_file_base_dir)",
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
+    "epochs": 300,
+    "num_interval_per_valid": 1,
+    "learning_rate": 0.0001,
+    "amp": true,
+    "network_def": {
+        "_target_": "SegResNet",
+        "blocks_down": [
+            1,
+            2,
+            2,
+            4
+        ],
+        "blocks_up": [
+            1,
+            1,
+            1
+        ],
+        "init_filters": 16,
+        "in_channels": 4,
+        "out_channels": 3,
+        "dropout_prob": 0.2
+    },
+    "network": "$@network_def.to(@device)",
+    "loss": {
+        "_target_": "DiceLoss",
+        "smooth_nr": 0,
+        "smooth_dr": 1e-05,
+        "squared_pred": true,
+        "to_onehot_y": false,
+        "sigmoid": true
+    },
+    "optimizer": {
+        "_target_": "torch.optim.Adam",
+        "params": "$@network.parameters()",
+        "lr": "@learning_rate",
+        "weight_decay": 1e-05
+    },
+    "lr_scheduler": {
+        "_target_": "torch.optim.lr_scheduler.CosineAnnealingLR",
+        "optimizer": "@optimizer",
+        "T_max": "@epochs"
+    },
+    "train": {
+        "preprocessing_transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": [
+                    "image",
+                    "label"
+                ]
+            },
+            {
+                "_target_": "ConvertToMultiChannelBasedOnBratsClassesd",
+                "keys": "label"
+            },
+            {
+                "_target_": "NormalizeIntensityd",
+                "keys": "image",
+                "nonzero": true,
+                "channel_wise": true
+            }
+        ],
+        "random_transforms": [
+            {
+                "_target_": "RandSpatialCropd",
+                "keys": [
+                    "image",
+                    "label"
+                ],
+                "roi_size": [
+                    224,
+                    224,
+                    144
+                ],
+                "random_size": false
+            },
+            {
+                "_target_": "RandFlipd",
+                "keys": [
+                    "image",
+                    "label"
+                ],
+                "prob": 0.5,
+                "spatial_axis": 0
+            },
+            {
+                "_target_": "RandFlipd",
+                "keys": [
+                    "image",
+                    "label"
+                ],
+                "prob": 0.5,
+                "spatial_axis": 1
+            },
+            {
+                "_target_": "RandFlipd",
+                "keys": [
+                    "image",
+                    "label"
+                ],
+                "prob": 0.5,
+                "spatial_axis": 2
+            },
+            {
+                "_target_": "RandScaleIntensityd",
+                "keys": "image",
+                "factors": 0.1,
+                "prob": 1.0
+            },
+            {
+                "_target_": "RandShiftIntensityd",
+                "keys": "image",
+                "offsets": 0.1,
+                "prob": 1.0
+            }
+        ],
+        "final_transforms": [
+            {
+                "_target_": "ToTensord",
+                "keys": [
+                    "image",
+                    "label"
+                ]
+            }
+        ],
+        "preprocessing": {
+            "_target_": "Compose",
+            "transforms": "$@train#preprocessing_transforms + @train#random_transforms + @train#final_transforms"
+        },
+        "dataset": {
+            "_target_": "Dataset",
+            "data": "@train_datalist",
+            "transform": "@train#preprocessing"
+        },
+        "dataloader": {
+            "_target_": "DataLoader",
+            "dataset": "@train#dataset",
+            "batch_size": 1,
+            "shuffle": true,
+            "num_workers": 4
+        },
+        "inferer": {
+            "_target_": "SimpleInferer"
+        },
+        "postprocessing": {
+            "_target_": "Compose",
+            "transforms": [
+                {
+                    "_target_": "Activationsd",
+                    "keys": "pred",
+                    "sigmoid": true
+                },
+                {
+                    "_target_": "AsDiscreted",
+                    "keys": "pred",
+                    "threshold": 0.5
+                }
+            ]
+        },
+        "handlers": [
+            {
+                "_target_": "LrScheduleHandler",
+                "lr_scheduler": "@lr_scheduler",
+                "print_lr": true
+            },
+            {
+                "_target_": "ValidationHandler",
+                "validator": "@validate#evaluator",
+                "epoch_level": true,
+                "interval": "@num_interval_per_valid"
+            },
+            {
+                "_target_": "StatsHandler",
+                "tag_name": "train_loss",
+                "output_transform": "$monai.handlers.from_engine(['loss'], first=True)"
