Center for Cognitive and Neurobiological Imaging

Recent advances in in-vivo spectroscopy methods at CNI

Background

Interest in measuring metabolic changes via MRS techniques and combining that information with functional MRI measurements in neuroscience studies continues to grow. CNI continues to support the research of its user community by developing and incorporating for general use new data acquisition and data analysis capabilities as recommended by the at-large MRS community. Through collaborative efforts, the special interest spectroscopy group at CNI has enabled education and participated in experimental design, guided analyses, and interpretation of results. If you are interested in learning more about these methods, or using them, please consult with Laima Baltusis.

Examples of studies at the CNI using in-vivo spectroscopy techniques include characterization of biomarkers following transcranial magnetic stimulation, and metabolite characterization for conditions such as addiction, pain, depression, and various forms of dementia. Collaborative studies performed at CNI include:

• Method applications (DeSouza DD, Stimpson K, Baltusis L, Sacchet MD, Gu M, Hurd R, Wu H, Yeomans DC, Williams N, Spiegel D. Association between anterior cingulate neurochemical concentration and individual differences in hypnotizability. Cerebral Cortex, Volume 30, Issue 6, June 2020, Pages 3644–3654),
• New method developments (Gu M, Hurd R, Noeske R, Baltusis L, Hancock R, Sacchet MD, Gotlib IH, Chin FT, Spielman DM (2018) GABA editing with macromolecule suppression using an improved MEGA-SPECIAL sequence, Magnetic Resonance in Medicine 79:41-47),
• Participation in multi-site spectroscopy studies (Hui et al. Frequency drift in MR spectroscopy at 3T, NeuroImage 241(2021) 118430),
• Voxel placement method development (James H. Bishop, Andrew Geoly, Naushaba Khan, Claudia Tischler, Ruben Krueger, Poorvi Keshava, Heer Amin, Laima Baltusis, Hua Wu, David Spiegel, Nolan Williams, Matthew D. Sacchet Real-Time Semi-Automated and Automated Voxel Placement using fMRI Targets for Repeated Acquisition Magnetic Resonance Spectroscopy, Journal of Neuroscience Methods 392 (2023) 109853).

The most recent spectroscopy results were shared in a poster ENC-2024-Poster-final presented at the 65^th ENC, the premier conference for nuclear magnetic research (April 7 – 11, 2024). The key technical points in the poster are summarized below. Additional details and supporting figures are in the poster attached.

Technical Notes

For all spectroscopy research projects CNI actively follows the recommended best practices from the ISMRM (International Society for Magnetic Resonance in Medicine)  Spectroscopy Study Group and recently published experts’ consensus recommendations. Currently the semi-LASER sequence is the ISMRM consensus method, particularly for multi-site 3T MRS studies. We have transitioned studies at CNI from previously recommended sequences (PRESS, as example) to the semi-LASER sequence (a GE WIP (Works-in-Progress)) for both short TE single voxel and MRSI studies.

In this year’s poster we presented both single voxel (MRS) and focal multi-voxel (MRSI) methods. Both single and focal multi-voxel scans are acquired at short TE (30 msec) using  semi-LASER volume selection.  Focal MRSI has proven to be a better choice in challenging and therefore less well studied but important areas of the human brain such as the subcortical basal ganglia regions. These regions are often studied by investigators trying to understand movement, reward, cognition, and emotion.

To measure and improve data quality of small metabolite differences in brain subregions of interest, we have been optimizing and evaluating focal 2D MRSI data acquisition in both cortical and subcortical regions of the brain as an alternative to single voxel MRS data acquisition.

