Introduction

Diffusion spectrum imaging (DSI) is a model-free diffusion MRI technique capable of resolving complex fiber orientations [1]. Previously, the DSI protocol within clinically acceptable scan time has been established and optimized on 3T MRI [2]. Furthermore, a comparative diffusion tensor imaging (DTI) study has demonstrated that higher field strength is advantageous for DTI by assessing effects of SNR on DTI quantitative metrics [3]. However, the capability of DSI at different field strength and spatial resolution remains unexplored. In this study, we aimed to investigate the effects of magnetic field strength and spatial resolution on the reproducibility of DSI. A multi-site and repeated DSI study have been conducted at 3T and 7T. The reproducibility has been quantified by orientation distribution function (ODF) and deviation angle, and further evaluated by statistical comparisons.

Materials & Methods

The overall workflow of this study is shown in figure 1. A total of 8 healthy subjects were recruited in this study. MR experiments were performed on a 7T MRI system (5 subjects; Magnetom 7T, Siemens Healthineers, Erlangen, Germany) and a 3T MRI system (3 subjects; Prisma 3T, Siemens Healthineers, Erlangen Germany). For each subject, the MR experiments were performed twice to obtain two DSI datasets (scan#1 and scan#2). DSI data was acquired by using the diffusion-weighted spin-echo EPI sequence. A DSI protocol with 203 grid sampling points and maximum b-value of 4500 s/mm2 was used within clinically acceptable scan time. We acquired three different voxel sizes (2.5 mm, 2.0 mm, and 1.5 mm isotropic resolution) on both systems. The FOV (mm) / matrix size / slice number are 240 / 96 / 36 for 2.5-mm resolution, 216 / 108 / 46 for 2.0-mm resolution, and 216 / 144 / 60 for 1.5-mm resolution. The TR and TE were optimized by using the parallel imaging (GRAPPA) on both systems. On 7T, TR/TE (ms) are 4400/91 for 2.5 mm, 5800/92 for 2.0 mm, and 7700/94 for 1.5 mm. On 3T, they are 5500/59 for 2.5 mm, 7800/60 for 2.0 mm, and 10800/61 for 1.5 mm. All diffusion-weighted images (DWIs) were pre-processed using FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki), including eddy current distortion and motion correction, co-registration and normalization of T1-weighted images (T1WIs) onto MNI space. DSI reconstruction was performed on DSI Studio to obtain the voxel-wise ODF, major fiber orientations (the first five in our case), and fractional anisotropy along the maximum ODF (FA_ODFMAX) (http://dsi-studio.labsolver.org/). The white matter (WM) region was segmented by T1WIs and a FA_ODFMAX threshold (0.3 in our case). All voxels within the WM region were divided into two groups, i.e. single-fiber and crossing-fiber, based on the ratio of lengths of second ODF and the maximum ODF (0.4 in our case). Two quantitative metrics, i.e. orientational similarity (SIM) and deviation angle (Angle_Dev) for assessing the reproducibility in both fiber groups were calculated.

Results

Figure 2 shows the maps of major fiber orientations of DSI among different spatial resolutions and field strengths. For each field strength, the enlarged views in fiber-crossing regions show that the reconstructed fiber orientations are visually consistent between two scans in each spatial resolution. Compared with lower spatial resolution, higher spatial resolution reveals more details of intra-voxel fiber orientations. Figure 3 shows the voxel-wise SIM map and plot of averaged SIM among different spatial resolutions and field strengths. Generally, SIM of single-fiber groups (0.9 – 0.95) are consistently higher than SIM of crossing-fiber groups (0.65 – 0.8). In single-fiber group, there are no obvious differences of SIM between different spatial resolutions and field strengths. In crossing-fiber group, SIM tends to increase as the voxel size becomes larger for both field strengths. By comparing 3T and 7T, higher field strength yields relatively higher SIM among all spatial resolutions. Figure 4 shows the voxel-wise Angle_Dev map and plot of averaged Angle_Dev among different spatial resolutions and field strengths. Generally, Angle_Dev of single-fiber groups (5 – 10 degrees) are consistently lower than Angle_Dev of crossing-fiber groups (20 – 35 degrees). Similar as SIM results, in single-fiber group, there are no obvious differences of Angle_Dev between different spatial resolutions and field strengths. In crossing-fiber group, Angle_Dev tends to decrease as the voxel size becomes larger for both field strengths. By comparing 3T and 7T, higher field strength yields relatively lower Angle_Dev among all spatial resolutions except 1.5 mm.

Discussion and Conclusion

To our best knowledge, this is the first study to quantitatively investigate the effects of field strengths and spatial resolutions on DSI reproducibility. Our results show that DSI at 3T and 7T for single-fiber group are comparable, whereas DSI at 7T could provide relatively better reproducibility than that at 3T for crossing-fiber group. Among different spatial resolutions, no differences of reproducibility have been found in single-fiber group. However, our results of SIM and Angle_Dev for crossing-fiber group show that larger voxel size has incrementally better reproducibility than smaller voxel size, suggesting that partial volume effect may be less dominant than signal-to-noise ratio (SNR). A future work to control the SNR by manipulating the sequence parameters is needed to further understand the effects of partial volume and SNR on DSI reproducibility and their associations with field strengths and spatial resolutions.

Acknowledgements

This work was supported by the grants from the Ministry of Science and Technology (MOST-109-2221-E-400-001-MY2), the National Health Research Institutes (NHRI-BN-109-PP-06 and NHRI-BN-109-SP-11), and the Japan Society for the Promotion of Science (19K22985).