voxelkit 0.1.6

Library-first toolkit for inspecting, previewing, and QA reporting imaging datasets.

VoxelKit

Inspect. Preview. QA-check. One toolkit, every format.

Documentation · Getting Started · CLI Reference · Python API

---

You just received 500 brain scans. Before you write a single line of training code — do you know if the data is actually healthy?

VoxelKit answers that in one command.

voxelkit report-batch data/scans/

No format-specific boilerplate. No juggling nibabel, h5py, and tifffile separately. One consistent interface across every file type you work with.

---

Features

• 🔍 Inspect — shape, dtype, and metadata from any supported file, instantly
• 🖼️ Preview — PNG slice from any 3D/4D volume, one command
• 📊 QA Reports — per-file stats and warnings (NaNs, constant arrays, zero-dominated volumes)
• 📁 Batch QA — scan an entire directory, emit JSON or a self-contained HTML report with thumbnails
• 🩻 DICOM Support — single .dcm files and full series directories; PHI stripped by default
• 🛡️ Anonymisation — voxelkit anonymise ./dicoms/ --output ./clean/ scrubs PHI in bulk
• 🔄 DICOM → NIfTI — voxelkit convert scan.dcm scan.nii.gz
• 🧬 Embedding Analysis — dead dimensions, outlier samples, collapsed spaces
• 🌐 REST API — FastAPI server for HTTP-based workflows
• 🖥️ Local GUI — optional Streamlit interface, runs entirely offline

Supported Formats

Format	Extensions / Input
NIfTI	.nii .nii.gz
HDF5	.h5 .hdf5
NumPy	.npy .npz
TIFF	.tif .tiff
DICOM	.dcm file or directory of slices (series)

Install

pip install voxelkit

pip install voxelkit[gui]   # optional Streamlit GUI

Requires Python ≥ 3.9.

Quick Start

# What's in this file?
voxelkit inspect scan.nii.gz

# Is the data healthy?
voxelkit report scan.nii.gz

# Show me a slice
voxelkit preview scan.nii.gz --plane axial --output preview.png

# Check a whole dataset at once (JSON or self-contained HTML)
voxelkit report-batch data/scans/
voxelkit report-batch data/scans/ --html report.html

DICOM workflows:

# Inspect a single .dcm or a whole series directory (PHI stripped by default)
voxelkit inspect scan.dcm
voxelkit inspect ./series/

# Scrub PHI from every .dcm under a directory
voxelkit anonymise ./incoming/ --output ./anonymised/

# Convert DICOM to NIfTI
voxelkit convert ./series/ volume.nii.gz

Or from Python:

from voxelkit import inspect_file, report_file, report_batch

info   = inspect_file("scan.nii.gz")        # also handles .dcm + series dirs
report = report_file("scan.nii.gz")
batch  = report_batch("data/scans/")

# DICOM-specific helpers
from voxelkit.dicom import anonymise_directory, dicom_to_nifti
anonymise_directory("./incoming/", "./anonymised/")
dicom_to_nifti("./series/", "volume.nii.gz")

Full Documentation

Everything else — complete CLI reference, Python API docs, REST API, QA warning explanations, and contributing guide — lives at:

arsalaanahmad.github.io/VoxelKit

Contributing

Contributions are welcome. See the Contributing Guide for setup instructions, workflow, and how to report issues.

Citation

If you use VoxelKit in research, please cite:

@software{voxelkit,
  author  = {Arsalaan Ahmad},
  title   = {VoxelKit},
  doi     = {10.5281/zenodo.19774569},
  url     = {https://github.com/ArsalaanAhmad/VoxelKit},
}

Disclaimer

VoxelKit is a developer and research utility. It is not a clinical decision system and must not be used for diagnosis or treatment.