+            },
+            {
+                "_target_": "TensorBoardStatsHandler",
+                "log_dir": "@output_dir",
+                "tag_name": "train_loss",
+                "output_transform": "$monai.handlers.from_engine(['loss'], first=True)"
+            }
+        ],
+        "key_metric": {
+            "train_mean_dice": {
+                "_target_": "MeanDice",
+                "include_background": true,
+                "output_transform": "$monai.handlers.from_engine(['pred', 'label'])"
+            }
+        },
+        "trainer": {
+            "_target_": "SupervisedTrainer",
+            "max_epochs": "@epochs",
+            "device": "@device",
+            "train_data_loader": "@train#dataloader",
+            "network": "@network",
+            "loss_function": "@loss",
+            "optimizer": "@optimizer",
+            "inferer": "@train#inferer",
+            "postprocessing": "@train#postprocessing",
+            "key_train_metric": "@train#key_metric",
+            "train_handlers": "@train#handlers",
+            "amp": "@amp"
+        }
+    },
+    "validate": {
+        "preprocessing": {
+            "_target_": "Compose",
+            "transforms": "$@train#preprocessing_transforms + @train#final_transforms"
+        },
+        "dataset": {
+            "_target_": "Dataset",
+            "data": "@val_datalist",
+            "transform": "@validate#preprocessing"
+        },
+        "dataloader": {
+            "_target_": "DataLoader",
+            "dataset": "@validate#dataset",
+            "batch_size": 1,
+            "shuffle": false,
+            "num_workers": 4
+        },
+        "inferer": {
+            "_target_": "SlidingWindowInferer",
+            "roi_size": [
+                240,
+                240,
+                160
+            ],
+            "sw_batch_size": 1,
+            "overlap": 0.5
+        },
+        "postprocessing": {
+            "_target_": "Compose",
+            "transforms": [
+                {
+                    "_target_": "Activationsd",
+                    "keys": "pred",
+                    "sigmoid": true
+                },
+                {
+                    "_target_": "AsDiscreted",
+                    "keys": "pred",
+                    "threshold": 0.5
+                },
+                {
+                    "_target_": "SplitChanneld",
+                    "keys": [
+                        "pred",
+                        "label"
+                    ],
+                    "output_postfixes": [
+                        "tc",
+                        "wt",
+                        "et"
+                    ]
+                }
+            ]
+        },
+        "handlers": [
+            {
+                "_target_": "StatsHandler",
+                "iteration_log": false
+            },
+            {
+                "_target_": "TensorBoardStatsHandler",
+                "log_dir": "@output_dir",
+                "iteration_log": false
+            },
+            {
+                "_target_": "CheckpointSaver",
+                "save_dir": "@ckpt_dir",
+                "save_dict": {
+                    "model": "@network"
+                },
+                "save_key_metric": true,
+                "key_metric_filename": "model.pt"
+            }
+        ],
+        "key_metric": {
+            "val_mean_dice": {
+                "_target_": "MeanDice",
+                "include_background": true,
+                "output_transform": "$monai.handlers.from_engine(['pred', 'label'])"
+            }
+        },
+        "additional_metrics": {
+            "val_mean_dice_tc": {
+                "_target_": "MeanDice",
+                "include_background": true,
+                "output_transform": "$monai.handlers.from_engine(['pred_tc', 'label_tc'])"
+            },
+            "val_mean_dice_wt": {
+                "_target_": "MeanDice",
+                "include_background": true,
+                "output_transform": "$monai.handlers.from_engine(['pred_wt', 'label_wt'])"
+            },
+            "val_mean_dice_et": {
+                "_target_": "MeanDice",
+                "include_background": true,
+                "output_transform": "$monai.handlers.from_engine(['pred_et', 'label_et'])"
+            }
+        },
+        "evaluator": {
+            "_target_": "SupervisedEvaluator",
+            "device": "@device",
+            "val_data_loader": "@validate#dataloader",
+            "network": "@network",
+            "inferer": "@validate#inferer",
+            "postprocessing": "@validate#postprocessing",
+            "key_val_metric": "@validate#key_metric",
+            "additional_metrics": "@validate#additional_metrics",
+            "val_handlers": "@validate#handlers",
+            "amp": "@amp"
+        }
+    },
+    "training": [
+        "$monai.utils.set_determinism(seed=123)",
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@train#trainer.run()"
+    ]
+}

