T2*-weighted (T2*w) gradient-echo (GRE) sequences are commonly used in neuroimaging to depict hemorrhage, calcification and iron deposition. Two-dimensional (2D) GRE sequences are often employed for T2*w imaging. However, compared to three-dimensional (3D) GRE sequences, 2D GRE sequences are more sensitive to the deleterious T2* effects at air-tissue interfaces and implants due to the thicker slices used in 2D imaging¹. 3D Cartesian high-resolution T2*w GRE sequences usually require long scan times, because of the preferred long TRs and TEs. In this study, we propose to improve the scan efficiency of 3D T2*w imaging with a distributed spiral² in-out trajectory.

A schematic diagram of the GRE sequence with spherical distributed spiral² (SDS) in-out readout is shown in Fig. 1 (a). When coronal or sagittal scans are performed, the slab selection gradient (denoted as s_ex in the diagram) is moved from the slice-encoding axis to one of the readout axes. In other words, the slab is always excited in the axial direction, whereas the spiral acquisition changes for axial, coronal or sagittal scans. 3D and 2D views of one interleave of the spiral trajectory are shown in Fig. 1 (b) and (c), respectively.

Axial and coronal volunteer data were collected on a 3T Philips Ingenia scanner with a 15-channel head coil. In all scans, TE was 16 ms. Cartesian 2D and SDS data were acquired for axial slices. The parameters for the 2D Cartesian scan include: FOV = 230×178×139 mm³, resolution =1×1 mm², slice thickness = 4 mm with 1mm gap, slice number = 28, flip angle = 18°, TR = 822 ms, and readout time = 4.62 ms. The 3D spiral scan covered the same anatomical region as the Cartesian scan, with a resolution of 1×1×2 mm³, TR of 38 ms and a readout time of 19.59 ms. The flip angle was adjusted³ to 3.8° to keep the T1 effect similar to the 2D Cartesian scan. The scan time was 2:30 for the 2D Cartesian scan and 3:04 for the spiral scan respectively. SDS and 3D Cartesian data were collected for coronal slices with a FOV of 190×190×210 mm³ and resolution of 1×1×2 mm³. In the SDS scan, TR was 37 ms and the flip angle was 3.8°. TR was 26 ms and the flip angel was 3.2° for the Cartesian scan. Pure noise data sets were also acquired for signal-to-noise ratio (SNR) evaluation⁴. In all the spiral scans, data were acquired at two TEs with the second TE shifted by a ΔTE=1.15 ms. Water and fat images were extracted and deblurred from two spiral-in images and two spiral-out images separately with a map of B0 inhomogeneity obtained from a low-resolution scan⁵, which were then averaged to form the combined in-out images.

The results of axial scans are shown in Fig. 2. The spiral water images are shown in the middle column and the water/fat combined images are shown in the left column. The severe signal loss (indicated by the yellow arrows) near the air-tissue interfaces is mitigated in the 3D spiral images because of the thinner slices. Coronal images reconstructed from spiral-in part, spiral-out part, and the in-out combined images are compared to the Cartesian images in Fig. 3. The spiral-out images demonstrate a slightly higher T2* contrast than the spiral-in images. SNR estimated from SDS water images and Cartesian images in the coronal scans, as well as the scan efficiency (defined as SNR/scan time^1/2) are summarized in Fig. 4. The spiral-in and spiral-out images have similar SNR. The combination of the two increases the SNR by a factor close to $$$\sqrt{2}$$$. The SDS in-out imaging achieves similar SNR as that of the Cartesian scan, even though it is more than 2 times faster than the Cartesian scan. It is also worth noting that the data collected in the 3D Cartesian scan are elliptical in the k_y-k_z plane, so that the actual scan voxel is slightly larger than that of the spiral scans. The improvement of scan efficiency with spiral imaging would be even more significant if this factor was taken into account.This work has demonstrated the GRE SDS in-out imaging as a fast, scan efficient sequence for 3D T2*w imaging.