CLI Usage
Info
Get basic project info
$ mdai project info --help
Usage: mdai project info [options]
Display project info
Options:
-i, --project-id <id> project id
-h, --help display help for command
Example:
$ mdai project info --project-id MwBeWjNr
Project MwBeWjNr - Test Project
--------------------------------------------------
Description:
Private: YES
Created: 2018-10-07 19:10:16
Updated: 2018-10-07 19:11:39
Get list of project datasets
$ mdai project list-datasets --help
Usage: mdai project list-datasets [options]
List project datasets
Options:
-i, --project-id <id> project id
-h, --help display help for command
Example:
$ mdai project list-datasets --project-id MwBeWjNr
Project MwBeWjNr: Test Project
--------------------------------------------------
Datasets:
ID Name Type Source Created Updated Processing
D_NbrqAB Head CT DICOM upload 2018-10-07 19:11:39 2018-10-07 19:11:39
Get dataset info
$ mdai dataset info --help
Usage: mdai dataset info [options]
Display dataset info
Options:
-i, --dataset-id <id> dataset id
-h, --help display help for command
Example:
$ mdai dataset info --dataset-id D_NbrqAB
Dataset D_NbrqAB: Head CT
--------------------------------------------------
Type DICOM
Source upload
Description
Created 2018-10-07 19:11:39
Updated 2018-10-07 19:14:50
# Studies 1
# Series 1
# Images 11
Creating projects and datasets
Create new project
$ mdai project create --help
Usage: mdai project create [options]
Create new project
Options:
-n, --name <name> project name (use quotes if contains spaces)
-d, --description <description> [optional] project description (use quotes if contains spaces)
-m, --model <name> [optional] sample model name
-s, --setting <pathname> [optional] project settings
-h, --help display help for command
Example:
$ mdai project create --name 'New Project' --description 'This is a _markdown_ formatted description.'
Successfully created project New Project [id: L4RJwoNP]
Note
Project name is required, but providing a description is optional.
Create new dataset
$ mdai dataset create --help
Usage: mdai dataset create [options]
Create new dataset in project
Options:
-i, --project-id <id> project id
-s, --source <source> dataset source, one of: [ upload, dicom-push, google-cloud-storage, google-cloud-healthcare-api, amazon-s3 ]
-n, --name <name> dataset name (use quotes if contains spaces)
-d, --description <description> [optional] dataset description (use quotes if contains spaces)
-h, --help display help for command
Example:
$ mdai dataset create --project-id L4RJwoNP --source upload --name 'New Dataset' --description 'This is a _markdown_ formatted description.'
Successfully created DICOM dataset New Dataset with upload source [id: D_GmxvgB]
Note
Providing a description is optional.
Clone existing dataset
$ mdai dataset clone --help
Usage: mdai dataset clone [options]
Clone dataset to a different project
Options:
-i, --dataset-id <id> dataset id to clone from
-p, --project-id <id> target project to clone to
-h, --help display help for command
Example:
$ mdai dataset clone --dataset-id D_6BWvjY --project-id VpBQwNA6
Successfully cloned dataset Test_Dataset [id: D_6BWvjY] to Project Test_Project [id: VpBQwNA6]
Loading DICOM resources
Load DICOM into dataset (upload source only)
$ mdai dataset load --help
Usage: mdai dataset load [options] <pathname>
Load files/folders into dataset (upload source only)
Options:
-i, --dataset-id <id> dataset id
-o, --order-exams-by <ordering> [optional] order exams by [ default, patient_id, study_date_time, study_desc, random ] (default: "default")
-h, --help display help for command
Note
All files and folders under the specified path will be included, including any symlinked files or folders.
Example:
$ mdai dataset load --dataset-id D_GmxvgB dicom_dir
Preparing to upload 'dicom_dir' to DICOM dataset D_GmxvgB...
File count: 11
Total size: 5.52 MB
Uploading:
100% [==================================================] 11/11 [0.0s<0.0s, 478/s]
Completed. No errors.
Done uploading for batch id 1e6ddfed-a32f-4cd4-8113-c683b9122134.
Run `mdai dataset progress -i D_GmxvgB` to track progress.
Load DICOM into dataset (google-cloud-storage or amazon-s3 source only)
For datasets created with source google-cloud-storage or amazon-s3, provide the bucket name (with optional file prefix to filter on).
