Untitled

ProFit-based Quantitation of Cerebral Metabolites using 2D L-COSY at 3T
E. Frias-Martinez1, N. Rajakumar1, X. Liu1, A. Singhal1, S. Banakar1, S. Lipnick1, G. Verma1, S. Ramadan1, A. Kumar2, and
M. A. Thomas1
1Radiological Sciences, UCLA, Los Angeles, CA, United States, 2Department of Psychiatry, UCLA, Los Angeles, CA, United States Introduction: The ProFit algorithm has been developed for quantitation of 2D JPRESS (J-resolved spectroscopy) signals [1]. When
compared with one-dimensional (1D) MRS data processed using LC Model, the results indicated an increase in the number of
metabolites that can be detected (from 8 in LC Model to 17 in ProFit) with decreased Cramer-Rao lower bound (CRLB) values [1,2].
In contrast to JPRESS, localized 2D shift correlated spectroscopy (L-COSY) [3] has improved spectral dispersion along the second
spectral dimension, which may offer improved specificity for quantitation. The goal of this work was to adapt ProFit for processing
2D L-COSY spectra and to test if the increased dispersion along the second dimension improves the overall quantitation accuracy.
Methods: A maximum-echo sampling 2D L-COSY sequence containing three slice-selective RF pulses (900, 1800, 900) was
implemented on a Siemens 3T Trio-Tim scanner (Siemens Medical Systems, Germany) [3]. The following parameters were used:
TR/TE=2s/30ms, 3x3x3cm3 voxel, 8 averages per Δt1 and 100 Δt1 increments. A white matter brain phantom containing fifteen
metabolites (pH=7.3) was used for processing 10 in vitro measurements. Eight healthy volunteers have been investigated so far. The
2D MRS voxel was mainly localized in the occipital white/gray matter. The modified ProFit algorithm uses MATLAB (Mathworks,
Natlick, MA, USA, ver. 7.3) and was executed on an Intel 2.8GHz with Windows XP. ProFit algorithm uses prior knowledge
constraints and a combined linear and non-linear optimization for fitting concentrations. The algorithm uses a prior knowledge basis-
set generated using the GAMMA library [4] in combination with the chemical shift and J-coupling values reported the literature [5].
The main modifications implemented for quantitation of L-COSY include: (1) programming COSY sequence using GAMMA for
obtaining prior knowledge metabolite spectra; (2) inclusion of optimal sine bell filters along t1 and t2 dimensions both for display and
quantitation, and (3) redefining the size of the region of interest (ROI) used for covering the entire COSY cross peaks along the 2nd
dimension. Prior knowledge generated for L-COSY included 20 metabolites: creatine (Cr), N-acetylaspartate (NAA),
glycerylphosphocholine (GPC), phosphorylcholine (PCh), free choline (Cho), alanine (Ala), aspartate (Asp), γ-aminobutyric acid
(GABA), glucose (Glc), glutamine (Gln), glutamate (Glu), glycine (Gly), glutathione (GSH), lactate (Lac), myo-inositol (mI), N-
acetylaspartylglutamate (NAAG), phosphoethanolamine (PE), taurine (Tau), scyllo-inositol (Scy) and ascorbate (Asc). 2D L-COSY
spectra were then processed using the modified ProFit code and the measurement accuracy was characterized using CRLB [6].
Results and Discussion: Fig. 1 presents a typical 2D L-COSY spectrum recorded in human brain in vivo. Shown in Fig. 2 is the
prior knowledge obtained from combining the spectra of the metabolites considered for fitting [4-5], where the cross peaks produced
by L-COSY can be clearly seen, and Fig. 3 shows the residual obtained after the fitting process. ProFit quantitation of the 3T L-COSY
spectra in vivo enabled the quantitation of all the previously enumerated metabolites, with concentrations expressed as ratios to
creatine of (mean±SD): 1(Cr), 1.27±0.21(NAA), 0.1±0.02(GPC), 012±0.02(PCh), 0.1±0.02(Cho), 0.12±0.03(Ala), 0.43±0.01(Asp),
0.41±0.2(GABA), 0.31±0.1(Glc), 0.43±0.1(Gln), 1.48±0.3(Glu), 0.12±0.09(Gly), 0.15±0.07(GSH), 0.14±0.05(Lac), 0.83±0.1(mI),
0.33±0.09(NAAG), 0.29±0.07(PE), 0.21±0.08(Tau), 0.05±0.00(Scy), 0.3±0.21(Asc). These concentrations, for the best part, agree
with those reported in the literature when using ProFit processed JPRESS [1], but, the CRLBs (mean) reported by L-COSY, for in vivo
and in vitro, are smaller than those reported by ProFit-JPRESS.

Conclusion: This work demonstrates the first quantitation of L-COSY spectra using a modified version of ProFit algorithm. Initial
results indicate that 20 metabolites can be quantitated successfully with better accuracy than JPRESS. This preliminary study indicates
that the initial assumption, that improved dispersion of COSY when compared to JPRESS implies better quantitation, holds true.
References:
1. Schulte, RF and Boesiger, P. NMR Biomed 2006; 19: 225-263
2. Provencher SW. Mag. Reson. Med. 1993; 30: 672-679.
3. Thomas MA, Yue K, Binesh N et al. Magn. Reson. Med. 2001; 46: 58-67
4. Smith SA, Levante TO, et al. J. Magn. Reson. 1994; 106: 75-105
5. Govindaraju V, Young K and Maudsley AA. NMR Biomed 2000; 13: 129-53
6. Cavassila S, Deval S, Huegen C, et al. NMR Biomed. 2001; 14: 278-283

Source: http://enriquefrias-martinez.info/yahoo_site_admin/assets/docs/rad1f8681.8370331.pdf