Volumes (DICOM)


In this tutorial we will focus on working with volumes using Supervisely SDK.

You will learn how to:

📗 Everything you need to reproduce this tutorial is on GitHub: source code and demo data.

How to debug this tutorial

Step 1. Prepare ~/supervisely.env file with credentials. Learn more here.

Step 2. Clone repository with source code and demo data and create Virtual Environment.

git clone https://github.com/supervisely-ecosystem/tutorial-volume.git

cd tutorial-volume


Step 3. Open repository directory in Visual Studio Code.

code -r .

Step 4. Change workspace ID in local.env file by copying the ID from the context menu of the workspace.

WORKSPACE_ID=654 # ⬅️ change value

Step 5. Download sample volumes

Step 6. Place downloaded files in the project structure as shown below:

├── .vscode
├── src
│     ├── upload
│     │     ├── MRHead_dicom        <-- sample dicom files
│     │     │   ├── 000000.dcm
│     │     │   ├── 000001.dcm
│     │     │   └── ...
│     │     └── nrrd
│     │         ├── CTACardio.nrrd  <-- sample NRRD
│     │         ├── CTChest.nrrd    <-- sample NRRD
│     │         └── MRHead.nrrd     <-- sample NRRD
│     └── main.py
├── .gitignore
├── create_venv.sh
├── local.env
├── README.md
└── requirements.txt

Step 7. Start debugging src/main.py.

Import libraries

import os

from dotenv import load_dotenv
from pprint import pprint
import supervisely as sly

Init API client

First, we load environment variables with credentials and init API for communicating with Supervisely Instance.

if sly.is_development():

api = sly.Api()

Get variables from environment

In this tutorial, you will need an workspace ID that you can get from environment variables. Learn more here

workspace_id = sly.env.workspace_id()

Create new project and dataset

Create new project with ProjectType.VOLUMES type.

Source code:

project = api.project.create(
    "Volume tutorial",

print(f"Project ID: {project.id}")


# Project ID: 16342

Create new dataset.

Source code:

dataset = api.dataset.create(project.id, "dataset_1")

print(f"Dataset ID: {dataset.id}")


# Dataset ID: 54698

Upload volumes from local directory to Supervisely

Upload NRRD format volume

Source code:

local_path = "src/upload/nrrd/MRHead.nrrd"

nrrd_info = api.volume.upload_nrrd_serie_path(
print(f'"{nrrd_info.name}" volume uploaded to Supervisely with ID:{nrrd_info.id}')


# "NRRD_1.nrrd" volume uploaded to Supervisely with ID:18562981

Upload volume as NumPy array

Source code:

np_volume, meta = sly.volume.read_nrrd_serie_volume_np(local_path)

nrrd_info_np = api.volume.upload_np(

print(f"Volume uploaded as NumPy array to Supervisely with ID:{nrrd_info_np.id}")


# Volume uploaded as NumPy array to Supervisely with ID:18562982

Upload DICOM series from local directory

Inspect you local directory and collect all dicom series.

Source code:

dicom_dir_name = "src/upload/MRHead_dicom/"

series_infos = sly.volume.inspect_dicom_series(root_dir=dicom_dir_name)

Upload DICOM series from local directory to Supervisely platform.

Source code:

for serie_id, files in series_infos.items():
    item_path = files[0]
    name = f"{sly.fs.get_file_name(path=item_path)}.nrrd"
    dicom_info = api.volume.upload_dicom_serie_paths(
    print(f"DICOM volume has been uploaded to Supervisely with ID: {dicom_info.id}")

Set anonymize=True if you want to anonymize DICOM series and hide PatientID and PatientName fields.


