Uzay E Emir1,2, Tianmeng Lyu3, Dinesh K Deelchand2, James M Joers2, Diane Hutter2, Christopher M Gomez4, Khalaf O Bushara5, Lynn E Eberly3, and Gulin Oz2
1FMRIB Centre, University of Oxford, Oxford, United Kingdom, 2Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, United States, 3Division of Biostatistics, School of Public Health, University of Minnesota, Minneapolis, MN, United States, 4Department of Neurology, University of Chicago, Chicago, IL, United States, 5Department of Neurology, Medical School, University of Minnesota, Minneapolis, MN, United States
Synopsis
To evaluate diagnostic accuracy of
state-of-the-art MRS in early neurodegenerative disease, we measured
neurochemical profiles in the vermis, cerebellar hemisphere and brainstem of genetically
confirmed subjects with spinocerebellar ataxia type 1 and controls by 3T and 7T
1H MRS. Concentrations of major metabolites obtained at 3T and 7T were strongly
correlated. While 3T showed great potential by enabling detection of abnormal
metabolite levels even in the presymptomatic stage, the increased sensitivity
at 7T enabled group separation with higher significance and identification of
subtle neurochemical alterations in early symptomatic disease stage more
robustly than at 3T. Introduction
In
the post-genomics era movement disorders research is shifting toward more
specialized clinical trials that subdivide patients into groups based on
molecular characteristics. In other words, personalized medicine promises
specialized therapies for each individual by delivering more effective drug
treatments, while avoiding or reducing adverse drug reactions. Hereditary
spinocerebellar ataxia type 1 (SCA1), a polyglutamine movement disorder, was
the first SCA for which the genetic defect was uncovered (1). As potential treatments for SCA1 enter the
pipeline, robust, noninvasive biomarkers of cerebral pathology are urgently
needed for early diagnosis, monitoring of treatment and post-treatment
evaluation. A prior MRS study demonstrated neurochemical alterations in SCA1
using a 4T research scanner (2). With wide availability of 3T and increasing
availability of 7T scanners for clinical research and trials, a need arises to
establish the added value of the ultra-high field (7T) relative to high field
(3T). Therefore, here we compared the sensitivity of single-voxel, short-echo
MRS using a FASTMAP+semi-LASER protocol at 3T vs. 7T to distinguish genetically
confirmed subjects with SCA1 at early-moderate disease stage from age and
gender frequency-matched controls. Specifically, we evaluated the ability of
the protocol to make a diagnostic decision on a single subject basis.
Methods
18 individuals with early-moderate SCA1
(age 50.9 ± 10.2 years, mean ± SD, 8M/10F, Scale for the Assessment and Rating
of Ataxia (SARA) score, 9.0 ± 3.5) and 29 age range matched healthy volunteers (53.3
± 15.1 years, 13M / 16F, SARA, 0.1 ± 0.2) were studied. Two patients were
presymptomatic (SARA=0-1). Subjects were scanned at both 3T and 7T (Siemens) on
consecutive days. Spectra were acquired from the vermis (10 x 25 x 25 mm
3),
cerebellar hemisphere (17 x 17 x17 mm
3) and pons (16 x 16 x 16 mm
3)
using a semi-LASER sequence (TR=5s, TE= 26-28ms, NEX=64) with VAPOR water
suppression and outer volume saturation (3). Metabolites were quantified with LCModel
(4) using the unsuppressed water signal as reference. Only those measured
reliably (Cramér Rao lower bounds < 50%, cross correlation coefficients r
> -0.5) from more than half of the spectra at a given field strength and
subject group were reported.
Concentrations were corrected for the amount of CSF present in each VOI.
Finally, the quantified metabolites from all brain regions were used as an
input to distance-weighted discrimination (DWD) for classification (5).
Leave-one-out Cross-Validation (LOOCV) was used to correct for overfitting when
estimating the mis-classification rate.
Results and Discussion
Spectra with good signal-to-noise ratio and
spectral resolution were consistently obtained from both patients and controls
at 3T and 7T (Fig. 1). The spectral quality enabled the quantification of a
neurochemical profile consisting of 15 metabolites in the vermis and cerebellar
hemisphere and 13 in the pons at 7T (Fig. 2). Concentrations of major
metabolites (total NAA (tNAA), total choline (tCho), total creatine (tCr), myo-inositol, Glutamate+Glutamine) obtained
at 3T and 7T were strongly correlated (r
2 = 0.86, p <0.001). Higher myo-inositol and tCr and lower tNAA were
detected in patients at both magnetic fields, in agreement with previous
findings (2).
The projection space plot of the DWD
classifier showed a distinct clustering with almost complete separation between
SCA1 and controls at both fields, except for two patients who were
misclassified at 3T, one of whom was again misclassified at 7T (Fig. 3). The
subject who had the smallest SARA among symptomatic patients was incorrectly
classified at 3T, but correctly classified at 7T. One presymptomatic patient, who
did show neurochemical (high tCr and myo-Inositol
and low tNAA) signs of disease was correctly classified with the SCA1 group at
both fields, while the other presymptomatic individual who effectively had no
signs of disease onset, both from the neurochemical and clinical perspective,
was classified as control at both 3T and 7T. In addition, the improved
sensitivity and resolution at 7T resulted in a higher significance level for
group separation than 3T (7T: t = 9.2, p-value = 4.7e-09; 3T: t = 8.5, p-value
= 1.8e-08). Finally, the DWD score
significantly correlated with the SARA score for patients pooled with controls
at both fields (Spearman r=0.8, p<0.0001).
In conclusion, the increased sensitivity
at 7T enabled group separation with higher significance and identification of
subtle neurochemical alterations in early symptomatic disease stage more
robustly than at 3T, while 3T showed great potential by enabling detection of
abnormal metabolite levels in an individual in the presymptomatic disease stage.
Furthermore, the demonstration of a high correlation between MRS and clinical scores
supports a role for the methodology in clinical applications at both magnetic
fields.
Acknowledgements
Supported by NIH R01 NS070815, R01 NS080816, P41 EB015894 and P30 NS076408.References
1. Orr HT, Chung
MY, Banfi S, Kwiatkowski TJ, Jr, Servadio A, Beaudet AL, McCall AE, Duvick LA,
Ranum LP, Zoghbi HY. Expansion of an unstable trinucleotide CAG repeat in
spinocerebellar ataxia type 1. Nat Genet. 1993;4:221-226.
2. Oz G, Hutter D,
Tkac I, Clark HB, Gross MD, Jiang H, Eberly LE, Bushara, KO, Gomez CM.
Neurochemical alterations in spinocerebellar ataxia type 1 and their
correlations with clinical status. Mov. Disord. 2010;25:1253-1261.
3. Oz G and Tkac
I. Short-echo, single-shot,
full-intensity proton magnetic resonance spectroscopy for neurochemical
profiling at 4 T: validation in the cerebellum and brainstem. Magn Reson
Med. 2011;65:901.
4. Provencher,
SW. Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR
Biomed. 2001;14:260-264.
5. Marron JS, Todd
MJ, Ahn J. Distance-weighted Discrimination. JASA. 2007;102:1267-1271.