Readout Segmentation for Increased Spectral Bandwidth in High Spatial and Spectral Resolution (HiSS) MRI
David Andrew Porter1 and Marco Vicari1

1Fraunhofer MEVIS, Bremen, Germany


A novel method of echo-planar spectroscopic imaging is introduced, in which readout segmentation is used to reduce the echo spacing and provide a substantial increase in spectral bandwidth. Results are presented, showing how the technique avoids the aliasing problems that affect conventional applications of high-resolution, spectroscopic imaging at 3T and serves as a robust method for providing spectrally-selective fat and water images. The method is also a promising option for high-bandwidth, spectroscopic imaging studies of metabolites at high field strengths.


High spectral and spatial resolution (HiSS)1 MRI is a well-established technique for examining the spectral characteristics of proton signals at voxel locations in high-resolution magnetic resonance images. The method can be used to robustly separate contributions from water and fat signals2 and to probe tissue composition by analysing the spectral components of the water and fat resonances3. Data acquisition is performed using EPSI4,5 to allow rapid spatial and spectral encoding with a scan duration that is suitable for clinical application. However, EPSI is limited by the requirement that the echo spacing is sufficiently long to perform the required spatial encoding in the readout direction, resulting in a corresponding limit on the spectral bandwidth that is available. In turn, this leads to aliasing of fat or water signals, which complicates the data analysis. The problem scales with field strength due to the increasing separation of fat and water resonances. This paper introduces a readout-segmented version of EPSI, which makes it possible to perform HiSS with significantly increased spectral bandwidth, which is independent of the spatial resolution or available gradient strength; it is demonstrated how this modified technique avoids the aliasing problems that are encountered using standard EPSI at a field strength of 3T. A drawback is the increased acquisition time due to the additional scans required to acquire the multiple readout segments


Fig. 1 shows the pulse diagram for the sequence used in the study. A train of gradient echoes is sampled using a sinusoidal readout gradient, which is preceded by stepped encoding gradients in both readout and phase-encoding directions. Data were acquired from the knee of a healthy volunteer using a Siemens 3T Skyra system and a wrap-around flex coil. Imaging parameters were as follows: FOV 200mm; matrix 256 x 256; slice thickness 2mm; TR 150ms; 192 echoes with spacing 360μs, corresponding to a spectral bandwidth of 2.8kHz; scan time 8 mins. 19 secs. After acquisition and standard reconstruction of images for each echo time, Fourier transformation was applied in the echo dimension to provide a proton spectrum for each voxel location. Fat and water images were generated by integrating the respective regions of the spectra..


Fig. 2 shows four images from the multi-echo data set, showing a high level of anatomical detail and no discernable artefacts relating to the readout-segmented encoding applied in the head-feet direction. As seen in fig. 3, the spectroscopic analysis of the data made it possible to cleanly separate the signal contributions from water and fat resonances and to generate the corresponding spectrally selective images. Fig. 4 shows single-voxel spectra from four anatomical regions of the knee, each showing a dominant spectral component according to the respective tissue type.


Readout segmentation is used routinely with echo-planar imaging (EPI) to reduce the echo spacing and associated susceptibility artefacts6, but has not been used previously in conjunction with EPSI. As demonstrated by the images and spectra shown in this study, the technique decouples the spatial resolution in the readout direction from the echo spacing, thereby making it possible to simultaneously achieve a high spatial resolution and a high spectral bandwidth. The compromise for this improvement is an increase in the overall scan time due to the acquisition of multiple readout segments. This could be mitigated by using partial Fourier in the readout direction to reduce the number of acquired readout segments7 and by matching the echo spacing carefully to the bandwidth requirements of the application. The results presented in this study could be improved by introducing time-domain filtering before generating the proton spectra; this would be particularly beneficial in the case of the the fat image of fig. 3, whose SNR is affected by the rapid T2 decay of fat during the multi-echo readout.


This preliminary study has demonstrated how the application of readout segmentation to multiple-gradient-echo sequences substantially increases the spectral bandwidth that can be achieved when imaging at high spatial resolution. This approach promises to have a high clinical impact by improving the reliability of data analysis in HiSS studies in general and by facilitating the transfer of the technique to higher field strengths of 3T and above. Furthermore, the propsed sequence will also be a significant advantage for metabolite imaging using EPSI, in particiular for studies at ultra-high field strengths or with nuclei other than 1H, when a higher spectral bandwidth is required.


No acknowledgement found.


1. Du, W., Du, Y. P., Bick, U., Fan, X., MacEneaney, P. M., Zamora, M. A., ... & Karczmar, G. S. (2002). Breast MR Imaging with High Spectral and Spatial Resolutions: Preliminary Experience 1. Radiology, 224(2), 577-585.

2. Medved, M., Newstead, G. M., Abe, H., Zamora, M. A., Olopade, O. I., & Karczmar, G. S. (2006). High spectral and spatial resolution MRI of breast lesions: preliminary clinical experience. American Journal of Roentgenology, 186(1), 30-37.

3. Medved, M., Newstead, G. M., Fan, X., Du, Y. P., Olopade, O. I., Shimauchi, A., ... & Karczmar, G. S. (2009). Fourier component imaging of water resonance in the human breast provides markers for malignancy. Physics in medicine and biology, 54(19), 5767.

4. Mansfield, P. (1984). Spatial mapping of the chemical shift in NMR. Magnetic Resonance in Medicine, 1(3), 370-386. 5.

5. Sarkar, S., Heberlein, K., Metzger, G. J., Zhang, X., & Hu, X. (1999). Applications of high-resolution echoplanar spectroscopic imaging for structural imaging. Journal of Magnetic Resonance Imaging, 10, 1-7. 1.

6. Porter, D. A., & Heidemann, R. M. (2009). High resolution diffusion-weighted imaging using readout-segmented echo-planar imaging, parallel imaging and a two-dimensional navigator-based reacquisition. Magnetic Resonance in Medicine, 62(2), 468-475.

7. Frost, R., Jezzard, P., Douaud, G., Clare, S., Porter, D. A. and Miller, K. L. (2015), Scan time reduction for readout-segmented EPI using simultaneous multislice acceleration: Diffusion-weighted imaging at 3 and 7 Tesla. Magn Reson Med, 74: 136–149.


Fig.1. Diagram for the HiSS sequence with a segmented readout. A sinusoidal readout waveform allows fast gradient switching with lower acoustic noise. The prephasing readout gradient lobe performs a kx offset for the different readout segments acquired at successive excitations. Data are continuously sampled and then regridded as part of image reconstruction. A common phase encoding value is applied to all echoes in the echo train. The number of available echoes is limited by the T2* decay.

Fig.2. Gradient echo images from the knee of a healthy volunteer in sagittal orientation at increasing echo times along the echo train: a) 1st echo time at 4.1 ms, b) 9th echo time at 7.0ms, c) 27th echo time at 13.5ms and d) 81th echo time at 32.9ms.Tissue signal-to-noise and contrast-to-noise ratio change with the prevailing T2* decay.

Fig.3: a) Fat and b) water images obtained from the corresponding resonance peaks in the measured signal spectrum. Voxel intensities for each image type are calculated by integrating the appropriate region of the spectral bandwidth.

Modulus proton spectra corresponding to individual voxels in readout-segmented, multiple-gradient-echo images of the knee. Each spectrum corresponds to an individual voxel of size 0.8 x 0.8 x 2.0 mm3. The spectra illustrate the high spectral bandwidth that can be achieved with the proposed technique, enabling easy identification of resonances without interference from aliasing or instriumental Nyquist artefacts.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)