pyMocoKit

MRI reconstruction pipeline with retrospective motion correction and option to reverse a prospectively applied correction. Supports raw data (Siemens .dat files) from MPRAGE and T2-SPACE sequences (an extension to diffusion EPI support is planned).

Includes: noise prewhitening, OS removal, iPAT/GRAPPA, non-Cartesian gridding (KbNUFFT), PF/POCS, coil combination, NIfTI export.

Installation

Option 1: conda

conda env create -f environment.yml
conda activate moco
pip install -e .

Option 2: pip

pip install -e .

# For GPU
pip install -r requirements-gpu.txt
pip install -e ".[gpu]"

Usage

Basic usage:

mocokit -i /path/to/folder/dat \
    -tcl -td /path/to/tcl_dir \
    -reverse -smooth \
    -orig -center \
    -device cuda:0 \
    --cuda-visible-devices 0 \
    --headless \
    --numpy-precision 6 \
    -v

Parameters

• -i: Input directory containing .dat files
• -tcl: Enable TCL processing
• -td: TCL directory path
• -reverse: Reverse motion correction
• -smooth: Apply smoothing
• -orig: Use original coordinates
• -center: Center reconstruction
• -device: Specify compute device (e.g., cuda:0, cpu)
• --cuda-visible-devices: Set visible CUDA devices
• --headless: Run without GUI
• --numpy-precision: Set numerical precision
• -v: Verbose output

Requirements

• Python 3.10+
• CUDA-compatible GPU (optional)
• Required packages listed in environment.yml

Citation

If you use this repository, please cite:

Zariry Z, Lamberton F, Frost R, Gaass T, Troalen T, Rayson H, Slipsager JM, Richard N, van der Kouwe A, Bonaiuto J, Hiba B. An in-vivo approach to quantify intra-MRI head motion tracking accuracy: comparison of markerless optical tracking versus fat-navigators. medRxiv [Preprint]. 2025 Jul 17:2025.04.23. DOI: [10.1101/2025.04.23.25326185] (https://www.medrxiv.org/content/10.1101/2025.04.23.25326185v2)