1349798 - 3D joint T1/T1rho mapping and water-fat imaging at 0.55T

Native T1 and T1ρ mapping has shown potential as a non-contrast alternative to Late-Gadolinium Enhancement (LGE) to detect focal and diffuse scarring in patients with suspected myocardial infarction [1]. In [2] we proposed a novel free-breathing, 3D whole-heart joint T1/T1ρ mapping sequence with Dixon encoding to provide native 3D T1 and T1ρ maps with isotropic resolution and co-registered water and fat volumes for myocardial tissue characterisation at 1.5T. Lower T1 relaxation times (improved T1 sensitivity), reduced SAR (higher spin-locking frequencies) and fewer B0/B1 inhomogeneities (fat-water separation) make low-field MRI an attractive alternative for joint T1/T1ρ mapping. Here, we present preliminary results from this sequence modified to operate on a 0.55T scanner.

Methods:

The ECG-triggered 3D joint T1/T1ρ prototype sequence (Fig. 1) consists of a repeating set of preparation modules over 4 heartbeats (HBs): IR preparation, no preparation, no preparation and T1ρ preparation [2]. A 2-point Dixon GRE read-out (TE₁/TE₂/TR=2.6/6.5/9.6ms, flip angle=8^o, bandwidth=451 Hz/pixel) is used every HB to acquire pseudo in-phase (IP) and opposed-phase (OP) echoes. A 4x undersampled variable-density Cartesian trajectory with spiral profile order and golden angle step is used to acquire the 3D data. 2D image navigators (iNAVs) acquired prior to each spiral are used to perform beat-to-beat translational respiratory motion estimation and correction, as well as to bin the 3D data into 4 respiratory bins for estimation of bin-to-bin 3D non-rigid motion. The 8 3D image contrasts (OP and IP echoes for each HB) are reconstructed using non-rigid motion correction with patch-based multi-dimensional low-rank regularisation (HD-PROST) [3]. Water and fat images are estimated from the reconstructed IP/OP HB images [4]. EPG simulation was used to generate a dictionary with Np~6.7k T1/T1ρ parameter combinations, and mapping performed voxel-by-voxel by maximising the inner-product of the dictionary and water HB images.

Results:

Data was acquired on a 0.55T MR scanner (MAGNETOM Free.MAX, Siemens Healthcare, Erlangen, Germany). Phantoms: P</span>hantom experiments (Fig. 2) were performed by varying simulated heart rate (HR), with good precision for both T1 and T1ρ values for high HRs, and good agreement with 2D MOLLI and 2D T1ρ [5]. In-vivo: Data was acquired in one healthy volunteer (HR = 55+/-3 bpm), with an acquisition window of ~150 ms at mid-diastole. Total scan-time was ~11.7 mins (spatial resolution 2 mm isotropic). Fig. 3 illustrates water, fat, T1 and T1ρ maps estimated from dictionary matching in mid-coronal and short-axis views through the 3D reconstruction.

Conclusion:

Preliminary 3D joint T1/T1ρ mapping results demonstrate good agreement in phantoms, and promising results in-vivo at 0.55T. Further work will include validation in phantoms using 2D spin-echo reference measurements, and acquisition of a larger cohort of healthy subjects and patients with suspected myocardial infarction.