Introduction

Echo planar imaging (EPI) is widely used in the MR community, and its relatively high temporal resolution makes it particularly adept for numerous dynamic MR applications, such as fMRI or DWI. Recent advances in EPI techniques extend its use for the depiction of neuronal activation with a cortical depth-dependence at ultra-high fields. Therefore, enhanced performance with reliable image quality is highly desirable in EPI. For these applications, a commercial, vendor-supplied, EPI sequence is usually employed. However, for advanced imaging options, which cannot be supported by the commercial EPI sequence, a custom-made EPI sequence is often developed, and, of these, the one from CMRR (Center for Magnetic Resonance Research) is the most widely used.^1-4 In this work, we present our own in-house designed EPI sequence (INM4-EPI), developed in a Siemens environment⁵, and evaluate its performance in comparison to two other widely used EPI sequences: Siemens-EPI and CMRR-EPI. For this purpose, this work configures several imaging protocols at 7T and demonstrates the improved performance offered by our in-house designed sequence and online image reconstruction software.

Methods

The in-house designed EPI sequence was developed using the standard Siemens sequence development platform, IDEA (Integrated Development Environment for Applications) VE12U. The development was based on the manual implementation of all of the sequence components necessary for 2D gradient-echo EPI, but without the Siemens EPI sequence as a basis template code. The sequence was designed to support the combination of parallel imaging, partial Fourier and multi-band techniques, in which the data reconstruction for the first two techniques was handled by the default reconstruction routine from Siemens. The reconstruction of the multi-band technique was performed by our in-house implemented ‘Slice-GRAPPA’ method,⁶ which was subsequently integrated into the default Siemens reconstruction routine, i.e. ICE (Image Calculation Environment). The performance of the three EPI sequences (Siemens, CMRR: version R016 and INM4) was evaluated in a typical fMRI setting: TR = 2500 ms, FA = 90°, FOV = 210 × 210 mm² and sequence type = 2D gradient-echo. Various imaging protocols were configured with a number of matrix sizes and acceleration conditions (Fig. 1). Data sets from a uniform spherical phantom and two healthy subjects, screened with a standard safety procedure, were acquired at a Siemens Magnetom Terra 7T scanner with a 1-Tx/32-Rx head coil. The performance difference of the three sequences was investigated by applying our in-house implemented reconstruction routine to the Siemens and CMRR data. Finally, the image reconstruction times of the three EPI sequences were compared.

Results

Figure 2 shows the reconstructed phantom images obtained from four accelerated protocols. The figure shows that for all resolution cases, aliasing artefacts arising from the acceleration techniques were clearly eliminated in the INM4 data, whereas a significant amount remain for the Siemens and CMRR data, causing substantial signal loss. The degree of aliasing artefacts in these two sequences was much greater when a larger parallel imaging factor was applied (i.e. 2nd row; three-fold) than a smaller one (i.e. 1st raw; two-fold). Here, particularly for the ‘Protocol D’ (1.0 mm), which was shown to be the worst-case in terms of the aliasing artefacts, the Siemens and CMRR data were reconstructed with our in-house implemented image reconstruction function. As a result, the aliasing artefacts observed in the two sequences were significantly reduced in the new reconstructed images (3rd and 4th columns in Fig. 3) and were observed to be very comparable to the image from INM4-EPI (5th column in Fig. 3). Figure 4a shows the reconstructed in vivo images obtained from ‘Protocol D’ (1.0 mm). Similar to the phantom case, aliasing artefacts or signal loss can be observed in the Siemens and CMRR data around the regions marked by the yellow arrows. Figure 4b shows in vivo data from another subject, demonstrating that the signal loss observed in the Siemens and CMRR data was significantly recovered by means of our image reconstruction routine. Figure 5 shows the image reconstruction time of the three sequences, measured for the following data sets: ‘Protocol A’ (1.6 mm), multi-band factors of 1, 2, …, 8 and 50 temporal volumes. Here, the corresponding images are shown alongside. The graph demonstrates that Siemens-EPI had the fastest reconstruction performance. However, the reconstruction time of INM4-EPI becomes comparable to Siemens-EPI as the multi-band factor increases; for a multi-band factor of 8, the reconstruction times of INM4-EPI and Siemens-EPI are very similar each other.

Discussion and conclusions

This work demonstrates the superior performance of our in-house designed EPI sequence (INM4-EPI) compared to two other widely employed EPI sequences: Siemens-EPI and CMRR-EPI. Results from the Siemens and CMRR data show that the degree of aliasing artefacts became greater for relatively large parallel imaging and multi-band acceleration conditions. However, these aliasing artefacts were effectively eliminated by our in-house implemented image reconstruction routine. Furthermore, our image reconstruction software showed image reconstruction times comparable to the Siemens case. It is expected that the improved performance demonstrated with our 2D GE-EPI in this work can be directly transferred to other types of EPI (e.g 2D diffusion-EPI, 2D SE-EPI, etc.). Additionally, future work will attempt to repeat these investigations on other 7T scanners to determine if the improvements are of general validity.

Acknowledgements

No acknowledgement found.