fsantini/pyvoxel

Fast Pythonic data structures and tools for wrangling medical images.

Voxel

Documentation | Installation | Basic Usage

Voxel provides fast Pythonic data structures and tools for wrangling with medical images.

Installation

Voxel requires Python 3.7+. The core module depends on numpy, nibabel, pydicom, requests, and tqdm.

To install Voxel, run:

pip install pyvoxel

Features

Simplified, Efficient I/O

Voxel provides efficient readers for DICOM and NIfTI formats built on nibabel and pydicom. Multi-slice DICOM data can be loaded in parallel with multiple workers and structured into the appropriate 3D volume(s). For example, multi-echo and dynamic contrast-enhanced (DCE) MRI scans have multiple volumes acquired at different echo times and trigger times, respectively. These can be loaded into multiple volumes with ease:

import voxel as vx

xray = vx.load("path/to/xray.dcm")
ct_scan = vx.load("path/to/ct/folder/")

multi_echo_scan = vx.load("/path/to/multi-echo/scan", group_by="EchoNumbers")
dce_scan = vx.load("/path/to/dce/scan", group_by="TriggerTime")

Data-Embedded Medical Images

Voxel's MedicalVolume data structure supports array-like operations (arithmetic, slicing, etc.) on medical images while preserving spatial attributes and accompanying metadata. This structure supports NumPy interoperability intelligent reformatting, fast low-level computations, and native GPU support. For example, given MedicalVolumes mv_a and mv_b we can do the following:

# Reformat image into Superior->Inferior, Anterior->Posterior, Left->Right directions.
mv_a = mv_a.reformat(("SI", "AP", "LR"))

# Get and set metadata
study_description = mv_a.get_metadata("StudyDescription")
mv_a.set_metadata("StudyDescription", "A sample study")

# Perform NumPy operations like you would on image data.
rss = np.sqrt(mv_a**2 + mv_b**2)

# Move to GPU 0 for CuPy operations
mv_gpu = mv_a.to(vx.Device(0))

# Take slices. Metadata will be sliced appropriately.
mv_subvolume = mv_a[10:20, 10:20, 4:6]

Easily Prepare Data for AI Pipelines

Voxel enables you to preprocess DICOM images for deep learning in a few lines of code:

# Load a scan, and prepare it for AI/visualization
mv = (
  vx.load("/dicoms")
  .apply_rescale()
  .apply_window()
  .to_grayscale()
)

# Zero-copy to PyTorch
arr = mv.to_torch()

Connect with PACS

Voxel provides easy access to data stored in a PACS environment through DICOMweb. This makes loading data from a remote server just as easy as using the local filesystem.

# Download an MRI from a local Orthanc instance
mv = vx.load("http://localhost:8042/dicom-web/studies/x/series/y", params={"Modality": "MR"})

# Re-use the session for multiple requests
with vx.HttpReader(verbose=True) as hr:
  mv_a = hr.load("http://localhost:8042/dicom-web/studies/v/series/w")
  mv_b = hr.load("http://localhost:8042/dicom-web/studies/x/series/y")

Contribute

If you would like to contribute to Voxel, we recommend you clone the repository and install Voxel with pip in editable mode.

git clone [email protected]:pyvoxel/pyvoxel.git
cd pyvoxel
pip install -e '.[dev,docs]'
make dev

To run tests, build documentation and contribute, run

make autoformat test build-docs

Citation

Voxel is a refactored version of the DOSMA package that focuses on medical image data structures and I/O. If you use Voxel in your research, please cite the following work:

@inproceedings{desai2019dosma,
  title={DOSMA: A deep-learning, open-source framework for musculoskeletal MRI analysis},
  author={Desai, Arjun D and Barbieri, Marco and Mazzoli, Valentina and Rubin, Elka and Black, Marianne S and Watkins, Lauren E and Gold, Garry E and Hargreaves, Brian A and Chaudhari, Akshay S},
  booktitle={Proc 27th Annual Meeting ISMRM, Montreal},
  pages={1135},
  year={2019}
}

In addition to Voxel, please also consider citing the work that introduced the method used for analysis.