Introduction

We recently proposed an ultrafast 3D gradient-echo imaging sequence using spatiotemporal encoding (SPEN) with constant TE, called ERASE (Equal-TE Rapid Acquisition and Sequential Excitation)¹, which has high immunity to B₀ inhomogeneities. We previously reported that ERASE provides higher SNR, tSNR and t-score maps than multislice 2D gradient-echo EPI (GE-EPI) when thermal noise is more dominant than physiological noise,¹ e.g., in high spatial resolution. In this work, we examined the noise characteristics such as the ratio of physiological and thermal noise (σ_P/σ₀), lambda (λ), SNR and tSNR of ERASE in in vivo rat brain imaging at 9.4T, in comparison with 2D-EPI and 3D-EPI.

Method

Sequence: Figure1 illustrates the ERASE sequence diagram (A) and schematic representation of its sequential and local excitation and rephasing mechanism in the SPEN direction (B). ERASE provides a constant TE across slices via the insertion of a rephasing gradient between spin excitation and acquisition processes along the SPEN direction.1 Experiment: In vivo rat brain imaging was performed at 9.4T (BioSpec, Bruker). Four male Sprague Dawley (SD) rats (280–310 g, 8–9 weeks of age) were used with the approval of the Institutional Animal Care and Use Committee (IACUC) at Sungkyunkwan University. Three imaging sequences of ERASE, 2D-EPI and 3D-EPI were simultaneously applied to each rat and 30 volumes were acquired with each sequence. Scan parameters are summarized in Table 1. ERASE was reconstructed offline with MATLAB (ver.8.2.0; R2013b) using the super­resolution(SR) algorithm by Chen et. al.3 in the SPEN direction.
Data analysis: To reduce the effect of head motion, image registration in each time series was performed with a reference to the first image using SPM8 motion-correction software. Noise characteristics were analyzed based on evaluation of SNR and tSNR. The ratio of physiological thermal noise (sP/s0) was evaluated to assess the relative contribution of physiological and thermal processes to each imaging technique by the following equation:
$$\frac{\sigma_{p}}{\sigma_{o}}=\sqrt{(\frac{SNR_{o}}{tSNR})^{2} - 1} .[1]$$
Here, SNR0 was calculated pixelwise by dividing each pixel value by the standard deviation of background noise (s0). tSNR was also calculated pixelwise by dividing the mean pixel value across the 30 time-series images by the temporal fluctuations (sF), where σ_F² = σ_P² + σ₀². On the other hand, σ_P is commonly expressed as λ×S (λ, constant; S, the temporal mean signal)⁴, where λ represents a measure of SNR degradation by signal-dependent fluctuations and can be determined by the relationship between tSNR and SNR₀:^4,5
$$tSNR=\frac{SNR_{0}}{\sqrt{1+\lambda^{2}SNR_{0}^{2}}} .[2]$$

Results

Fgiure 2A shows presents the tSNR and SNR maps of three center slices selected from rat brain images obtained using 2D GE-EPI, 3D GE-EPI and ERASE. Image intensities were scaled within a range of 0–60 in the SNR map, and 0–30 in the tSNR map. For better representation, the scatter plots of all tSNR and SNR values in the whole brain were also presented (Fig.2B). Overall, ERASE provided better tSNR than 2D and 3D GE-EPI, whereas it provided intermediate SNR between them. In Fig.3A, the mean SNR and tSNR values were plotted in a bar graph by averaging all the results over the whole brain, as obtained from four rats. σ_P/σ₀ and l were also presented in a bar graph (Figs. 3B, C). The σ_P/σ₀ and l values of ERASE were lower than 2D and 3D GE-EPI.

Discussion & Conclusion

Here, we compared ERASE with 2D and 3D GE-EPI in terms of noise characteristics, SNR and tSNR ERASE provided better tSNR than both of them and intermediate SNR between them, and showed lower σ_P/σ₀ and l values, which closely relate to physiological noise, than both 2D and 3D GE-EPI. It is interesting to note that the physiological noise contribution to ERASE is less than 2D GE-EPI because 3D GE-EPI is generally known to have more physiological noise than 2D GE-EPI.5 In addition, from the fact that l showed a similar value in 2D and 3D GE-EPI, it is inferred that thermal noise was dominant over physiological noise due to high spatial resolution in this study. The reason why ERASE had less physiological noise contribution might be understood in the following respects: 1) Less sensitivity to the breathing-induced B₀ modulation over the whole brain owing to its robustness against B₀ inhomogeneity. 2) Less sensitivity to EPI artifacts such as Nyquist ghosting and image distortion, which are the main sources of signal fluctuations. A further study is needed to investigate the dependence on a field strength and spatial resolution.

Acknowledgements

This work was supported by NRF-2019M3C7A1031993 and IBS-R015-D1.