An open-source framework for volume-to-volume motion estimation in d/fMRI and PET, and Eddy-current-derived distortion estimation in dMRI.
Important
NiFreeze is a fork of eddymotion
In November 2023, the NiPreps Steering Committee brought to the Bi-monthly Roundup the discussion about re-branding eddymotion to better reflect its aspirations to perform on diverse modalities.
The repository of the project has been archived, and development will continue under the NiFreeze project.
The contributor list of eddymotion is found under the credit file
.maint/EDDYMOTION.md in this repository.
Diffusion and functional MRI (d/fMRI) generally employ echo-planar imaging (EPI) for fast
whole-brain acquisition.
Despite the rapid collection of volumes, typical repetition times are long enough for head motion
to occur, which has been proven detrimental to both diffusion [1] and functional [2] MRI.
In the case of dMRI, additional volume-wise, low-order spatial distortions are caused by
eddy currents (EC), which appear as a result of quickly switching diffusion gradients.
Unaccounted for EC distortion can result in incorrect local model fitting and poor downstream
tractography results [3], [4].
FSL's eddy [5] is the most popular tool for EC distortion correction, and
implements a leave-one-volume-out approach to estimate EC distortions.
However, FSL has commercial restrictions that hinder application within open-source initiatives
such as NiPreps [6].
In addition, FSL's development model discourages the implementation of alternative data-modeling
approaches to broaden the scope of application (e.g., modalities beyond dMRI).
NiFreeze is an open-source implementation of eddy's approach to estimate artifacts
that permits alternative models that apply to, for instance, head motion estimation in fMRI
and positron-emission tomography (PET) data.
| [1] | Yendiki et al. (2014) Spurious group differences due to head motion in a diffusion MRI study. NeuroImage 88:79-90. |
| [2] | Power et al. (2012) Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage 59:2142-2154. |
| [3] | Zhuang et al. (2006) Correction of eddy-current distortions in diffusion tensor images using the known directions and strengths of diffusion gradients. J Magn Reson Imaging 24:1188-1193. |
| [4] | Andersson et al. (2012) A comprehensive Gaussian Process framework for correcting distortions and movements in difussion images. In: 20th SMRT & 21st ISMRM, Melbourne, Australia. |
| [5] | Andersson & Sotiropoulos (2015) Non-parametric representation and prediction of single- and multi-shell diffusion-weighted MRI data using Gaussian processes. NeuroImage 122:166-176. |
| [6] | Esteban (2025) Standardized preprocessing in neuroimaging: enhancing reliability and reproducibility. In: Whelan, R., & Lemaître, H. (eds.) Methods for Analyzing Large Neuroimaging Datasets. Neuromethods, vol. 218, pp. 153-179. Humana, New York, NY. doi:10.1007/978-1-0716-4260-3_8. |