Our results indicate fewer artifacts, fewer baseline issues, and better signal-to-noise in focal 2D MRSI data relative to single voxel MRS data as input for data processing. Additionally, the use of open-source software (FreeSurfer) for automated segmentation of the subvoxel space allows for the separation of gray matter and white matter within the prescription to identify and correct for potential influences on neurometabolite estimates from different tissue amounts in each subvoxel using percentage of gray matter as a covariant.  Single-voxel MRS in the right dorsolateral prefrontal cortex (DLPFC) was statistically compared with 2D focal MRSI (right DLPFC subvoxels extracted) in 7 controls at two timepoints. Analysis of the results showed that there was less variability with 2D focal MRSI than single voxel MRS across timepoints (test-retest) for each data analysis used (NAA concentration, gray matter determination, white matter determination).

Initial basic data reconstruction and visualization employed SAGE, a GE proprietary analysis and visualization tool with LCModel used as a separate module for data fitting. We have now automated the data processing pipeline in Matlab.

For optimal quantification of metabolites in both focal 2D MRSI and single voxel MRS data analysis methods have included: evaluation of improved coil combination methods; improvements in data analysis using LCModel by (1) mitigation of baseline and macromolecular contributions for data analysis with LCModel and (2) improvement of the accuracy of the LCModel basis using a largely experimental 23 metabolite basis set; segmentation to identify and correct for potential influences on neurometabolite estimates from different tissue amounts in each subvoxel, using percentage of gray matter as a covariate.

Future research directions include correlating focal MRSI in subcortical regions with methods such as QSM (iron susceptibility) and resting state fMRI (network connectivity) to determine if there is a relationship with metabolites on iron and/or brain networks.

Recent advances in In-vivo spectroscopy methods at the CNI

Background

A growing number of people at the CNI are interested in measuring metabolic changes in the brain using in-vivo spectroscopy (MRS) techniques.  Often, they combine this information with functional MRI measurements. The special interest spectroscopy group at CNI  develops and supports new MRS data acquisition and data analysis capabilities, following the recommendations of the at-large MRS community. This message describes the initiatives for education, support of experimental design, guided analyses, and interpretation of MRS data.

Recent spectroscopy results were shared in a poster that we presented at the 64^th ENC, the premier conference for nuclear magnetic research (April 16 – 20, 2023). Examples of studies at the CNI using in-vivo spectroscopy techniques include characterization of biomarkers following transcranial magnetic stimulation, and metabolite characterization for conditions such as addiction, pain, depression, and various forms of dementia. Published, collaborative studies performed at CNI include

• method applications (DeSouza DD, Stimpson K, Baltusis L, Sacchet MD, Gu M, Hurd R, Wu H, Yeomans DC, Williams N, Spiegel D. Association between anterior cingulate neurochemical concentration and individual differences in hypnotizability. Cerebral Cortex, Volume 30, Issue 6, June 2020, Pages 3644–3654),
• new method developments (Gu M, Hurd R, Noeske R, Baltusis L, Hancock R, Sacchet MD, Gotlib IH, Chin FT, Spielman DM (2018) GABA editing with macromolecule suppression using an improved MEGA-SPECIAL sequence, Magnetic Resonance in Medicine 79:41-47),
• participation in multi-site spectroscopy studies (Hui et al. Frequency drift in MR spectroscopy at 3T, NeuroImage 241(2021) 118430), and
• a recent paper regarding voxel placement method development (James H. Bishop, Andrew Geoly, Naushaba Khan, Claudia Tischler, Ruben Krueger, Poorvi Keshava, Heer Amin, Laima Baltusis, Hua Wu, David Spiegel, Nolan Williams, Matthew D. Sacchet Real-Time Semi-Automated and Automated Voxel Placement using fMRI Targets for Repeated Acquisition Magnetic Resonance Spectroscopy, Journal of Neuroscience Methods 392 (2023) 109853.

Technical notes

Here is a description of the specific methods we now support.  If you are interested in learning more about these methods, or using them, please consult with Laima Baltusis.

