Myocardial perfusion imaging offers promising means to evaluate and characterize myocardial tissue health. High spatial resolution is desired in order to better identify and delineate sub-endocardial ischemic regions [1]. Several methods have been proposed that combine k-space undersampling methods and advanced reconstruction methods that offer high-resolution perfusion imaging [2-4]. Here we use undersampled radial simultaneous multi-slice (SMS) imaging with constrained reconstruction to obtain high in-plane spatial resolution perfusion imaging. The proposed method is tested in a resolution phantom and in two different types of perfusion acquisitions (i) a set of 3 simultaneous slices after a saturation pulse, repeated several times per beat at different cardiac phases, and (ii) a ‘hybrid’ perfusion acquisition with one saturation pulse per beat.

A saturation recovery radial simultaneous multi-slice imaging sequence was used to acquire data in a resolution phantom and in human subjects. The resolution phantom was 3D printed with line thickness varying from 1.25 mm to 3 mm and was placed in a 12 mM CuSO4 solution [5]. Myocardial perfusion data was acquired in vivo with (i) a multi-cardiac phase scheme in which the application of a saturation pulse and acquisition of multiple slices is repeated within a heartbeat to acquire the same set of slices in different cardiac phases (ii) a ‘hybrid’ ECG gated scheme where after every saturation pulse data is continuously acquired with the number of radial rays that can fit in a R-R interval. This scheme allows for flexibility of obtaining perfusion images at systole and at diastole post acquisition. Graphical descriptions of schemes (i) and (ii) are shown in Figure 1.

All of the data was acquired on a Siemens 3T Prisma scanner with TR=2.7 msec, TE=1.6 msec, number of slices=3, slice acceleration factor=3, distance between slices=8mm, and golden ratio based angular spacing. Other scan parameters are shown in Table 1.

Reconstruction from undersampled radial SMS data was performed in in a POCS framework [6,7] that uses temporal TV denoising and BM3D spatial denoising [8] while preserving fidelity to the acquired SMS data. Reconstruction steps involved projecting the estimated multiple slices on to the acquired SMS data followed by temporal TV and spatial BM3D denoising steps at each iteration. We chose a BM3D spatial constraint instead of a wavelet or non-local means constraint, as BM3D is a patch-based denoising method that better preserves image texture especially at lower signal levels and high spatial resolution. BM3D is a relatively fast method compared to other patch-based techniques.

Figure 2 shows a slice reconstructed from the resolution phantom data along with a schematic of line thickness for different structures. Different structures including some of the smallest structures appear to be well resolved. Figure 3 shows the post contrast images at three different phases within a heartbeat from scheme (i). Each row of slices is simultaneously excited. Figure 4 shows the results from scheme (ii), systolic and diastolic images are shown.Radial simultaneous multi-slice imaging with constrained reconstruction is a new method for high-resolution myocardial perfusion imaging that opens up a number of possibilities for new acquisition types. Promising results are shown in two different types of perfusion acquisitions. Further studies are needed to evaluate the utility and impact of high-resolution radial SMS imaging in myocardial perfusion acquisitions.