docs/README.md

@@ -0,0 +1,96 @@
+# Model Overview
+A pre-trained model for volumetric (3D) segmentation of brain tumor subregions from multimodal MRIs based on BraTS 2018 data. The whole pipeline is modified from [clara_pt_brain_mri_segmentation](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/med/models/clara_pt_brain_mri_segmentation).
+
+## Workflow
+
+The model is trained to segment 3 nested subregions of primary brain tumors (gliomas): the "enhancing tumor" (ET), the "tumor core" (TC), the "whole tumor" (WT) based on 4 aligned input MRI scans (T1c, T1, T2, FLAIR).
+- The ET is described by areas that show hyper intensity in T1c when compared to T1, but also when compared to "healthy" white matter in T1c.
+- The TC describes the bulk of the tumor, which is what is typically resected. The TC entails the ET, as well as the necrotic (fluid-filled) and the non-enhancing (solid) parts of the tumor.
+-  The WT describes the complete extent of the disease, as it entails the TC and the peritumoral edema (ED), which is typically depicted by hyper-intense signal in FLAIR.
+
+## Data
+
+The training data is from the [Multimodal Brain Tumor Segmentation Challenge (BraTS) 2018](https://www.med.upenn.edu/sbia/brats2018/data.html).
+
+- Target: 3 tumor subregions
+- Task: Segmentation
+- Modality: MRI
+- Size: 285 3D volumes (4 channels each)
+
+The provided labelled data was partitioned, based on our own split, into training (200 studies), validation (42 studies) and testing (43 studies) datasets.
+
+Please run `scripts/prepare_datalist.py` to produce the data list. The command is like:
+
+```
+python scripts/prepare_datalist.py --path your-brats18-dataset-path
+```
+
+## Training configuration
+
+This model utilized a similar approach described in 3D MRI brain tumor segmentation
+using autoencoder regularization, which was a winning method in BraTS2018 [1]. The training was performed with the following:
+
+- GPU: At least 16GB of GPU memory.
+- Actual Model Input: 224 x 224 x 144
+- AMP: True
+- Optimizer: Adam
+- Learning Rate: 1e-4
+- Loss: DiceLoss
+
+## Input
+
+Input: 4 channel MRI (4 aligned MRIs T1c, T1, T2, FLAIR at 1x1x1 mm)
+
+1. Normalizing to unit std with zero mean
+2. Randomly cropping to (224, 224, 144)
+3. Randomly spatial flipping
+4. Randomly scaling and shifting intensity of the volume
+
+## Output
+
+Output: 3 channels
+- Label 0: TC tumor subregion
+- Label 1: WT tumor subregion
+- Label 2: ET tumor subregion
+
+## Model Performance
+
+The achieved Dice scores on the validation data are:
+- Tumor core (TC): 0.8559
+- Whole tumor (WT): 0.9026
+- Enhancing tumor (ET): 0.7905
+- Average: 0.8518
+
+## commands example
+
+Execute training:
+
+```
+python -m monai.bundle run training --meta_file configs/metadata.json --config_file configs/train.json --logging_file configs/logging.conf
+```
+
+Override the `train` config to execute multi-GPU training:
+
+```
+torchrun --standalone --nnodes=1 --nproc_per_node=8 -m monai.bundle run training --meta_file configs/metadata.json --config_file "['configs/train.json','configs/multi_gpu_train.json']" --logging_file configs/logging.conf
+```
+
+Override the `train` config to execute evaluation with the trained model:
+
+```
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file "['configs/train.json','configs/evaluate.json']" --logging_file configs/logging.conf
+```
+
+Execute inference:
+
+```
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file configs/inference.json --logging_file configs/logging.conf
+```
+
+# Disclaimer
+
+This is an example, not to be used for diagnostic purposes.
+
+# References
+
+[1] Myronenko, Andriy. "3D MRI brain tumor segmentation using autoencoder regularization." International MICCAI Brainlesion Workshop. Springer, Cham, 2018. https://arxiv.org/abs/1810.11654.

