Fig. 1.

Schematic of the proposed single breath-hold 2D myocardial T1ρ mapping technique (A) with joint T1ρ fitting and model-based motion correction (B). T1ρ mapping is performed using a single-shot electrocardiogram (ECG)-triggered balanced steady-state free precession acquisition where five images with different spin lock times (TSL) are acquired over 13 heartbeats within a single breath-hold. The T1ρ preparation pulse includes 4 spin locks and 2 adiabatic refocusing pulses. Motion correction is performed by iterating between a Levenberg–Marquardt T1ρ fitting, generating a T1ρ map used for the simulation of T1ρ-weighted images. These simulated images have the same contrast than the acquired images and a pair-wise non-rigid motion estimation can thus be performed between the two sets of images. The obtained non-rigid motion fields are then used to transform the acquired images to the same motion state, and these steps are repeated until convergence