MONAI APP

class MONAILabelApp(app_dir, studies, train_stats_path=None)

Base Class for Any MONAI Label App

Parameters

  • app_dir – path for your App directory

  • studies – path for studies/datalist

  • train_stats_path – Path for Training stats json

batch_infer(request, datastore=None)

Run batch inference for an existing pre-trained model.

Parameters

  • request – JSON object which contains model, params and device

  • datastore –

    Datastore object. If None then use default app level datastore to fetch the images

    For example:

    {
    "device": "cuda"
    "model": "segmentation_spleen",
    "images": "unlabeled",
    "label_tag": "original"
}

Raises

MONAILabelException – When model is not found

Returns

JSON containing label and params

static deepgrow_infer_tasks(model_dir, pipeline=True)

Dictionary of Default Infer Tasks for Deepgrow 2D/3D

infer(request, datastore=None)

Run Inference for an exiting pre-trained model.

Parameters

  • request – JSON object which contains model, image, params and device

  • datastore –

    Datastore object. If None then use default app level datastore to save labels if applicable

    For example:

    {
    "device": "cuda"
    "model": "segmentation_spleen",
    "image": "file://xyz",
    "save_label": "true/false",
    "label_tag": "original"
}

Raises

MONAILabelException – When model is not found

Returns

JSON containing label and params

info()

Provide basic information about APP. This information is passed to client. Default implementation is to pass the contents of info.yaml present in APP_DIR

next_sample(request)

Run Active Learning selection. User APP has to implement this method to provide next sample for labelling.

Parameters

request –

JSON object which contains active learning configs that are part APP info

For example:

{
    "strategy": "random"
}

Returns

JSON containing next image info that is selected for labeling

on_save_label(image_id, label_id) → Dict[str, Any]

Callback method when label is saved into datastore by a remote client

scoring(request, datastore=None)

Run scoring task over labels.

Parameters

  • request – JSON object which contains model, params and device

  • datastore –

    Datastore object. If None then use default app level datastore to fetch the images

    For example:

    {
    "device": "cuda"
    "method": "dice",
    "y": "final",
    "y_pred": "original",
}

Raises

MONAILabelException – When method is not found

Returns

JSON containing result of scoring method

abstract train(request)

Run Training. User APP has to implement this method to run training

Parameters

request –

JSON object which contains train configs that are part APP info

For example:

{
    "device": "cuda"
    "epochs": 1,
    "amp": False,
    "lr": 0.0001,
}

Returns

JSON containing train stats

class Datastore
abstract add_image(image_id: str, image_filename: str) → str

Save a image for the given image id and return the newly saved image’s id

Parameters

  • image_id – the image id for the image; If None then base filename will be used

  • image_filename – the path to the image file

Returns

the image id for the saved image filename

abstract datalist() → List[Dict[str, str]]

Return a dictionary of image and label pairs corresponding to the ‘image’ and ‘label’ keys respectively

Returns

the {‘label’: image, ‘label’: label} pairs for training

abstract description() → str

Return the user-set description of the dataset

Returns

the user-set description of the dataset

abstract get_image(image_id: str) → Any

Retrieve image object based on image id

Parameters

image_id – the desired image’s id

Returns

return the “image”

abstract get_image_info(image_id: str) → Dict[str, Any]

Get the image information for the given image id

Parameters

image_id – the desired image id

Returns

image info as a list of dictionaries Dict[str, Any]

abstract get_image_uri(image_id: str) → str

Retrieve image uri based on image id

Parameters

image_id – the desired image’s id

Returns

return the image uri

abstract get_label(label_id: str) → Any

Retrieve image object based on label id

Parameters

label_id – the desired label’s id

Returns

return the “label”

abstract get_label_by_image_id(image_id: str, tag: str) → str

Retrieve label id for the given image id and tag

Parameters

  • image_id – the desired image’s id

  • tag – matching tag name

Returns

label id

abstract get_label_info(label_id: str) → Dict[str, Any]

Get the label information for the given label id

Parameters

label_id – the desired label id

Returns

label info as a list of dictionaries Dict[str, Any]

abstract get_label_uri(label_id: str) → str

Retrieve label uri based on image id

Parameters

label_id – the desired label’s id

Returns

return the label uri

abstract get_labeled_images() → List[str]

Get all images that have a corresponding label

Returns

list of image ids List[str]

abstract get_labels_by_image_id(image_id: str) → Dict[str, str]

Retrieve all label ids for the given image id

Parameters

image_id – the desired image’s id

Returns

label ids mapped to the appropriate LabelTag as Dict[str, LabelTag]

abstract get_unlabeled_images() → List[str]

Get all images that have no corresponding label

Returns

list of image ids List[str]

abstract list_images() → List[str]

Return list of image ids available in the datastore

Returns

list of image ids List[str]

abstract name() → str

Return the human-readable name of the datastore

Returns

the name of the dataset

abstract refresh() → None

Refresh the datastore

abstract remove_image(image_id: str) → None

Remove image for the datastore. This will also remove all associated labels.

