3D spiral-navigated simultaneous Cine, LGE and T1 mapping under free-breathing and self-gating (CAT-SPARCS) using dual-excitation flip-angle with randomized stack-of-spiral

Background: Cardiac MR (CMR) imaging is a widely used technique that provides important diagnostic and prognostic information in cardiac diseases. T1 mapping provides additional insight into myocardial viability and fibrosis by detecting changes in myocardial tissue properties. While previous methods required separate acquisitions for Cine and T1 mapping, our group has developed a simultaneous acquisition for both[1]. Here, we extend this approach to 3D whole-heart coverage Cine, LGE and T1 mapping in a single free-running spiral sequence completed in approximately 4 minutes.

Methods: Six patients were scanned on a 3T Siemens Vida/Skyra scanner before and 4 minutes after gadolinium contrast. A 3D golden-angle variable-density stack-of-spirals trajectory was employed, with inversion-recovery RF pulses applied every 4.6 s (pre-contrast) and 2.3 s (post-contrast). To improve tissue contrast and enable quantitative mapping, two flip angles (15° and 3°) were alternated across acquisitions (Fig. 1a).

Self-gating and motion compensation[2]: Cardiac motion was extracted via principal component analysis of spiral navigator signals, enabling retrospective cardiac phase binning (Fig. 1b). Respiratory motion was corrected using 3D rigid-body registration, aligning the acquired data across respiratory states (Fig. 2a–c).

Image reconstruction: Undersampled k-space data were reconstructed using non-uniform FFT (NUFFT) followed by signal flattening and motion-corrected volume combination (Fig. 2d). A dictionary-learning framework[3,4] (k-SVD trained dictionaries) was applied, with Bloch-simulated signal evolutions for both flip angles serving as the basis functions (Fig. 2e–f). Sparse coding was used to represent voxel-wise signal evolutions relative to the learned dictionary.

T1 mapping and LGE: Voxel-wise T1* values were estimated separately from the two flip angles. These were then combined using dual-flip-angle fitting equations to generate final quantitative whole-heart T1 maps. Cine images were reconstructed from steady-state spiral data using the 3D SPARCS[2,5] algorithm. Synthetic LGE images are reconstructed, where fibrotic or infarcted tissue appears hyperintense relative to nulled healthy myocardium.

Results: Cine reconstructions provided clear visualization of both diastolic and systolic phases throughout the left ventricle (Fig. 3a). Free-breathing spiral acquisitions enabled successful generation of post-contrast T1 maps, which demonstrated consistent blood–myocardium contrast and agreed with the reference MOLLI map (Fig. 3b,d). Synthetic LGE images derived from these T1 maps revealed myocardial scar (red arrow) with close correspondence to clinical PSIR LGE images (Fig. 3c).

Conclusion: We present a 3D free-breathing, self-gated technique that enables simultaneous cine, LGE, and T1 mapping in a 4-minute scan. The method provides whole-heart coverage with image quality comparable to standard MOLLI, potentially streamlining cardiac MR by reducing scan time.

Figure 1 (a) Schematic of the dual-flip-angle stack-of-spirals acquisition with inversion recovery pulses. (b) Workflow for self-gating, including navigator-based PCA for cardiac gating.
Figure 2. Reconstruction pipeline for simultaneous 3D T1 mapping. (a) Self-gating from spiral navigator signals enables cardiac phase sorting. (b) Non-uniform FFT (NUFFT) reconstruction and signal flattening are performed. (c) 3D rigid registration corrects respiratory motion. (d) Motion-corrected volumes are aligned across respiratory states. (e) Bloch-simulated dictionary curves for dual flip angles (3° and 15°). (f) Dictionary learning framework used for sparse reconstruction of under-sampled data. (g) Generation of T1* maps for each flip angle. (h) Fitting equations combining dual flip angles. (i) Final T1 map reconstructed with whole-heart coverage.
Figure 3: Representative image reconstructions. (a) End-diastolic and systolic cine images demonstrate preserved cardiac function. (b) Free-breathing (FB) spiral post-contrast T1 maps compared with a reference breath-hold (BH) MOLLI T1 map (left), showing consistent blood–myocardium contrast (color bar: 0–800 ms). (c) Synthetic LGE images generated from FB T1 maps compared with matched clinical PSIR LGE images. The red arrow highlights a myocardial scar, which is well depicted in both methods. (d) Quantitative post-contrast T1 values for blood pool and myocardium from BH MOLLI and FB spiral T1 mapping show close agreement.