3D self-gated (SG) cine imaging with TrueFISP not only provides excellent contrast between myocardium and blood, but also eliminates the need for ECG set up and permits free-breathing acquisitions [1]. However, such golden-angle radial sampling-based techniques are commonly used at 1.5 T due to the eddy current and SAR problems as well as time-consuming on data acquisition under the Nyquist sampling criteria. To achieve time-efficient 3T cine imaging, a novel accelerated SG method, named SparseSG, was proposed using a tiny golden angle and compressed sensing [2].

Theory: A 3D hybrid radial sampling pattern was adopted for the SparseSG [1]. In order to reduce the eddy current effect, a tiny golden angle of 32.039°, instead of 111.246°, was used for data acquisition (Fig.1). After the respiratory and cardiac motions were determined by processing the SG data as [1], the acquired data was retrospectively sorted into different respiratory and cardiac phases. A compressed sensing method exploiting the image sparsity in k-t space by solving a constrained convex optimization problem as Eq. (1) was then used for image reconstruction, thus effectively shortening the scan time and reducing SAR.

$$arg min \left\{{λ ‖T\cdotρ‖_1 }\right\} subject to ‖d-P\cdot F\cdotρ‖_2^2< ε (1)$$

where P is sampling matrix, F is the NUFFT operator defined on the radial acquisition pattern, ρ denotes is the image series to be reconstructed in x-y-t-coil space, d is the acquired radial k-t-coil space data, T is the temporal total-variation (TV) operator (sparsifying transform), imposed on the l₁ norm,and λ is the regularization weight that controls the tradeoff between the data consistency and sparsity. The SparseSG reconstruction was initially implemented in MATLAB, using a tailored version of bregman algorithm [3].

IRB-approved cardiac imaging was performed on 5 healthy subjects (2M, 3F, age 20~26) at 3T (Siemens Tim Trio, Germany) with a standard 6-channel body coil and a spine coil. Scan parameters included: 3D imaging with standard short-axis, TR=3.8ms, TE =1.9ms, spatial resolution = 1.3×1.3×8.0 mm³, bandwidth =1502 Hz/Pixel, partition number = 10. The acceleration factors were R = 4 and 8, corresponding to scan time 0.76 min and 0.38 min. The standard breath-hold 2D multi-slice ECG-triggering and conventional self-gating methods with the same spatial and temporal resolutions were also conducted for comparison.All MR scans were successfully conducted. SparseSG allowed a whole-heart coverage of 3D cine imaging within 1 min, which was much shorter than those of the standard ECG-triggering and conventional SG methods (Fig.2.b). As the acceleration factor increased, the reconstructed cine images of SparseSG become a little blurry. However, the left ventricle ejection fraction (LVEF) and cardiac structure obtained from SparseSG were in good agreement with those from standard ECG-triggering and conventional SG methods, even if a higher acceleration factor R=8 was used (Fig.2.a&c). An accelerated SG technique, SparseSG, was developed to realize 3D cardiac cine imaging at 3T without ECG and breath-holding. Preliminary in vivo study demonstrated that a whole heart coverage of 3D cine imaging can be achieved within 1 min and the technique had excellent performance compared to the standard ECG-triggering and conventional SG methods. This warrants further evaluation of SparseSG on more volunteers and patients.