Skip to content

What's in NIRWizard

Everything the application does after the recording, from quality control to the published figure: what it needs, how it is licensed and how to cite it.

Workshop slides Download 0.2.4

Capabilities

Studies, not folders

  • A study holds subjects, and each subject holds its runs
  • One run policy for the cohort: per-run fits combined by fixed effects, or concatenated
  • Import proposals split files into subjects and runs before you commit
  • Batch first level with one export folder per run and per estimate

Signal and quality

  • SNIRF input, with a sample recording bundled
  • Scalp coupling index and channel marking with reasons
  • Optical density, TDDR, wavelet and spline motion correction, filtering, Beer–Lambert law
  • Cedalion as the reference engine, MNE-NIRS as a compatibility backend

Statistics

  • Block averaging and GLM at the first level
  • Group inference where each participant enters once, however many runs they have
  • Connectivity and laterality
  • Time series, spectrogram and heatmaps for inspection

Diffuse optical tomography

  • Subject T1 anatomy, or the bundled Colin27 and fsaverage heads
  • Optode-to-scalp projection and per-optode fluence on a voxel grid
  • Per-channel sensitivity and regularised inverse reconstruction
  • Cortical HbO and HbR images in the 3D viewport, fed into the statistics

Visualization

  • Stacked time series with event bands, spectra and heatmaps
  • The probe map of the loaded recording
  • The 3-D head and cortex, with a result placed on it from the panel that computed it
  • Figures and exports for publication

Provenance

  • Every result records its inputs, parameters, exclusions and statistics
  • Every export carries that record beside the data
  • A published number traces back to the file and settings that made it
  • Study files refuse cached results when a run has changed

From channels to the cortex

The DOT chain runs inside the same session as the channel-space analysis, and the image it reconstructs goes into the same statistics. It is validated against an independent implementation in Cedalion.

  1. SNIRF recording probe and signals
  2. Subject T1 surfaces and MRI volume
  3. Optode to scalp projection
  4. Fluence per optode, voxel grid
  5. Sensitivity per channel
  6. Reconstruction regularised inverse
  7. Cortical image HbO and HbR, 3D

Optional external tools

NIRWizard runs without any of these. Each one unlocks one part of the application.

  • Python 3.11+ with Cedalion

    Unlocks preprocessing

    The reference engine. Install the pinned tag, since PyPI is stale.

    pip install "cedalion @ git+https://github.com/ibs-lab/cedalion@v26.5.1"

  • Monte Carlo eXtreme

    Unlocks photon simulation

    mcx must be on PATH.

  • Docker and a FreeSurfer licence

    Unlocks anatomy segmentation

    For segmenting a subject T1 into head and brain surfaces.

How it's built

Interface
Svelte 5, TypeScript, ECharts, Three.js
Desktop
Tauri 2
Core
Rust, with nalgebra, ndarray, rayon and rustfft
Data
HDF5 for SNIRF, neuroformats for MRI and surfaces

One study, two tools

Before the recording: the montage and the cap.

What's in WizardHat

After it: preprocessing, analysis and visualization.

Free, and not open source

NIRWizard is free for research, teaching, evaluation and private study, at a university, a hospital or a company, on as many machines as you like. Publish whatever results you get from it.

The source is published so you can inspect, build and verify what it computes. You may not publish, sell, fork publicly or otherwise supply the software or a modified version of it.

Commercial use needs a separate licence. Write to post@askeengineering.no.

Citing NIRWizard

If you use NIRWizard in published work, cite the version you used. A permanent citable record is pending.

Aske, A. E. (2026). NIRWizard [Computer software]. PLACEHOLDER: citable record pending. https://github.com/adamaske/nirwizard-releases.git

PLACEHOLDER: nirwizard-releases.git

Questions

Does it cost anything?
No. NIRWizard is free for research, teaching, evaluation and private study, on as many machines as you like. Commercial use, such as reselling it, running it as a paid service or embedding it in a product, needs a separate licence, granted on request.
Which platforms does it run on?
Windows 10 and 11 (x64) and 64-bit Linux, as an AppImage, .deb or .rpm. There is nothing to build.
What does the “tester” in the file name mean?
Pre-1.0 builds are marked tester: the same application with the licence key screen switched off, so anyone can run it without a key.
Does it design montages or run experiments?
No. NIRWizard works on recordings that already exist. WizardHat designs a montage and makes the cap for it. NIRWizard does not record or run the experiment.
Do I need Python?
Only for preprocessing. Loading, quality control, analysis and the 3D viewport work without any external tools.
How do I cite it?
Cite the release you used. A citable record is pending: PLACEHOLDER, see https://github.com/adamaske/nirwizard-releases.git.