Note
Bucket permissions must be maintained for the corresponding MD.ai external data service account with at least the Storage Object Viewer role. See the web UI for the service account email, or see the resulting error message for this command when permissions are not present or sufficient.
$ mdai dataset connect --help
Usage: mdai dataset connect [options]
Connect cloud storage bucket into dataset (google-cloud-storage or amazon-s3 source only)
Options:
-i, --dataset-id <id> dataset id
-b, --bucket <bucket> bucket name
-p, --prefix <prefix> [optional] folder prefix
-o, --order-exams-by <ordering> [optional] order exams by [ default, patient_id, study_date_time, study_desc, random ] (default: "default")
-h, --help display help for command
Example:
$ mdai dataset connect --dataset-id D_1j4M2B --bucket mdai-test-bucket --prefix test
Dataset D_1j4M2B - New Dataset 2
──────────────────────────────────────────────────
Bucket: mdai-test-bucket
Prefix: test
Order Exams: default
Successfully connected bucket. Run `mdai dataset progress -i D_1j4M2B` to track progress.
Check progress of DICOM processing for dataset
$ mdai dataset progress --help
Usage: mdai dataset progress [options]
Check dataset processing progress
Options:
-i, --dataset-id <id> dataset id
-h, --help display help for command
Example:
$ mdai dataset progress --dataset-id D_NbrqAB
Dataset D_NbrqAB: Head CT
--------------------------------------------------
Type: DICOM
Source: upload
Processing progress for worker batch id 1e6ddfed-a32f-4cd4-8113-c683b9122134
Status FINISHED
Subtask: process uploaded files
# Completed 11
# Remaining 0
# Errors 0
Subtask: generate series thumbnails
# Completed 1
# Remaining 0
# Errors 0
Started 2018-10-07 19:13:36
Elapsed time 1
Percent complete 100%
Deleting DICOM resources
Delete existing dataset
$ mdai dataset delete --help
Usage: mdai dataset delete [options]
Delete dataset
Options:
-i, --dataset-id <id> dataset id
-h, --help display help for command
Example:
$ mdai dataset delete -i D_Zb5Eb4
Dataset D_Zb5Eb4 - VAL
This cannot be undone! Enter the dataset name VAL to comfirm deletion:
VAL
Start deleting dataset VAL
Done deleting dataset VAL
Delete existing studies/series/instances from dataset by UIDs
Delete studies:
$ mdai dataset remove-study --help
Usage: mdai dataset remove-study [options]
Remove studies from dataset by UIDs
Options:
-i, --dataset-id <id> dataset id
-U, --study-uids <uids> StudyInstanceUIDs, separated by commas(no spaces)
-h, --help display help for command
Delete series:
$ mdai dataset remove-series --help
Usage: mdai dataset remove-series [options]
Remove series from dataset by UIDs
Options:
-i, --dataset-id <id> dataset id
-U, --series-uids <uids> SeriesInstanceUIDs, separated by commas(no spaces)
-h, --help display help for command
Delete instances:
$ mdai dataset remove-instance --help
Usage: mdai dataset remove-instance [options]
Remove instances from dataset by UIDs
Options:
-i, --dataset-id <id> dataset id
-U, --instance-uids <uids> SOPInstanceUIDs, separated by commas(no spaces)
-h, --help display help for command
Example:
$ mdai dataset remove-study -i D_vBnWb1 -U 1.2.276.0.7230010.3.1.2.8323329.15697.1531787477.263823,1.2.276.0.7230010.3.1.2.8323329.15695.1531787477.240488
Dataset D_vBnWb1 - TRAIN
Successfully deleted DICOM studies with StudyInstanceUID:
1.2.276.0.7230010.3.1.2.8323329.15697.1531787477.263823
1.2.276.0.7230010.3.1.2.8323329.15695.1531787477.240488
$ mdai dataset remove-series -i D_vBnWb1 -U 1.2.276.0.7230010.3.1.3.8323329.15825.1531787477.370926
Dataset D_vBnWb1 - TRAIN
Successfully deleted DICOM series with SeriesInstanceUID:
1.2.276.0.7230010.3.1.3.8323329.15825.1531787477.370926
$ mdai dataset remove-instance -i D_vBnWb1 -U 1.2.276.0.7230010.3.1.4.8323329.15772.1531787477.330136,1.2.276.0.7230010.3.1.4.8323329.15650.1531787477.198749,1.2.276.0.7230010.3.1.4.8323329.15664.1531787477.222040
Dataset D_vBnWb1 - TRAIN
Successfully deleted DICOM series with SOPInstanceUID:
1.2.276.0.7230010.3.1.4.8323329.15772.1531787477.330136
1.2.276.0.7230010.3.1.4.8323329.15650.1531787477.198749
1.2.276.0.7230010.3.1.4.8323329.15664.1531787477.222040
User management
Add users
$ mdai project add-user --help
Usage: mdai project add-user [options]
Add user to project
Options:
-i, --project-id <id> project id
-e, --email <email> user email
-r, --role <role> role of project user [ READER, WRITER, QA, ADMIN ]
-h, --help display help for command
Example:
$ mdai project add-user --project-id maBjAq5 --email user@email.com --role WRITER
Successfully added user U_eMzVMx with WRITER role.
