Supplementary Material
Cortico-Striatal GABAergic and Glutamatergic Dysregulations in Subjects at Ultra-High
Risk for Psychosis Investigated with Proton Magnetic Resonance Spectroscopy
Camilo de la Fuente-Sandoval, Francisco Reyes-Madrigal, Xiangling Mao, Pablo León-Ortiz,
Oscar Rodríguez-Mayoral, Rodolfo Solís-Vivanco, Rafael Favila, Ariel Graff-Guerrero, Dikoma
C. Shungu
MRS Data Quality Assessment Criteria and Procedures
In this study, spectral quality assessment, which determined whether the results of
individual MRS data sets were ultimately included in and rejected from group analyses,
was established using a number of criteria along the spectral data acquisition and
processing pipeline. As can be appreciated by examining the sample spectrum
presented in Figure 1B, which is typical of all successful acquisitions, we achieved
relatively high spectral quality and SNRs. Therefore, no spectra were rejected strictly
due to poor quality or SNR. Rather, spectra were rejected because either (1) the shim
quality was poor (defined as a full-width at one-half maximum of the water resonance of
more than 20 Hz, and/or spectra with unresolved total creatine and total choline
resonances at 3.03 and 3.22 ppm), or (2) there was excessive head motion during a
scan. Our criteria for detecting and rejecting motion-degraded spectra were: (1) a very
large residual water resonance in the difference or edited spectra due to poor
cancelation upon subtracting 2 subspectra in which the water signal was differentially
affected by the head motion; (2) peak phase distortions in all the spectra that could not
be automatically adjusted using the phases derived from the unsuppressed voxel water
resonance; and (3) degraded SNR in the edited spectra due to incoherent summing of
subspectra with motion-induced peak phase and position shifts.
All the spectra that met those initial quality assessment criteria were processed
as illustrated in Figure 1B (traces a-e) to obtain the area under the GABA and Glx
peaks, which are proportional to the concentration of each neurotransmitter in the voxel
of interest. Briefly, the GABA and Glx resonances in the J-edited difference spectra
were modeled as a linear combination pseudo-Voigt lineshape functions and then fitted
in the frequency domain using a robust and highly optimized public-domain Levenberg–
Marquardt nonlinear least-squares minimization routine, MPFIT (Markwardt, 2009) (The
IDL fitting routine, 'MPFIT', is available at http://purl.com/net/mpfit; Last Modified on
2013-08-14 10:55:25 by Craig Markwardt). The pseudo-Voigt lineshape function
enables more precise analysis of line shapes that consist of mixtures of Lorentzian and
Gaussian functions (Marshall et al., 2000), as is often the case for in vivo spectra.
At convergence, MPFIT, like all nonlinear least-squares estimation procedures,
reports the minimized sum of squared residuals, or 2, as the “goodness of fit”
(Bevington and Robinson, 2003). A fit is considered acceptable if at convergence, 2 is
approximately equal to the number of degrees of freedom, defined as the difference
between the number of spectral data points (N) and the number of estimated
parameters (M), or if 2 /(N-M), known as reduced 2, is approximately equal to 1.
Higher 2 values than these limits generally indicate an incorrect fitting model or a failed
fit, while lower values would indicate overestimation of measurement errors. Due to the
relative simplicity and high SNR and quality of the spectra in this study, excellent
“goodness of fit,” based on both 2 values and visual inspection of the fits, was
consistently obtained, so that no spectra that survived our initial quality assessment
criteria were rejected due to poor fit.
Test-Retest Reliability and Mobile Macromolecule Contamination
We previously evaluated the test-rested reliability of brain GABA and Glx measurement
using the J-editing and found it to be high (Kegeles et al., 2006). We also assessed the
contribution of macromolecules to the total detected GABA and found it to be relatively
constant across brain regions, ranging from 41% to 49%, a nonsignificant regional
variation (P = .58) (Kegeles et al., 2007).
REFERENCES
Bevington PR, Robinson DK (2003) Data reduction and error analysis for the physical
sciences, 3rd Edition. Boston: McGraw-Hill.
Kegeles LS, Mao X, Gonzales R, Shungu DC (2007) Evaluation of anatomic variation in
macromolecule contribution to the GABA signal using metabolite nulling and the
j-editing technique at 3.0 T. Proc Intl Soc Mag Reson Med 15:1391.
Kegeles LS, Mao X, Dyke J, Gonzales R, Soones T, Shungu DC (2006) Test-retest
reliability of dorsolateral prefrontal cortical GABA measurement using an 8channel phased-array head coil with the J-editing technique at 3T. Proc Intl Soc
Mag Reson Med 14:489.
Markwardt CB. Non-linear Least-squares fitting in IDL with MPFIT. Paper presented at:
Proceedings of Astronomical Data Analysis Software and Systems XVIII 2008,
2009; San Francisco, CA.
Marshall I, Bruce SD, Higinbotham J, MacLullich A, Wardlaw JM, Ferguson KJ, Seckl J
(2000) Choice of spectroscopic lineshape model affects metabolite peak areas
and area ratios. Magn Reson Med 44:646–649.
Figure S1
Scatterplots with linear regression fit for the associations between GABA and Glx: (A) in
the MPFC and (B) in dorsal caudate across all participants. The open circles (○)
represent the subjects at UHR and the open squares (□) the control subjects. The
GABA vs Glx associations across the full sample of participants were significant for both
voxels of interest. The curves bracketing the data points represent the 95% confidence
interval.
Figure S2
Scatterplots with linear regression fit for the associations between GABA and Glx: (A) in
the MPFC of the control participants and (B) in dorsal caudate of the subjects at UHR.
None of the associations was found to be statistically significant. The curves bracketing
the data points represent the 95% confidence interval.
Figure S3
(A) Scatterplot with linear regression fit for the association between GABA in the
caudate and Glx in the MPFC within the control group.(B) Scatterplot with linear
regression fit for the association between Glx in the MPFC and the SIPS positive
subscores within the UHR group. None of the associations was statistically significant.
The curves bracketing the data points represent the 95% confidence interval.