Comparison of STEAM and Semi-Laser for neurochemical separation at 7T
Poster No:
2391
Submission Type:
Abstract Submission
Authors:
Tiffany Bell1, Dana Goerzen2, Jamie Near3, Ashley Harris1
Institutions:
1University of Calgary, Calgary, Alberta, 2Cornell University, Ithaca , NY, 3University of Toronto, Toronto , Ontario
First Author:
Co-Author(s):
Introduction:
Proton Magnetic Resonance Spectroscopy (MRS) is commonly used to quantify neurochemicals of interest. Typically quantified neurochemicals are: total N-acetyl aspartate (tNAA), total Creatine (tCr), total Choline (tCho), and Glx. However, these are composites of signals from multiple overlapping neurochemicals. For example, Glx signal consists of overlapping signals glutamate (Glu) and Glutamine (Gln).
Higher field strengths (i.e. 7T) increase signal-to-noise ratio (SNR) and spectral resolution, allowing better signal separation. Several sequences have been proposed for 7T MRS. STEAM allows for very short echo times (TEs), minimizing signal modulation due to J-coupling and T2 relaxation, whereas Semi-LASER (sLASER) has reduced chemical shift displacement and better resilience to B1 inhomogeneity (Wilson et al., 2019).
Previous work suggested sLASER (TE=28ms) produced a more robust measure of Glu than STEAM (TE=7.5ms) at 7T (Marsman et al., 2017). When comparing TEs 45-225ms with sLASER localization, the optimal echo time of Glu signal evolution was 105ms (Wong et al., 2018). However, these studies only focused on Glu.
This 7T study directly compares the separation of neurochemicals using STEAM, short TE sLASER, and long TE sLASER acquisitions.
Higher field strengths (i.e. 7T) increase signal-to-noise ratio (SNR) and spectral resolution, allowing better signal separation. Several sequences have been proposed for 7T MRS. STEAM allows for very short echo times (TEs), minimizing signal modulation due to J-coupling and T2 relaxation, whereas Semi-LASER (sLASER) has reduced chemical shift displacement and better resilience to B1 inhomogeneity (Wilson et al., 2019).
Previous work suggested sLASER (TE=28ms) produced a more robust measure of Glu than STEAM (TE=7.5ms) at 7T (Marsman et al., 2017). When comparing TEs 45-225ms with sLASER localization, the optimal echo time of Glu signal evolution was 105ms (Wong et al., 2018). However, these studies only focused on Glu.
This 7T study directly compares the separation of neurochemicals using STEAM, short TE sLASER, and long TE sLASER acquisitions.
Methods:
Simulations were conducted to assess the effects of measurement error on between- and within-subject variation to determine the best metric for comparisons.
14 healthy participants (18-40y) were scanned at 7T (Siemens Terra). Voxels were placed in the parietal cortex centred on the midline. Data were collected using 3 sequences: STEAM_8 (TE/TR/TM=8/6000/40ms), sLASER_34 (TE/TR=34/5000ms), and sLASER_105 (TE/TR=105/5000ms), all with 64 averages. Sequence order was counterbalanced across subjects. Data were preprocessed using FID-A (Simpson et al., 2017) and quantified with LCModel (Provencher, 2001). To assess repeatability, the scans were split into two (each with 32 averages), then processed as above.
The analysis focused on the following combinations of metabolites: tNAA: NAA+N-acetylaspartylglutamate (NAAG), tCr: Cr+phosphocreatine (PCr), tCho: Cho+phosphocholine (PC)+glycerophosphocholine (GPC), Glx: Glutamate+Glutamine, and glutathione (GSH). The fit of the model was assessed using CRLB's, and repeatability was assessed using coefficient of variation (CV).