Get list of users
List users in project with their user id, name, email, and role.
$ mdai project list-users --help
Usage: mdai project list-users [options]
List project users
Options:
-i, --project-id <id> project id
-j, --json [optional] output as JSON
-h, --help display help for command
Example:
$ mdai project list-users --project-id MwBeWjNr
ID Name Email Role Pending
-----------------------------------------------------------------
U_y01508 Jane Doe jane@md.ai ADMIN No
Get list of assignments
List user assignments in project with their user id, name, label group restrictions, and exams list (dicomStudyRestriction).
$ mdai project list-user-assignments --help
Usage: mdai project list-user-assignments [options]
List assignments for project users
Options:
-i, --project-id <id> project id
-j, --json [optional] output as JSON
-h, --help display help for command
Example:
$ mdai project list-user-assignments --project-id MwBeWjNr
userId name labelGroupRestriction dicomStudyRestriction
U_8dX2w8 femur@yopmail.com G_WjoBx1,G_ADkazj D_WmRdXB: 20-30,35,37
U_6vkyr8 ulna@yopmail.com G_A1AXoD None
Add --json to get output in JSON format:
$ mdai project list-user-assignments --project-id MwBeWjNr --json
"data": [
{
"userId": "U_8dX2w8",
"name": "femur@yopmail.com",
"labelGroupRestriction": [
"G_WjoBx1",
"G_ADkazj"
],
"dicomStudyRestriction": [
{
"dicomDatasetId": "D_WmRdXB",
"dicomDatasetStudyNumbers": "20-30,35,37"
}
]
},
{
"userId": "U_6vkyr8",
"name": "ulna@yopmail.com",
"labelGroupRestriction": ["G_A1AXoD"],
"dicomStudyRestriction": []
}
]
Add/update assignments
$ mdai project update-user-assignment --help
Usage: mdai project update-user-assignment [options]
Update assignments for project users
Options:
-i, --project-id <id> project id
-u, --user-id <id> user id
-d, --dataset-id <id> [optional: use with --exams] dataset id
-e, --exams <sequence> [optional: use with --dataset-id] for specified dataset, assign exams such as 1-3,3-4 (use '' to reset)
-g, --label-group <number(s)> [optional] label group number, or multiple comma-separated numbers (use '' to reset)
-h, --help display help for command
This command will assign a range of exams and specific label group(s) to a user programmatically. Users need to be invited to the project first in order to have a user id. The following values need to be known:
- project id
- user id
- dataset id, if specific exams are to be assigned
- label group number, if specific label groups are to be assigned
Info
Exam assignments are tied to a specific dataset so it must be provided together with a dataset id. For multiple datasets, multiple commands can be issued (one for each dataset).
Example (generic):
$ mdai project update-user-assignment --project-id PROJECT_ID --user-id USER_ID \
--dataset-id DATASET_ID --exams EXAM_NUMBERS \
--label-group LABEL_GROUP_NUMBER
Successfully updated project user assignment.
Example (specific):
$ mdai project update-user-assignment --project-id maBjAq5 --user-id U_oVjzV \
--dataset-id D_v0B1N --exams '1-30,47,51-100' \
--label-group '1,4'
Successfully updated project user assignment.
Info
Assignments cannot be made for project admin users. Invite the user with an alternate email address (which will create a separate user with a separate user id) if an assignment needs to be made for that user.
