Online Motion-Resolved 4D MRI with Pseudo-Golden-Angle Radial Acquisition and Deep Learning Reconstruction on a 1.5T MR-Linac

Motivation: The Elekta Unity MR-Linac uses pseudo-golden-angle (PGA) instead of true golden-angle (GA) radial sampling in its clinical stack-of-stars sequence, requiring modification for true GA sampling in 4D MRI. Goal(s): To develop and validate motion-resolved 4D MRI with PGA radial sampling, incorporating rapid deep learning reconstruction for routine clinical use. Approach: 4D MRI with GA and PGA sampling was tested on a dynamic phantom and a healthy volunteer. Image quality and motion measurements were compared between the two sampling methods. Results: No significant differences in image quality and motion measurements were found between GA and PGA sampling in 4D MRI. Impact: Motion-resolved 4D MRI is feasible with PGA radial sampling, presenting performance similar to GA radial sampling. The combination of PGA sampling and Movienet reconstruction would enable the integration of 4D MRI into the Elekta MR-Linac clinical workflow for adaptive radiotherapy.