Parameters

image_id – the image id for the image to be removed from datastore

abstract remove_label(label_id: str) → None

Remove label from the datastore

Parameters

label_id – the label id for the label to be removed from datastore

abstract remove_label_by_tag(label_tag: str) → None

Remove all labels for matching tags from the datastore

Parameters

label_tag – the label tag for the label to be removed from datastore

abstract save_label(image_id: str, label_filename: str, label_tag: str) → str

Save a label for the given image id and return the newly saved label’s id

Parameters

  • image_id – the image id for the label

  • label_filename – the path to the label file

  • label_tag – the user-provided tag for the label

Returns

the label id for the given label filename

abstract set_description(description: str)

A human-readable description of the datastore

Parameters

description – string for description

abstract set_name(name: str)

Set the name of the datastore

Parameters

name – a human-readable name for the datastore

abstract status() → Dict[str, Any]

Return current statistics of datastore

abstract update_image_info(image_id: str, info: Dict[str, Any]) → None

Update (or create a new) info tag for the desired image

Parameters

  • image_id – the id of the image we want to add/update info

  • info – a dictionary of custom image information Dict[str, Any]

abstract update_label_info(label_id: str, info: Dict[str, Any]) → None

Update (or create a new) info tag for the desired label

Parameters

  • label_id – the id of the label we want to add/update info

  • info – a dictionary of custom label information Dict[str, Any]

class MONAILabelError(value)
SERVER_ERROR - Server Error
UNKNOWN_ERROR - Unknown Error
CLASS_INIT_ERROR - Class Initialization Error
MODEL_IMPORT_ERROR - Model Import Error
INFERENCE_ERROR - Inference Error
APP_INIT_ERROR - Initialization Error
APP_INFERENCE_FAILED - Inference Failed
APP_TRAIN_FAILED - Train Failed
APP_ERROR APP - General Error
class MONAILabelException(error: monailabel.interfaces.exception.MONAILabelError, msg: str)

MONAI Label Exception

Tasks

class InferType

Type of Inference Model .. attribute:: SEGMENTATION - Segmentation Model

CLASSIFICATION - Classification Model
DEEPGROW - Deepgrow Interactive Model
DEEPEDIT - DeepEdit Interactive Model
POSTPROCS - Post-Processor Model
OTHERS - Other Model Type
class InferTask(path, network, type: monailabel.interfaces.tasks.infer.InferType, labels, dimension, description, model_state_dict='model', input_key='image', output_label_key='pred', output_json_key='result')

Basic Inference Task Helper

Parameters

  • path – Model File Path. Supports multiple paths to support versions (Last item will be picked as latest)

  • network – Model Network (e.g. monai.networks.xyz). None in case if you use TorchScript (torch.jit).

  • type – Type of Infer (segmentation, deepgrow etc..)

  • dimension – Input dimension

  • description – Description

  • model_state_dict – Key for loading the model state from checkpoint

  • input_key – Input key for running inference

  • output_label_key – Output key for storing result/label of inference

  • output_json_key – Output key for storing result/label of inference

abstract inferer()

Provide Inferer Class

For Example:

return monai.inferers.SlidingWindowInferer(roi_size=[160, 160, 160])
inverse_transforms()

Provide List of inverse-transforms. They are normally subset of pre-transforms. This task is performed on output_label (using the references from input_key)

Return one of the following.

  • None: Return None to disable running any inverse transforms (default behavior).

  • Empty: Return [] to run all applicable pre-transforms which has inverse method

  • list: Return list of specific pre-transforms names/classes to run inverse method

For Example:

return [
    monai.transforms.Spacingd,
]
abstract post_transforms()

Provide List of post-transforms

For Example:

return [
    monai.transforms.AddChanneld(keys='pred'),
    monai.transforms.Activationsd(keys='pred', softmax=True),
    monai.transforms.AsDiscreted(keys='pred', argmax=True),
    monai.transforms.SqueezeDimd(keys='pred', dim=0),
    monai.transforms.ToNumpyd(keys='pred'),
    monailabel.interface.utils.Restored(keys='pred', ref_image='image'),
    monailabel.interface.utils.ExtremePointsd(keys='pred', result='result', points='points'),
    monailabel.interface.utils.BoundingBoxd(keys='pred', result='result', bbox='bbox'),
]
abstract pre_transforms()

Provide List of pre-transforms

For Example:

return [
    monai.transforms.LoadImaged(keys='image'),
    monai.transforms.AddChanneld(keys='image'),
    monai.transforms.Spacingd(keys='image', pixdim=[1.0, 1.0, 1.0]),
    monai.transforms.ScaleIntensityRanged(keys='image',
        a_min=-57, a_max=164, b_min=0.0, b_max=1.0, clip=True),
]
run_inferer(data, convert_to_batch=True, device='cuda')