In order to remove/reset assignments or parts of an assignment, use an empty string. For example:
$ mdai project update-user-assignment --project-id maBjAq5 --user-id U_oVjzV \
--dataset-id D_v0B1N --exams '' \
--label-group ''
Models
Check model info
$ mdai model info --help
Usage: mdai model info [options]
Display model info
Options:
-i, --model-id <id> model id
-h, --help display help for command
Example:
$ mdai model info --model-id M_0N3ADm
Model M_0N3ADm - X-Ray Classifier
──────────────────────────────────────────────────
Description: Classifies X-rays as chest or abdomen
Scope: Image
Created: 2022-09-13 11:35:19
Updated: 2022-09-13 11:36:51
Privacy: Private
AutoRun: No
Exportable: No
Cloneable: No
──────────────────────────────────────────────────
Labels With Class Index
──────────────────────────────────────────────────────────────────
class index │ label id │ label name │ label type │ label scope
──────────────────────────────────────────────────────────────────
0 │ L_QRb9NQ │ Abdomen │ Global │ Image
──────────────────────────────────────────────────────────────────
1 │ L_gw3RNg │ Chest │ Global │ Image
──────────────────────────────────────────────────────────────────
──────────────────────────────────────────────────
Versions
─────────────────────────────────────────────────────────────────────────────────────────────────────
# │ id │ Docker Image:Tag │ Created │ Build Status │ Hardware
─────────────────────────────────────────────────────────────────────────────────────────────────────
1 │ 541 │ gcr.io/mdai-models/staging/model-198/v1 │ 2022-09-13 11:35:20 │ Success │ CPU
─────────────────────────────────────────────────────────────────────────────────────────────────────
Clone model
$ mdai model clone --help
Usage: mdai model clone [options]
Clone project models
Options:
--from-project-id <id> original project id
--to-project-ids <ids> target project ids, separated by commas
--from-model-ids <ids> [optional] original model ids in original project, separated by commas
--from-domain <domain> [optional] original project MD.ai domain (e.g., public.md.ai)
-c, --config-file <path> [optional] clone config file (.json or .yaml)
-h, --help display help for command
Example:
$ mdai model clone --from-project-id MmRa7Mqr --to-project-ids 0pqXObNZ
Clone result of project 0pqXObNZ: Clones 3 models into project
Clone models complete!
Update model
$ mdai model set --help
Usage: mdai model set [options]
Update model
Options:
-i, --model-id <id> model id
-n, --model-name <name> [optional] model name (use quotes if contains spaces)
-d, --model-description <description> [optional] model description (use quotes if contains spaces)
-p, --model-privacy <privacy> [optional] configure model privacy (private or public)
-a, --model-autorun-inference [autorun] [optional] configure model autorun (y/n, yes or no)
-e, --model-exportable [exportable] [optional] configure model exportable (y/n, yes or no)
-c, --model-cloneable [cloneable] [optional] configure model cloneable (y/n, yes or no)
-s, --model-clone-scope <scope> [optional] model clone scope (SITE_ONLY or ALL_DOMAINS)
-h, --help display help for command
Example:
Change model name:
$ mdai model set --model-id M_7bxME5 --model-name "Demo model"
Successfully set model [id: M_7bxME5] name to be Demo Models
Set model autorun inference to yes:
$ mdai model set --model-id M_7bxME5 --model-autorun-inference yes
Successfully set model [id: M_7bxME5] autorun inference to be yes.