# DICOM volume has been uploaded to Supervisely with ID: 18630608

Upload list of volumes from local directory

Source code:

local_dir_name = "src/upload/nrrd/"
all_nrrd_names = os.listdir(local_dir_name)
names = [f"1_{name}" for name in all_nrrd_names]
paths = [os.path.join(local_dir_name, name) for name in all_nrrd_names]

volume_infos = api.volume.upload_nrrd_series_paths(dataset.id, names, paths)
print(f"All volumes has been uploaded with IDs: {[x.id for x in volume_infos]}")


# All volumes has been uploaded with IDs: [18630605, 18630606, 18630607]

Now you can explore and label it in Supervisely labeling tool:

Get volume info from Supervisely

Get list of volumes infos from current dataset

Source code:

volume_infos = api.volume.get_list(dataset.id)

volumes_ids = [x.id for x in volume_infos]

print(f"List of volumes`s IDs: {volumes_ids}")


# List of volumes`s IDs: [18562986, 18562987, 18562988, 18562989, 18562990]

Get single volume info by id

Source code:

volume_id = volume_infos[0].id

volume_info_by_id = api.volume.get_info_by_id(id=volume_id)

print(f"Volume name:", volume_info_by_id.name)


# Volume name: NRRD_1.nrrd

Get single volume info by name

Source code:

volume_info_by_name = api.volume.get_info_by_name(dataset.id, name="MRHead.nrrd")

print(f"Volume name:", volume_info_by_name.name)


# Volume name: NRRD_1.nrrd

Download volume from Supervisely to local directory

Source code:

volume_id = volume_infos[0].id
volume_info = api.volume.get_info_by_id(id=volume_id)

download_dir_name = "src/download/"
path = os.path.join(download_dir_name, volume_info.name)
if os.path.exists(path):

api.volume.download_path(volume_info.id, path)

if os.path.exists(path):
    print(f"Volume (ID {volume_info.id}) successfully downloaded.")


# Volume (ID 18630603) successfully downloaded.

Get volume slices from local directory

Read NRRD file from local directory

Read NRRD file from local directory and get meta and volume (as NumPy array).

Source code:

# read NRRD file from local directory
nrrd_path = os.path.join(download_dir_name, "MRHead.nrrd")
volume_np, meta = sly.volume.read_nrrd_serie_volume_np(nrrd_path)



# {
#     'ACS': 'RAS',
#     'channelsCount': 1,
#     'dimensionsIJK': {'x': 130, 'y': 256, 'z': 256},
#     'directions': (1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0),
#     'intensity': {'max': 279.0, 'min': 0.0},
#     'origin': (-86.64489746093749, -121.07139587402344, -138.21430206298828),
#     'rescaleIntercept': 0,
#     'rescaleSlope': 1,
#     'spacing': (1.2999954223632812, 1.0, 1.0),
#     'windowCenter': 139.5,
#     'windowWidth': 279.0
# }

Get slices from volume

Get slices from current volume. In this example we will get sagittal slices.

Source code:

slices = {}

dimension = volume_np.shape[0]  # change index: 0 - sagittal, 1 - coronal, 2 - axial
for batch in sly.batched(list(range(dimension))):
    for i in batch:
        if i >= dimension:
        pixel_data = volume_np[i, :, :]  # sagittal
        # pixel_data = volume_np[:, i, :]  # coronal
        # pixel_data = volume_np[:, :, i]  # axial
        slices[i] = pixel_data

print(f"{len(slices.keys())} slices has been received from current volume.")


# 130 slices has been received from current volume.

Download slice from Supervisely

Download slice as NumPy from Supervisely by ID

Source code:

slice_index = 60

image_np = api.volume.download_slice_np(

print(f"Image downloaded as NumPy array. Image shape: {image_np.shape}")


# Image downloaded as NumPy array. Image shape: (256, 256, 3)

Save slice to local directory

✅ There is a built-in function supervisely.image.write which reads file extension from path and saves image (slice) with the desired format in local directory.


import supervisely as sly

sly.image.write("folder/slice.nrrd", image_np) # save as NRRD
sly.image.write("folder/slice.jpg", image_np) # save as JPG

Save slice as NRRD

Recommended way to save slice as NRRD file to preserve image quality (pixel depth)

Source code:

# save slice as NRRD file
save_dir = "src/download/"
nrrd_slice_path = os.path.join(save_dir, 'slice.nrrd')

sly.image.write(nrrd_slice_path, image_np)

Save slice as JPG

Source code:

# save slice as jpg
save_dir = "src/download/"
image_slice_path = os.path.join(save_dir, 'slice.jpg')

sly.image.write(jpg_slice_path, image_np)


In case you save slice using nrrd library, it is recommended to use C-order indexing.

save_dir = "src/download/"
slice_path = os.path.join(save_dir, 'slice.nrrd')

nrrd.write(slice_path, image_np, index_order='C')

Last updated