Table 2.
Advantages and limitations for deep learning methods in MRI motion correction
| Literature | Advantages | Limitations |
|---|---|---|
| Kustner et al. [17], Sommer et al. [71] | • Patch-based image to image translation networks and these methods are efficient in GPU memory and thus easier to implement in practice | • Motion artefact is of global appearance, and the patch-based methods may not be the optimal solution |
| Johnson et al. [73] | • A full 3D motion simulation and 3D DL network for motion correction | • 3D Unet is used, but the field of view coverage in slice direction is only limited to 8 slices |
| Lee et al. [79] | • Using information from multi-contrast images to correct motion artefact | • Multiple contrasts may not be available for all the studies, and the method is prone to image registration errors across multiple contrasts |
| Pawar et al. [19, 19] | • Full 3D motion simulation, large dataset for training, validation on simulated, as well as real motion degraded images in clinical setting | • Processes only 2D slices that may result in slice to slice variations when viewed from the other orthogonal plane |
| Bilgic et al. [41], Haskell et. al. [49] | • Provided estimates for both motion parameters and motion corrected images | • A two-step approach where DL is used as a preprocessing step for an iterative motion correction model, and potentially multiple sources of errors may add together |
| Ghodrati et al. [68], Terpstra et al. [75], Tamada et al. [23] | • Methods developed for dynamic MR imaging and can correct for non-rigid motion artefact including cardiac cine and DCE liver MRI | • Small dataset used for training with proof of concept validation with limited clinical evaluation |
| Khalili et al. [83], Duffy et al. [84], Gong et al. [85], Shaw et al. [74] | • Focused on practical application of motion correction methods by assessing the downstream tasks such as image segmentation, cortical surface reconstruction, and diffusion parameter estimation on motion corrected images | • Actual motion corrected images are not compared with ground truth images |