Delete model
$ mdai model delete --help
Usage: mdai model delete [options]
Delete model
Options:
-i, --model-id <id> model id
-h, --help display help for command
Example:
Create model task
$ mdai model create-task --help
Usage: mdai model create-task [options]
Create new model inference task on dataset
Options:
-m, --model-version-id <id> model version id
-i, --dataset-id <id> dataset id
-j, --json [optional] output as JSON
-h, --help display help for command
Example:
$ mdai model create-task --model-version-id 13 --dataset-id D_vbvwom
Successfully created model inference task [Task ID: 190] on dataset Finetuning [Dataset ID: D_vbvwom]
Clone model outputs as annotations
$ mdai model clone-output --help
Usage: mdai model clone-output [options]
Clone model outputs as annotations for entire model task
Options:
-i, --model-task-id <id> model task id
-h, --help display help for command
Example:
Project starter packs with models
List available models
Currently supported models for starter packs:
$ mdai model create --list-models
─────────────────────────────────────────────────────────────────────────
name │ description
─────────────────────────────────────────────────────────────────────────
mdai:xray-classifier │ Classifies X-rays as chest or abdomen
─────────────────────────────────────────────────────────────────────────
mdai:lung-segmentation │ Generates segmentation for lungs in xray images
─────────────────────────────────────────────────────────────────────────
mdai:pixel-deid │ Pixel level De-identification
─────────────────────────────────────────────────────────────────────────
Create new project with selected models
$ mdai project create --help
Usage: mdai project create [options]
Create new project
Options:
-n, --name <name> project name (use quotes if contains spaces)
-d, --description <description> [optional] project description (use quotes if contains spaces)
-m, --model <name> [optional] sample model name
-s, --setting <pathname> [optional] project settings
-h, --help display help for command
Example:
$ mdai project create --name "Test Project" --model mdai:xray-classifier
Successfully created project Test Project [id: Y2qrE9Nv] and model X-Ray Classifier [id: M_0N3ADm]
Add model to project
$ mdai model create --help
Usage: mdai model create [options]
Create model
Options:
-l, --list-models list available sample models
-i, --project-id <id> project id
-n, --model-name <name> sample model name
-h, --help display help for command
Example:
$ mdai model create --project-id Y2qrE9Nv --model-name mdai:xray-classifier
Successfully created model X-Ray Classifier [id: M_0N3ADm] in project [id: Y2qrE9Nv]
De-identification (De-ID)
Create De-ID task
Creates a DeID task that can be found in the De-ID panel on the UI. Can be used to create both pixel redaction and dicom tag redaction tasks.
$ mdai deid create-task --help
Usage: mdai deid create-task [options]
Create De-ID task
Options:
-i, --dataset-id <id> dataset id
-a, --auto-inference [value] [optional] Apply automatic inference (y/n, yes or no)
-m, --auto-inference-model <name> [optional] Automatic inference model name (defaults to "mdai:pixel-deid")
-l, --label-ids <ids> [optional] label ids, separated by commas (no spaces)
-r, --remove-private-tags [value] [optional] Remove private tags (y/n, yes or no)
-c, --tags-config <path or "default"> [optional] path to DICOM tags modification JSON config file (load default if "default")
-j, --json [optional] output as JSON
-h, --help display help for command
Example:
- To run the entire De-identification workflow (including adding the latest De-ID model to the project, running infernece, and redacting all the model outputs by running the De-ID task) using one single command -
$ mdai deid create-task -i D_emLA9j -a yes
Successfully created De-ID task [id: 126] for dataset [id: D_emLA9j]
- To just run a part of the workflow where you want to redact manually created or inferenced labels you can choose to add the label IDs to redact -
$ mdai deid create -i D_emLA9j -l L_EM2rPN,L_ZgDraE
Successfully created De-ID task [id: 127] for dataset [id: D_emLA9j]
- To create a dicom tag redaction task with the default tag redaction config file -
$ mdai deid create -i D_emLA9j -c default
Successfully created De-ID task [id: 128] for dataset [id: D_emLA9j]
- To create a dicom tag redaction tasl the utilizes your own config file for tags -
# Export default tags config file, exported config file name would be `mdai_deid_tags_config.json`
mdai deid export-tags
# ......
# Do some custom modification for tags config file
# ......
# Use custom tags config file when create De-ID task
mdai deid create -i D_emLA9j -c mdai_deid_tags_config.json
Successfully created De-ID task [id: 126] for dataset [id: D_VNJgbG]
Get task info
$ mdai deid task-info --help
Usage: mdai deid task-info [options]
Display deid task info
Options:
-i, --deid-task-id <id> deid task id
-j, --json [optional] output as JSON
-h, --help display help for command
Example
$ mdai deid task-info -i 125
DEID Task 125
──────────────────────────────────────────────────
Type: DEID
Status: SUCCEEDED
Progress: 100
Created at: 2024-01-02 16:30:25
Start Time: 2024-01-02 16:30:29
End Time: 2024-01-02 16:40:19
Success Count: 3
Error Count:
New Dataset ID: D_LNqELN
Export default dicom tag redaction config
$ mdai deid export-tags --help
Usage: mdai deid export-tags [options]
Export default DICOM tags modification config file
Options:
-d, --export-dir <path> [optional] export dir
-h, --help display help for command
Example