Introduction

T₂^*-weighted (T₂^*w) gradient echo (GRE) MRI has been widely used in clinical research for high resolution MRI^1,2. However, high resolution T₂^*w MRI is sensitive to both motion and B₀ fluctuation, limiting its potential clinical application. We hypothesize that when using multi-shot multi-echo 3D echo-planar imaging (EPI) for T₂^*w data acquisition, the reduced scan time can improve the reliability of quantitative metrics, including R₂^* and susceptibility χ. In this study, we evaluated this in a group of subjects by comparing the test-retest reliability and accuracy of 3D EPI with 3D GRE.

Methods

MRI experiment

Experiments were performed on a 3T MRI scanner (Prisma, Siemens) with a 64-channel head-neck RF coil. Eight healthy subjects were recruited with signed consent (age: 46±20 year old, 4 males). The imaging protocol included T₂^*w multi-echo 3D GRE and multi-echo 3D EPI (Table 1). The major difference between GRE and EPI protocols was the EPI factor, which defines the number of k-space lines per echo. Acquisition bandwidth was chosen such that the EPI data had similar SNR per unit time and readout train length as the GRE data. Scans were repeated for test-retest measurement. In all scans, 3D EPI navigators were obtained with 6x5.6x15 mm³ spatial and 0.36 s temporal resolution for correction of head motion and B₀ changes³. An MPRAGE scan with isotropic 1 mm resolution was performed for segmentation and region-based group analysis.

In another experiment, 3D EPI was used to acquire T₂^*w data with nearly isotropic 0.45 mm resolution in about 10 minutes. This was performed in one healthy subject on a 7 T MRI (Terra, Siemens) with a 32-channel head coil (Nova Medical) using accelerated 3D EPI and 3D GRE protocols as shown in Table 2.

Image reconstruction

The T₂^*-weighted images were reconstructed with a custom MATLAB software which used the navigator information to perform motion and B₀ corrections ³. The images were corrected with either motion and linear B₀ correction (corrected mode) or only global average B₀ correction (uncorrected mode). Eddy current-related 1D phase correction was performed for EPI data.

Data analysis

The decay rate R₂^* was calculated using least square nonlinear fitting in MATLAB. Susceptibility χ was calculated using the JHU/KKI QSM toolbox⁴. These maps were aligned based on the T₂^*w magnitude data to correct for inter-scan motion.

Test-retest reliability was calculated as the difference of R₂^* and χ between two measurements. For EPI, each measurement contained averaged data from 3 or 5 scans according to the EPI factor in order to match SNR to the GRE data. In addition, agreement of R₂^* and χ between GRE and EPI methods were evaluated based on the mean value of different deep brain regions of interest (ROI) in the MNI space.

Results

Fig. 1 shows that with higher EPI factor or equivalently shorter single scan duration, a larger portion of voxels exhibited reduced inter-scan difference of R₂^* and χ and thus improved test-retest reliability. This trend applies to both corrected and uncorrected data. The average SNR in the GRE and averaged (averages of 3 or 5) EPI data at TE=14 ms was estimated to be 60 based on the noise measurement. The noise-equivalent inter-scan difference of R₂^* and χ based on average R₂^*=18 s^-1 were estimated and indicated in Fig. 1. This confirmed that the test-retest reliability of GRE result is not only affected by thermal noise but also more by other sources compared to the EPI data.

Fig. 2 illustrates the level of consistency of R₂^* and χ between corrected EPI (x5) and GRE data. In Fig. 2A, the group-average R₂^* difference is less than 2 s^-1 without significant residual anatomical information visible except in regions where the MRI signal is low (skull) or signal fluctuation is high (ventricles and large vessels). In Fig. 2B, the group-average χ difference is low, on the order of 0.005 ppm, but with slight systematic difference, e.g. in the white matter area. We suspect that this is due to imperfect EPI phase correction for the bipolar readout gradient or different sensitivity in the MR phase signal to the applied readout gradient. Overall, ROI-average values showed high correlation, with a correlation coefficient of 1.0 and a slope of 1.000 ± 0.003 for R₂^* and 1.0 and 1.004 ± 0.008 for χ, respectively.

At 7 T, the high resolution T₂*-weighted EPI image demonstrated similar anatomical details and brain morphology compared to the GRE image as shown in Fig. 3. It is worth noting that even with only 10 minutes of scan time, there appears to be sufficient SNR in the EPI image.

Discussion

We demonstrated improved test-retest reliability of R₂^* and χ quantification using multi-shot multi-echo 3D EPI sequences compared to 3D GRE, which is the commonly used acquisition method. The observed gain is attributed to reduced motion, physiology and instrument instability occurring in reduced scan time. This gain occurred while preserving measurement consistency. A very small systematic difference in χ between methods was noticed and requires further investigation to determine its origin.

Conclusion

Multi-shot multi-echo 3D EPI can be a useful alternative acquisition method for quantitative T₂^*-weighted MRI with improved reliability and similar accuracy compared to 3D GRE in clinical applications.

Acknowledgements

This work was partly supported by a faculty startup fund of UT Southwestern Medical Center and the intramural research program of the National Institute of Neurological Disorders and Stroke.

References

[1] J. H. Duyn, P. van Gelderen, T.-Q. Li, J. A. de Zwart, A. P. Koretsky, and M. Fukunaga, “High-field MRI of brain cortical substructure based on signal phase,” Proc. Natl. Acad. Sci. U.S.A., vol. 104, no. 28, pp. 11796–11801, Jul. 2007, doi: 10.1073/pnas.0610821104.

[2] E. M. Haacke, S. Mittal, Z. Wu, J. Neelavalli, and Y.-C. N. Cheng, “Susceptibility-Weighted Imaging: Technical Aspects and Clinical Applications, Part 1,” AJNR Am J Neuroradiol, vol. 30, no. 1, pp. 19–30, Jan. 2009, doi: 10.3174/ajnr.A1400.

[3] J. Liu, P. van Gelderen, J. A. de Zwart, and J. H. Duyn, “Reducing motion sensitivity in 3D high-resolution T₂^*-weighted MRI by navigator-based motion and nonlinear magnetic field correction,” Neuroimage, vol. 206, p. 116332, 2020, doi: 10.1016/j.neuroimage.2019.116332.

[4] X. Li et al., “Multi-atlas tool for automated segmentation of brain gray matter nuclei and quantification of their magnetic susceptibility,” Neuroimage, vol. 191, pp. 337–349, May 2019, doi: 10.1016/j.neuroimage.2019.02.016.