Quantifying global-brain metabolite level changes with whole-head proton MR spectroscopy at 3T

Abstract

Background and purpose: To assess the sensitivity of non-localized, whole-head ¹H-MRS to an individual's serial changes in total-brain NAA, Glx, Cr and Cho concentrations - metabolite metrics often used as surrogate markers in neurological pathologies.

Materials and methods: In this prospective study, four back-to-back (single imaging session) and three serial (successive sessions) non-localizing, ~3min ¹H-MRS (TE/TR/TI=5/10⁴/940ms) scans were performed on 18 healthy young volunteers: 9 women, 9 men: 29.9±7.6 [mean±standard deviation (SD)] years old. These were analyzed by calculating a within-subject coefficient of variation (CV=SD/mean) to assess intra- and inter-scan repeatability and prediction intervals. This study was Health Insurance Portability and Accountability Act compliant. All subjects gave institutional review board-approved written, informed consent.

Results: The intra-scan CVs for the NAA, Glx, Cr and Cho were: 3.9±1.8%, 7.3±4.6%, 4.0±3.4% and 2.5±1.6%, and the corresponding inter-scan (longitudinal) values were: 7.0±3.1%, 10.6±5.6%, 7.6±3.5% and 7.0±3.9%. This method is shown to have 80% power to detect changes of 14%, 27%, 26% and 19% between two serial measurements in a given individual.

Conclusions: Subject to the assumption that in neurological disorders NAA, Glx, Cr and Cho changes represent brain-only pathology and not muscles, bone marrow, adipose tissue or epithelial cells, this approach enables us to quantify them, thereby adding specificity to the assessment of the total disease load. This will facilitate monitoring diffuse pathologies with faster measurement, more extensive (~90% of the brain) spatial coverage and sensitivity than localized ¹H-MRS.

Keywords: Diffuse neurological disorders; MR spectroscopy; Metabolic quantification; Proton MR spectroscopy; Whole brain MR spectroscopy.

Fig. 1

VeSPA analysis of three serial (inter-scan) whole-head ¹H-MRS from a 35-year-old male. Top, a: The raw spectrum and the NAA, Glx, Cr, Cho and myo-inositol (mI) model functions used in the VeSPA analyses, all on the same intensity and chemical shift scales. Center, b: Raw data overlaid with VeSPA fitted baseline and model functions in a. Bottom, c: Residual from the difference between the “fitted+baseline” and “raw” data. Note the effective lipid suppression performance of the sequence, the exceedingly high SNR, reflecting signal from the entire head, the visually similar raw data between the three serial scans, reflecting the repeatability of the sequence, and the near vanishing residual, reflecting the goodness of VeSPA’s spectral modeling performance.

Fig. 2

Box plots showing the first, second (median) and third quartiles, ±95^th percentiles (whiskers), outliers (*) and means (+) of the inter-scan measurement (white) and intra-scan measurement (gray) mean CV distributions of the four metabolites in N=18 subjects. Note that: (i) all metabolites’ inter-scan CVs are typically lower than 10%; and (ii) the intra-scan CVs are, as expected, much lower (<5%) that the inter-scans’, leading to the prediction intervals in the Table 1.