1. Experimental spectroscopy data acquisition methods continue to evolve. Currently the semi-LASER sequence is the ISMRM (International Society for Magnetic Resonance in Medicine) consensus method, particularly for multi-site 3T MRS studies. We have, therefore, transitioned studies at CNI to the semi-LASER sequence (a GE WIP (Works-in-Progress)) for both single voxel and to MRSI studies. The ENC poster presents both single voxel and new focal 2D MRSI results using the semi-LASER sequence in challenging and therefore less well studied but important areas of the human brain such as the basal ganglia regions. This region is often studied by investigators trying to understand movement, cognition, and emotion, and the single voxel data in this region has been of lower quality. To improve the data quality, and measure small metabolite differences in subregions of interest, we optimized a focal 2D MRSI data acquisition method (presented in the poster). The improvements for focal 2D MRSI and single-voxel MRS includes B0 shimming, encoding and acceleration methods, spatial selection, water suppression, extracranial lipid suppression, and RF (Radio Frequency) performance.
2. In addition, current basic data reconstruction and visualization can employ SAGE, a GE proprietary analysis and visualization tool with LCModel used as a separate module for data fitting. The next phase of this particular research project will continue to explore additional optimizations of data collection but will focus more on data processing for optimal quantification of metabolites. Areas of study will include improvements in data analysis using LCModel by (1) mitigation of baseline and macromolecular contributions for data analysis with LCModel and (2) improvement of the accuracy of the LCModel basis set by the addition and validation of experimental lower concentration metabolites (glutathione and ascorbate, as examples) to an otherwise synthetic LCModel basis set. The use of Bayesian analysis will also be explored as a technique to remove the broad baseline and macromolecule content without bias. We anticipate presenting these findings at the 2024 ENC conference.

You may have been wondering about Laima’s experimental work over the last few years.  You can now read about it in a paper co-authored by folks who work at the CNI, GE, and Lucas.  Nice collaboration!

Acquiring a high quality fieldmap of the mean magnetic field is crucial to obtaining a good measurement.  Making sure you set the parameters so that this field map is obtained as part of your protocol is something we help you with.  There are cases, including new sequences, where this can require some extra attention.

The problem: The CNI team recently helped diagnose a confusing fieldmap result.  The user was acquiring a fieldmap using a spiral acquisition, but without first performing a high-order shim (HOS) acquisition or setting Autoshim (linear-order shim).  They subsequently acquired EPI data with Autoshim.  This resulted in acceptable imaging quality but also created some problems. The disadvantage was that high-order shim corrections were not used, and the spiral fieldmap data did not incorporate the linear-order shim corrections that were imposed by the Autoshim.  As a result, the data quality were not optimized.

The solution: We recommend that a HOS be acquired as part of your protocol, especially when EPI sequences are used for subsequent data acquisition.  We describe these procedures on our Wiki entries, as well as some troubleshooting steps in a blog entry:

http://cni.stanford.edu/wiki/MR_Protocols#HOS_WB_HRBRAIN
http://cni.stanford.edu/wiki/Troubleshooting#HIGH_ORDER_SHIMS_.28HOS.29
https://cni.stanford.edu/high-order-shim-errors/

If this overhead is too time-consuming, then it is important to ensure an Autoshim is acquired, which performs a linear-order shim correction. Please note: when Autoshim acquisition is set to “Auto”, and an HOS is acquired, the system should use the HOS acquisition rather than acquire a new linear correction.

Some details: A variation in the B0 main field from the desired B0 leads to an image shift in the phase encode direction of EPI acquired images.  When the B0 field is spatially varying, this leads to varying image shifts which can result in signal pileup, voids, and image warping. There are two methods described in our Wiki for correcting these distortions.  First, you can repeat your EPI data acquisition twice, with a reversal of the phase-encoding direction between data sets as we describe in a Wiki entry: CNI Data Processing

Secondly, you can acquire an explicit fieldmap using the spiral fieldmap sequence and then process the subsequently acquired EPI data as described in another Wiki entry: Fieldmaps

Please note that if the explicit fieldmap correction approach is adopted, it is important to ensure that the subsequent EPI acquisition does not modify the shim corrections.  In this case setting Autoshim to “Off” in your protocol would be advisable.

 
Adam and Hua