Cartesian vs. non-cartesian under-sampling for high resolution 3D double echo steady state (DESS) knee imaging at 7 T

Motivation: Double-echo-steady-state (DESS) is an attractive 3D sequence for musculoskeletal imaging at 7T, facilitating isotropic high-resolution knee cartilage imaging. Long acquisitions necessitate advanced undersampling techniques with high acceleration factors, while the optimal scheme and its limits are unknown. Goal(s): Compare image quality (IQ) of cartesian and non-cartesian undersampling with up to 30-fold acceleration (R). Approach: IQ evaluation in DESS images of 40 knees acquired with varying undersampling factors (R=4-30) and schemes (GRAPPA, CAIPIRINHA, Compressed Sensing (CS)) at 7T. Imaging data were reconstructed with a deep-learning (DL) algorithm. Results: 16-fold CS (01:12min:sec) showed no significant IQ difference to 4-fold GRAPPA (5:22min:sec) and 8-fold CAIPIRINHA (2:22min:sec). Impact: The results suggest significant advantages of non-cartesian versus cartesian undersampling strategies in DESS imaging with high acceleration factors at 7T. The short acquisition times may allow clinical high-resolution 3D knee cartilage imaging, e.g., for quantification of cartilage morphology.