INTRODUCTION

Cerebral glucose has been measured using ¹H MRS. The glucose peaks in the range of 3-4 ppm are analyzed using spectral analysis software like LCModel. To fit such weak peaks overlapped with other high-concentrated metabolite peaks, the prior knowledge about an intensity ratio to others (e.g., three big methyl peaks) is usually required. Even if the glucose peaks are fitted, it is difficult for glucose and taurine to quantitate separately due to the spectral pattern similarity.¹

H1-α-glucose peak at 5.23 ppm has also been measured at 4T^2-3 and 7T^4-5 which overcome the low sensitivity. Because the H1-α-glucose peak is free of spectral overlap, the detection of the H1-α-glucose has a potential of the accurate quantification of cerebral glucose concentration. However, if the water signal is poorly suppressed, the H1-α-glucose peak and surrounding baselines would be overlapped with the residual water peak. Hence, a proper baseline correction is important to fit the low-SNR H1-α-glucose peak. In this study, the H1-α-glucose peak was analyzed using LCModel with data excluding a range of the residual water peak, and the effects into the baseline around the H1-α-glucose peak were observed.

METHODS

This study was approved by an institutional review board, and written informed consent was obtained from subjects. The 20 healthy subjects (aged 20-24 years) were scanned on an investigational whole-body 7T scanner (MAGNETOM 7T, Siemens Healthcare, Erlangen, Germany). A single-channel transmit, and 32-channel receive head coil (Nova Medical, MA, USA) was used. Siemens investigational prototype sequence of MPRAGE for T₁ anatomical imaging was utilized to position an MRS voxel at the posterior cingulate cortex (27 mL) in the mid-sagittal plane on the images. B₀ shimming was achieved using FASTESTMAP⁶ (Siemens investigational prototype sequence). B₁ amplitude was adjusted for the MRS voxels.

MR spectra were acquired using Siemens investigational prototype sequences of semi-LASER (TR/TE = 6500-8000/32 ms, signal average = 96, scan time = 11-13 min) and short-TE STEAM (TR/TE/TM = 5700-7280/5/45 ms, signal average = 128, scan time = 12-15.5 min) with VAPOR water suppression⁷ and outer volume suppression. Identical MRS parameters were RF transmitter offset = 4.2 ppm, spectral width = 6 kHz, data points = 2048. The TRs were determined as the shortest values under SAR limitation. Water spectra in the MRS voxels were acquired too.

Spectra were analyzed using LCModel (version 6.3-1L). A semi-LASER basis set was developed in-house. Basis sets only for the doublet of the H1-α-glucose were generated to quantitate glucose concentration via the peak. Eddy current correction⁸ and water-scaling for quantification⁹ were performed using the water spectra. The analysis window was 0.2-6.0 ppm with the gaps of 4.20-5.00 ppm and 4.47-5.00 ppm in the semi-LASER and the short-TE STEAM, respectively, to exclude the residual water peak from the analysis using PPMGAP parameter. Baseline correction for glucose quantification via the H1-α-glucose peaks was evaluated using residuals (the acquired spectrum minus the fit) in 5.03-5.43 ppm. The precision of quantification was evaluated using Cramer-Rao lower bounds (CRLB, %SD). The SNR was calculated with the fit peak height divided by the SD of the residuals (i.e., baseline-corrected noises) in 0.2-0.6 ppm. The spectra with the SNR of the H1-α-glucose higher than 3, which is an acceptable value for estimating a detection limit in analytical chemistry guideline,¹⁰ were evaluated for glucose quantification. Statistical analysis was conducted, and a P < 0.05 was considered to be statistically significant.

RESULTS

Figure 1 shows the spectra acquired from one representative subject and analyzed using PPMGAP. The insets of the figure 1 shows that the H1-α-glucose peaks are fitted using the generated basis-sets. Figure 2 shows the spectra in the range of 4.0-6.0 ppm analyzed with and without PPMGAP. The insets of the figure 2 shows averages and SDs of the residuals around the H1-α-glucose peak analyzed with (blue) and without (black) PPMGAP. Evaluating with the averaged residuals, the errors of baseline fitting in the right-hand side near the water peak are reduced with PPMGAP to the level of those of fitting the H1-α-glucose peak. Evaluating with the SDs of the residuals, the errors of baseline fitting among the subjects are reduced in the semi-LASER with PPMGAP.

Results from the spectral analysis are shown in Table 1. The mean and detection limit of glucose concentrations were estimated to be 1.0 mM and 0.6 mM, respectively, via both sequences. The CRLBs of fitting the H1-α-glucose peak were significantly smaller in the semi-LASER (33.1 %) than those in the short-TE STEAM (49.8 %). The water peaks were significantly better suppressed in the short-TE STEAM than those in the semi-LASER.

DISCUSSION

Inadequate water suppression distorted the baseline around the H1-α-glucose peak. The distortions were severe in the semi-LASER due to poor water suppression. The averaged residuals in the semi-LASER had a peak at 5.08 ppm, which would be formed with the sum of multiple components including the tail of the residual water peak. Excluding data of the residual water peak with PPMGAP corrected the baseline distortion and reduced the variation among the subjects. The post-processing approach will simplify the scanning operation without the optimization of water suppression and improve the quantification of cerebral glucose concentration via the H1-α-glucose peak detection.

Acknowledgements

The authors thank Dr. Lana G Kaiser (University of California, Berkeley) to provide the semi-LASER basis set, Dr. Gerald R Moran (Siemens Healthcare Canada) to provide the prototype FASTESTMAP sequence, Dr. Tobias Kober (Siemens Healthcare GmbH) to provide the prototype MPRAGE sequence and Mr. Koji Yasuda (LA Systems) to support LCModel.