Run Inferer over pre-processed Data. Derive this logic to customize the normal behavior. In some cases, you want to implement your own for running chained inferers over pre-processed data

Parameters

  • data – pre-processed data

  • convert_to_batch – convert input to batched input

  • device – device type run load the model and run inferer

Returns

updated data with output_key stored that will be used for post-processing

static run_transforms(data, transforms, inverse=False, log_prefix='POST')

Run Transforms

Parameters

  • data – Input data dictionary

  • transforms – List of transforms to run

  • inverse – Run inverse instead of call/forward function

  • log_prefix – Logging prefix (POST or PRE)

Returns

Processed data after running transforms

writer(data, extension=None, dtype=None)

You can provide your own writer. However this writer saves the prediction/label mask to file and fetches result json

Parameters

  • data – typically it is post processed data

  • extension – output label extension

  • dtype – output label dtype

Returns

tuple of output_file and result_json

class TrainTask
abstract amp()

Use AMP

abstract device()

Provide device name

abstract event_names()

Provide iteration update while training

abstract loss_function()

Provide Lost function

abstract max_epochs()

Max number of epochs to run

abstract network()

Provide network

abstract optimizer()

Provide Optimizer

abstract train_additional_metrics()

Provide any additional metrics to be collected while training

abstract train_data_loader()

Provide Dataloader for training samples

abstract train_handlers()

Provide List of training handlers

abstract train_inferer()

Provide Inferer to be used while training

abstract train_iteration_update()

Provide iteration update while training

abstract train_key_metric()

Provide List of Key Metrics to be collected while training

abstract train_post_transforms()

Provide List of Post-Transforms for training

abstract train_pre_transforms()

Provide List of Pre-Transforms for training

abstract val_additional_metrics()

Provide any additional metrics to be collected while validation

abstract val_data_loader()

Provide Dataloader for validation samples

abstract val_handlers()

Provide List of handlers for validation

abstract val_inferer()

Provide Inferer to be used for validation step

abstract val_iteration_update()

Provide iteration update while validation

abstract val_key_metric()

Provide List of Key metrics to be collected during validation

abstract val_post_transforms()

Provide List of Post-Transforms for validation step

abstract val_pre_transforms()

Provide List of Pre-Transforms for validation step

class Strategy(description)

Basic Active Learning Strategy

Utils

class BasicTrainTask(output_dir, train_datalist, val_datalist, network, load_path=None, load_dict=None, publish_path=None, stats_path=None, device='cuda', max_epochs=1, amp=True, lr=0.0001, train_batch_size=1, train_num_workers=0, train_save_interval=50, val_interval=1, val_batch_size=1, val_num_workers=0, final_filename='checkpoint_final.pt', key_metric_filename='model.pt', **kwargs)

This provides Basic Train Task to train segmentation models over MSD Dataset.

Parameters

  • output_dir – Output to save the model checkpoints, events etc…

  • train_datalist – Training List of dictionary that normally contains {image, label}

  • val_datalist – Validation List of dictionary that normally contains {image, label}

  • network – If None then UNet with channels(16, 32, 64, 128, 256) is used

  • load_path – Initialize model from existing checkpoint

  • load_dict – Provide dictionary to load from checkpoint. If None, then net will be loaded

  • publish_path – Publish path for best trained model (based on best key metric)

  • stats_path – Path to save the train stats

  • device – device name

  • max_epochs – maximum epochs to run

  • amp – use amp

  • lr – Learning Rate (LR)

  • train_batch_size – train batch size

  • train_num_workers – number of workers for training

  • train_save_interval – checkpoint save interval for training

  • val_interval – validation interval (run every x epochs)

  • val_batch_size – batch size for validation step

  • val_num_workers – number of workers for validation step

  • final_filename – name of final checkpoint that will be saved

amp()

Use AMP

device()

Provide device name

event_names()

Provide iteration update while training

loss_function()

Provide Lost function

max_epochs()

Max number of epochs to run

network()

Provide network

optimizer()

Provide Optimizer

train_additional_metrics()

Provide any additional metrics to be collected while training

train_data_loader()

Provide Dataloader for training samples

train_handlers()

Provide List of training handlers

train_inferer()

Provide Inferer to be used while training

train_iteration_update()

Provide iteration update while training

train_key_metric()

Provide List of Key Metrics to be collected while training

abstract train_post_transforms()

Provide List of Post-Transforms for training

abstract train_pre_transforms()

Provide List of Pre-Transforms for training

val_additional_metrics()

Provide any additional metrics to be collected while validation

val_data_loader()

Provide Dataloader for validation samples

val_handlers()

Provide List of handlers for validation

abstract val_inferer()

Provide Inferer to be used for validation step

val_iteration_update()

Provide iteration update while validation

val_key_metric()

Provide List of Key metrics to be collected during validation

val_post_transforms()

Provide List of Post-Transforms for validation step

val_pre_transforms()

Provide List of Pre-Transforms for validation step