Motion-Based Reconstruction (MBR) and Prior-Based Reconstruction (PBR) are compressed sensing approaches for cardiac cine MRI that achieve high acceleration factors by exploiting temporal sparsity based on a prior image. It would be appealing to reconstruct only the image variation with respect to the prior image, as this is sparser than the image itself, but this leads to low signal-to-noise ratio and has been generally avoided.
In this work we propose a novel PRior based Image VAriation (PRIVA) reconstruction method that overcomes problems previously encountered with image variation reconstruction by using an image splitting approach. We tested PRIVA in combination with a MBR method (PRIVA-MBR), where motion is estimated from a prior image and then PRIVA is used to reconstruct the image variation with respect to the prior image. The prior image was reconstructed with the SpatioTemporal Total Variation (ST-TV) method. We analyzed PRIVA-MBR in terms of solution error and maximum achievable acceleration factor that maintained acceptable image quality for a prospective cardiac cine study in small animal MRI.
The prior, given by ST-TV, presented temporal-blurring effects at high acceleration factors. PRIVA-MBR, which takes motion into account, corrected for these effects, achieving acceleration factors of x7. In conclusion, we have validated that image variation reconstruction is feasible using PRIVA, and we have shown that the PRIVA-MBR method leads to high acceleration factors for cardiac cine MRI.