An open implementation of FSL’s eddy with volume-to-volume artifact estimation for neuroimaging beyond diffusion MRI

Motivation: Create an open-source alternative to FSL's eddy without commercial limitations that generalizes over problems beyond eddy current distortions of diffusion MRI. Goal(s): Evaluate the tool's performance vis-à-vis with corresponding tools. Approach: We implement a Gaussian Process regressor model, a Bayesian optimization layer on an existing registration framework and estimate the alignment of neuroimaging data enabled by the eddymotion framework. Results: Using publicly available neuroimaging data, we provide evidence about the effectiveness of our volume-to-volume modeling framework for generalized artifact identification and correction in neuroimaging. Impact: We present eddymotion, an open-source framework for volume-to-volume artifact estimation inspired by FSL eddy. Our tool allows for easy alternative model implementation, it can be used for non-dMRI imaging modalities and does not have a restrictive license.