Abstract
Introduction:
To evaluate the test-retest and inter-sequence reliability of neuro-metabolite concentrations in human brain measured by three different MR spectroscopy sequences: standard Point RESolved Spectroscopy (PRESS), J-resolved PRESS (JPRESS), and MEscher-GArwood Point RESolved Spectroscopy (MEGA-PRESS). For all three sequences, Inner Volume Saturation (IVS) was employed to minimize the chemical shift displacement. The present work aims at assessing the expected improvements in terms of test-retest reliability of the JPRESS and MEGA-PRESS sequences, compared to conventional PRESS
Methods:
18 healthy participants were scanned twice with identical protocols (range: one day to one week between the two sessions). Data were acquired on a 3-Tesla Philips MRI (Magnetic Resonance Imaging) scanner using the 32-channel SENSE head coil. Metabolite concentrations were estimated using LCModel (for PRESS and MEGA-PRESS) or ProFit2 (for JPRESS). The test-retest reliability was evaluated by Wilcoxon signed-rank tests, Pearson's r correlation coefficients, Bland-Altman plots (BA), intraclass correlation coefficients (ICC), and by inspecting the averages of the coefficients of variation (CV). The inter-sequence reliability was assessed with Wilcoxon signed-rank tests, Pearson's r correlation coefficients, and BA plots.
Results:
For GABA (gamma-aminobutyric acid), only the MEGA-PRESS sequence showed a moderate correlation between measurements (r = 0.53, ICC = 0.5, CV = 8.8%) and the related BA plot indicated for all three sequences a good test-retest agreement (p > 0.05). JPRESS provided less precise results (r = 0.38, ICC = 0.3, CV = 49.7%). Only PRESS was observed to be insensitive to GABA. In the case of Glx [Glutamate (Glu) + Glutamine (Gln)], the r and ICC values for PRESS (r = 0.86, ICC = 0.9, CV = 2.9%) and MEGA-PRESS (r = 0.69, ICC = 0.7, CV = 5.3%) reflect higher correlations, compared to JPRESS (r = 0.38, ICC = 0.4, CV = 20.1%). Furthermore, the inter-sequence reliability analysis demonstrated that JPRESS and MEGA-PRESS exhibit a significant agreement between them (p = 0.17, from the BA plot), whereas both the sequences report differences, if compared separately to PRESS (p value of 0.005 and <0.001, respectively). Finally, JPRESS exhibits a higher level of agreement for distinguishing Glu signals from Gln.
Conclusions:
MEGA-PRESS and JPRESS prove to be suitable for the detection of GABA, with a competing trade-off between accuracy and precision. For Glx, all three sequences are a viable solution for MRS (Magnetic Resonance Spectroscopy) investigations, although PRESS and MEGA-PRESS appear to be more precise than JPRESS. Nevertheless, JPRESS exhibited the highest reliability in the distinction between Glu and Gln. This may reflect the higher accuracy of JPRESS and MEGA-PRESS compared to PRESS in estimating Glx and also the Glu and Gln concentrations separately.
