Deep Learning Reconstruction with Different Undersampling Strategies for T1ρ Mapping with Magnetization-Prepared GRE

Motivation: To accelerate knee imaging for T1ρ mapping in clinical practiceGoal(s): To evaluate deep learning reconstruction from undersampled knee imagingApproach: Three different sampling strategies were proposed to further decrease scan time and reconstruct images with deep learningResults: Deep learning reconstruction results are in good agreement with reference images. Impact: This work will allow use of novel contrasts including T1ρ to be performed within clinical workflow and improve patient diagnosis