Spiral in-out readout is an efficient sampling scheme for T2-weighted (T2w) spin-echo (SE) sequences¹. Two sets of spiral in-out trajectories are typically acquired with a shift of ΔTE between the two acquisitions. Water and fat images can be successfully extracted and deblurred from two spiral-in images and two spiral-out images separately² and then combined. This method is referred to as the separate deblurring hereafter for simplicity. A new method (referred to as the joint deblurring hereafter) has been recently proposed to reconstruct water and fat images from a single set of spiral in-out data³, with a time delay ΔTE_io inserted between spiral-in and the spiral-out parts. The purpose of this work is to investigate the feasibility of reducing the scan time of the spiral T2w SE (by a factor of 2) with the joint deblurring by both experiments and numerical simulation.

Two volunteer data sets were collected on a 3T Philips Ingenia scanner, with one and two sets of spiral in-out acquisitions respectively. ΔTE was 1.15 ms for the latter. The readout time was 20.7 ms, including a ΔTE_io of 1.0 ms. Other scan parameters include: FOV = 230×230 mm², resolution =1×1 mm², slice thickness = 4 mm, slice gap = 1mm, slice number = 20, TR = 2350 ms and TE = 80 ms. A map of B0 inhomogeneity (Δf0) was obtained by a Cartesian mDixon scan. The spiral-in and spiral-out images were reconstructed after corrections for trajectories⁴ and B0 eddy currents^5-6. Water and fat images were reconstructed from the two data sets by the joint deblurring and the separate deblurring, respectively. In the joint deblurring, a slowly varying phase map (PM) was evaluated from the first-pass results³ by low-pass filtering (applying 20% Hann window in the k-space), which was removed from water and fat images in the subsequent process.

The water and fat images from the separate deblurring in the experiment were utilized to simulate the spiral in-out k-space data by the discrete Fourier transform. Noise was added to the simulated k-space data to characterize the dependence of the signal-to-noise ratio (SNR) on ΔTE_io. Furthermore, the influence of T2 decay was also evaluated in the simulation. The water images were approximately segmented to white matter, gray matter and cerebrospinal fluid (CSF) by thresholding. T2 was assigned 70 ms for white matter, 100 ms for gray matter, 2000 ms for CSF as shown in Fig. 1. T2 was assumed 70 ms for fat. A constant T₂' (the relaxation rate due to field inhomogeneity across a voxel) of 50 ms was also included in the simulation.

The water images reconstructed by the joint deblurring demonstrate similar overall sharpness and contrast in comparison with that of the separate deblurring (Fig. 2). The SNR reaches the highest value when ΔTE_io is between 1.1 and 1.3 ms (Fig. 3). The optimal SNR of the joint deblurring is about $$$\sqrt{2}/2$$$ of the optimal SNR of the separate deblurring that uses twice the scan time. Therefore, the scan efficiency (SNR/scan time^½) is comparable between the two methods.

The difference between the reconstructed images from the simulated data with and without T2 decay is negligible, which implies that the error due to T2 decay is insignificant. However, there is subtle but observable deviation in the reconstructed fat images when the evaluated PM (PM_eval) is used instead of the known PM (PM_true) as shown in Fig. 4. Therefore, inaccurate PM estimation may in part contribute to the artifacts in the fat images of the joint deblurring in the experiment.

This work has demonstrated the joint deblurring as a promising means of fast, scan efficient fat suppression for spiral T2w SE imaging.