Online Spatiotemporally Constrained Reconstruction for Real-Time Interactive MRI

To develop a low-latency dynamic MRI reconstruction that provides better spatial and temporal fidelity than contemporary online reconstruction, for localization.

We utilize online spatiotemporally constrained reconstruction (STCR) using a computationally efficient augmented Lagrangian solution and a limited number of iterations that terminate before the acquisition of the next frame. Online STCR is evaluated in the setting of fetal cardiac localization at 0.55 T, and is compared with online view-sharing and offline compressed sensing, using edge sharpness metrics.

Online STCR has the highest structural similarity index measure (SSIM) to offline STCR among the compared online reconstruction methods. Spatial edge sharpness improved by 13% $\pm$ 2% and temporal edge sharpness improved by 34% $\pm$ 4% compared to view-sharing.

We demonstrate low-latency reconstruction for 2D real-time MRI at 1.5 $\times$ 1.5 mm ${\hspace{0.17em}}^2$ spatial resolution, 31.8 ms temporal resolution, and 240 $\times$ 240 mm ${\hspace{0.17em}}^2$ FOV. We demonstrate that online STCR improves the depiction of rapidly moving boundaries, compared to standard view-sharing.