docs/license.txt

@@ -0,0 +1,49 @@
+Third Party Licenses
+-----------------------------------------------------------------------
+
+/*********************************************************************/
+i. Multimodal Brain Tumor Segmentation Challenge 2018
+    https://www.med.upenn.edu/sbia/brats2018/data.html
+/*********************************************************************/
+
+Data Usage Agreement / Citations
+
+You are free to use and/or refer to the BraTS datasets in your own
+research, provided that you always cite the following two manuscripts:
+
+[1] Menze BH, Jakab A, Bauer S, Kalpathy-Cramer J, Farahani K, Kirby
+[J, Burren Y, Porz N, Slotboom J, Wiest R, Lanczi L, Gerstner E, Weber
+[MA, Arbel T, Avants BB, Ayache N, Buendia P, Collins DL, Cordier N,
+[Corso JJ, Criminisi A, Das T, Delingette H, Demiralp Γ, Durst CR,
+[Dojat M, Doyle S, Festa J, Forbes F, Geremia E, Glocker B, Golland P,
+[Guo X, Hamamci A, Iftekharuddin KM, Jena R, John NM, Konukoglu E,
+[Lashkari D, Mariz JA, Meier R, Pereira S, Precup D, Price SJ, Raviv
+[TR, Reza SM, Ryan M, Sarikaya D, Schwartz L, Shin HC, Shotton J,
+[Silva CA, Sousa N, Subbanna NK, Szekely G, Taylor TJ, Thomas OM,
+[Tustison NJ, Unal G, Vasseur F, Wintermark M, Ye DH, Zhao L, Zhao B,
+[Zikic D, Prastawa M, Reyes M, Van Leemput K. "The Multimodal Brain
+[Tumor Image Segmentation Benchmark (BRATS)", IEEE Transactions on
+[Medical Imaging 34(10), 1993-2024 (2015) DOI:
+[10.1109/TMI.2014.2377694
+
+[2] Bakas S, Akbari H, Sotiras A, Bilello M, Rozycki M, Kirby JS,
+[Freymann JB, Farahani K, Davatzikos C. "Advancing The Cancer Genome
+[Atlas glioma MRI collections with expert segmentation labels and
+[radiomic features", Nature Scientific Data, 4:170117 (2017) DOI:
+[10.1038/sdata.2017.117
+
+In addition, if there are no restrictions imposed from the
+journal/conference you submit your paper about citing "Data
+Citations", please be specific and also cite the following:
+
+[3] Bakas S, Akbari H, Sotiras A, Bilello M, Rozycki M, Kirby J,
+[Freymann J, Farahani K, Davatzikos C. "Segmentation Labels and
+[Radiomic Features for the Pre-operative Scans of the TCGA-GBM
+[collection", The Cancer Imaging Archive, 2017. DOI:
+[10.7937/K9/TCIA.2017.KLXWJJ1Q
+
+[4] Bakas S, Akbari H, Sotiras A, Bilello M, Rozycki M, Kirby J,
+[Freymann J, Farahani K, Davatzikos C. "Segmentation Labels and
+[Radiomic Features for the Pre-operative Scans of the TCGA-LGG
+[collection", The Cancer Imaging Archive, 2017. DOI:
+[10.7937/K9/TCIA.2017.GJQ7R0EF

models/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:860ccb3f1c21c99d0410ad8a1ac4ef6b8fab60cec0a503b0ba42675741a750ae
+size 18840620

models/model.ts

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:729980a0bd9347bf2397701eb329e12517918dc282a2d09c40458e95b24ceed9
+size 18911784

scripts/prepare_datalist.py

@@ -0,0 +1,73 @@
+import argparse
+import glob
+import json
+import os
+
+import monai
+from sklearn.model_selection import train_test_split
+
+
+def produce_sample_dict(line: str):
+    names = os.listdir(line)
+    seg, t1ce, t1, t2, flair = [], [], [], [], []
+    for name in names:
+        name = os.path.join(line, name)
+        if "_seg.nii" in name:
+            seg.append(name)
+        elif "_t1ce.nii" in name:
+            t1ce.append(name)
+        elif "_t1.nii" in name:
+            t1.append(name)
+        elif "_t2.nii" in name:
+            t2.append(name)
+        elif "_flair.nii" in name:
+            flair.append(name)
+
+    return {"label": seg[0], "image": t1ce + t1 + t2 + flair}
+
+
+def produce_datalist(dataset_dir: str):
+    """
+    This function is used to split the dataset.
+    It will produce 200 samples for training, and the other samples are divided equally
+    into val and test sets.
+    """
+
+    samples = sorted(glob.glob(os.path.join(dataset_dir, "*", "*"), recursive=True))
+    datalist = []
+    for line in samples:
+        datalist.append(produce_sample_dict(line))
+    train_list, other_list = train_test_split(datalist, train_size=200)
+    val_list, test_list = train_test_split(other_list, train_size=0.5)
+
+    return {"training": train_list, "validation": val_list, "testing": test_list}
+
+
+def main(args):
+    """
+    split the dataset and output the data list into a json file.
+    """
+    data_file_base_dir = os.path.join(args.path, "training")
+    output_json = args.output
+    # produce deterministic data splits
+    monai.utils.set_determinism(seed=123)
+    datalist = produce_datalist(dataset_dir=data_file_base_dir)
+    with open(output_json, "w") as f:
+        json.dump(datalist, f)
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser(description="")
+    parser.add_argument(
+        "--path",
+        type=str,
+        default="/workspace/data/medical/brats2018challenge",
+        help="root path of brats 2018 dataset.",
+    )
+    parser.add_argument(
+        "--output", type=str, default="configs/datalist.json", help="relative path of output datalist json file."
+    )
+    args = parser.parse_args()
+
+    main(args)