Enhanced T2 Spectroscopic Imaging Using Backprojections (SI-WEB): A Simulation Study
Zohaib Iqbal1 and M. Albert Thomas1

1Radiological Sciences, University of California - Los Angeles, Los Angeles, CA, United States

Synopsis

Measuring the transverse relaxation times (T2s) of metabolites from different regions of the brain provides important insight into different pathologies, but requires long acquisition. A novel k-space acquisition using a hybrid technique involving both echo planar spectroscopic imaging (EPSI) and concentrically circular echo planar trajectories (SI-CONCEPT) was applied to simulated brain tumor spectroscopic imaging data. In addition, an algorithm incorporating the backprojection information from the EPSI scan was developed to enhance the spectroscopic images. This novel acquisition and reconstruction technique, called SI-WEB, decreases the total duration of the scan while producing similar results when compared to the true T2 spectroscopic images.

Purpose:

It has been shown that T2 values of brain metabolites, including N-acetyl aspartate (NAA) are altered in different pathologies such as brain tumor1,2. These studies have used Point resolved spectroscopy (PRESS) for single voxel localization3. Spectroscopic imaging4 requires an enormous amount of time even with the use of an echo planar bipolar gradient for readout5,6, and this may be the primary reason why T2 spectroscopic imaging is not implemented in vivo. The purpose of this study is to investigate the feasibility of a new type of k-space acquisition, which uses both echo planar spectroscopic imaging (EPSI) and spectroscopic imaging using concentrically circular echo planar trajectories (SI-CONCEPT)7. This type of acquisition has the potential to reduce scan time by two (2x) or three (3x) fold. In order to improve the quality of metabolite maps, backprojections from the acquired EPSI lines are used to reconstruct the images. This method, called spectroscopic imaging with enhancement by utilizing backprojections (SI-WEB), is applied to a virtual brain cancer phantom to assess performance.

Methods

Phantom: The virtual NAA phantom was simulated to replicate a slice with a spatial resolution of 32x32 with the characteristics of a glioma patient. Multi-voxel NAA spectra were simulated for B0 = 3T containing 512 temporal points (t) and a spectral bandwidth of 1190Hz. NAA concentration was chosen to be uniform throughout the brain phantom and decreased in the location of the tumor by a factor of two, and was simulated for three TE values (TE= 30, 90, 180ms). The T2 of the healthy tissue was set to 350ms and the T2 of the region mimicking the cancer tissue was set to 250ms (usually a 2x2 or 3x3 region of the tumor). These T2 values are similar to those reported in the literature2. Ten phantoms with different tumor locations were simulated and tested.

SI-WEB acquisition: The 2x SI-WEB sampling scheme can be seen in Figure 1. It is clear that k-space is only partially acquired using ten EPSI phase encoding lines (in red) and six SI-CONCEPT phase encoding lines (in blue). For the 3x SI-WEB mask (Figure 2), 6 EPSI and 4 SI-CONCEPT phase encoding lines are acquired.

Reconstruction using Backprojections: Two backprojections were acquired using EPSI: kx = 17 and ky = 17 lines. These two backprojections (0º and 90º) were used to ensure the data was consistent using the following optimization: $$\min_{X}\|MX-K\|_2^2+\lambda\sum_{b=1}^2\|X_{b}-K_{b}\|_2^2$$ M, X, K, λ, Xb, and Kb are the sampling mask, reconstructed data, acquired data, regularization term, projection of the reconstructed data along either 0º or 90º, and the acquired backprojection along either 0º or 90º respectfully. Essentially, the first term acted as a data fidelity term and the second term ensured consistency between the backprojections and reconstructed data. T2 maps were created by fitting the three NAA maps to $$$e^{-{TE}/T_2}$$$.

Quality Evaluation: Normalized root mean square error (nRMSE) images were calculated between the T2 maps of the virtual phantom (truth) and the 2x and 3x SI-WEB NAA maps. Qualitative comparisons were also made between the virtual phantoms and reconstructed SI-WEB maps at each simulated TE.

Results

A comparison between the virtual phantom and the reconstructed 2x SI-WEB at each TE can be seen in Figure 3A. Figure 3B shows the resulting NAA T2 maps by fitting the three TE points for the virtual phantom and 2x SI-WEB acquisition. Figure 4 shows the T2 maps of the virtual phantom (A), 2x SI-WEB (B), 3x SI-WEB (C), as well as the nRMSE for the 2x SI-WEB (D) and 3x SI-WEB (E). The largest error in the region of the glioma for the 2x SI-WEB was 16% and was 26% for the 3x SI-WEB for this particular phantom, whereas the smallest error in the region of the lesion was 5% for the 2x SI-WEB and 23% for the 3x SI-WEB. Other phantoms yielded very similar results.

Discussion and Conclusion

For ten different versions of the virtual phantom, twice and thrice accelerated (2x and 3x) SI-WEB showed similar NAA T2 maps when compared to the true T2 maps. The reconstruction employed here helped minimize noise where signal was not present, and this type of reconstruction would yield better results if more backprojections were used. Future work will focus on applying this method in human brain to validate this technique in vivo.

Acknowledgements

NIH R21 Grant (NS080649-02)

References

1. Manton DJ, Lowry M, Blackband SJ, et al. Determination of Proton Metabolite Concentrations and Relaxation Parameters in Normal Human Brain and Intracranial Tumors. NMR Biomed 1995;8:104-112.

2. Isobe T, Matsumura A, Anno I, et al. Quantification of cerebral metabolites in glioma patients with proton MR spectroscopy using T2 relaxation time correction. Magn Reson Imag 2002;20:343-9.

3. Bottomley PA. Spatial localization in NMR spectroscopy in vivo. Ann NY Acad Sci 1987;508:333-348.

4. Brown TR, Kincaid BM, Ugurbil K. NMR chemical shift imaging in three dimensions. Prot Natl Acad Sci 1982;79:3523-6.

5. Mansfield P. Spatial mapping of the chemical shift in NMR. J Phys D Appl Phys 1983;L235-8.

6. Posse S, Tedeschi G, Risinger R, et al. High speed 1H spectroscopic imaging in human brain by echo planar spatial-spectral encoding. Magn Reson Med 1995;33:34-40.

7. Furuyama JK, Wilson NE, Thomas MA, et al. Spectroscopic imaging using concentrically circular echo-planar trajectories in vivo. Magn Reson Med 2012;67:1515-1522.

Figures

Figure 1. The 2x SI-WEB mask is shown with EPSI acquisition in red and SI-CONCEPT acquisition in blue for a 32x32 slice. A total of 10 EPSI lines and 6 SI-CONCEPT lines are used for acquisition.

Figure 2. The 3x SI-WEB mask is shown with EPSI acquisition in red and SI-CONCEPT acquisition in blue for a 32x32 slice. A total of 6 EPSI lines and 4 SI-CONCEPT lines are used for acquisition.

Figure 3. Virtual phantom and 2x SI-WEB NAA maps at each TE value are shown (A). T2 maps are also compared between the two (B), where the T2 feature is noted with a black arrow. Even though reconstructed NAA maps appear noisy, overall T2 map quality is high.

Figure 4. T2 NAA metabolite maps for the virtual phantom (A), 2x SI-WEB (B), and 3x SI-WEB (C) are shown. In addition, the nRMSE maps are shown for the 2x SI-WEB (D) and 3x SI-WEB (E). In general, the decreased T2 feature is noticeable for both 2x and 3x SI-WEB.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3953