A Recurrent Neural Network Based Real-Time Reconstruction Method for MRI Navigation in Robot Assisted Intervention

Motivation: MRI navigation for intervention requires higher imaging efficiency, and deep learning-based algorithms significantly reduce reconstruction time. Goal(s): To develop a deep learning-based MRI reconstruction method suitable for interventional navigation tasks. Approach: This study proposes an unfolded neural network to reconstruct radial sampling sequences, integrating the data consistency term with both parallel and serial network structures, and incorporating a data-sharing term at the beginning. Results: Under limited computational resources, the proposed method demonstrates superior performance compared to the comparative methods on two tested datasets. Impact: We propose a promising MRI reconstruction algorithm suitable for navigation scenarios, which will help advance our goals in MRI-guided surgeries.