14 healthy participants (18-40y) were scanned at 7T (Siemens Terra). Voxels were placed in the parietal cortex centred on the midline. Data were collected using 3 sequences: STEAM_8 (TE/TR/TM=8/6000/40ms), sLASER_34 (TE/TR=34/5000ms), and sLASER_105 (TE/TR=105/5000ms), all with 64 averages. Sequence order was counterbalanced across subjects. Data were preprocessed using FID-A (Simpson et al., 2017) and quantified with LCModel (Provencher, 2001). To assess repeatability, the scans were split into two (each with 32 averages), then processed as above.
The analysis focused on the following combinations of metabolites: tNAA: NAA+N-acetylaspartylglutamate (NAAG), tCr: Cr+phosphocreatine (PCr), tCho: Cho+phosphocholine (PC)+glycerophosphocholine (GPC), Glx: Glutamate+Glutamine, and glutathione (GSH). The fit of the model was assessed using CRLB's, and repeatability was assessed using coefficient of variation (CV).
Results:
Simulations showed that, for between subject variation, there was a large overlap between distributions of different measurement errors. In contrast, for within subject variation, there was little overlap between distributions from small measurement errors and large measurement errors. This indicates within subject variation is a better indicator of measurement error, and subsequently we chose this metric for comparing sequences (Figure 1).
For most neurochemicals, sLASER_34 had the lowest CRLB and STEAM_8 had the lowest CV. However, for Cho and GPC, STEAM_8 had the lowest CRLB. For PC sLASER_105 had the lowest CRLBs, however these were still much higher than the typical cut-off points.
For most neurochemicals, sLASER_34 had the lowest CRLB and STEAM_8 had the lowest CV. However, for Cho and GPC, STEAM_8 had the lowest CRLB. For PC sLASER_105 had the lowest CRLBs, however these were still much higher than the typical cut-off points.
Conclusions:
Overall, sLASER_34 best quantified both the individual and summed signals. However, Cho and GPC were best quantified with STEAM_8. PC wasn't quantified well with any sequence, likely due to the strong overlap between PC and GPC.
STEAM_8 generally produced lower CV's, in contrast to previous work (Marsman et al., 2017), however this study compared measurements acquired two weeks apart. STEAM may produce more consistent results within session, but less consistent results across time. sLASER is likely less sensitive to repositioning effects due to reduced chemical shift displacement and better resilience to B1 inhomogeneity (Wilson et al., 2019).
We conclude short TE sLASER is better for quantification of most standard neurochemicals than long TE sLASER or short TE STEAM at 7T. Depending however, if there is a specific neurochemical of interest (i.e., choline containing neurochemicals), short-echo STEAM or longer echo sLASER may be preferred.
STEAM_8 generally produced lower CV's, in contrast to previous work (Marsman et al., 2017), however this study compared measurements acquired two weeks apart. STEAM may produce more consistent results within session, but less consistent results across time. sLASER is likely less sensitive to repositioning effects due to reduced chemical shift displacement and better resilience to B1 inhomogeneity (Wilson et al., 2019).
We conclude short TE sLASER is better for quantification of most standard neurochemicals than long TE sLASER or short TE STEAM at 7T. Depending however, if there is a specific neurochemical of interest (i.e., choline containing neurochemicals), short-echo STEAM or longer echo sLASER may be preferred.
Modeling and Analysis Methods:
Methods Development 2
Novel Imaging Acquisition Methods:
MR Spectroscopy 1
Keywords:
Design and Analysis
Glutamate
Magnetic Resonance Spectroscopy (MRS)
MR SPECTROSCOPY
1|2Indicates the priority used for review
Provide references using author date format
Provencher (2001). Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR Biomed, 14, 260–264.
Simpson et al., (2017). Advanced processing and simulation of MRS data using the FID appliance (FID-A)—An open source, MATLAB-based toolkit. Magnetic Resonance in Medicine, 77(1), 23–33.
Wilson et al., (2019). Methodological consensus on clinical proton MRS of the brain: Review and recommendations. Magnetic Resonance in Medicine, 82(2), 527–550.
Wong et al (2018). Optimized in vivo brain glutamate measurement using long-echo-time semi-LASER at 7 T. NMR in Biomedicine, 31(11), 1–13.