Higher field strength for proton MR spectroscopy.

MR imagers with magnetic fields (B0) greater than 1.5 T are offered by all major manufacturers. Although 4-, 6-, 7-, and even 8-T, whole-body instruments are currently operational, the most common high-B0 systems are the nearly 100 installed with 3-T magnets. The demand for bigger B0 systems has been driven almost exclusively by functional MR imaging (1); however, their proliferation and Food and Drug Administration approval for most MR systems raises the question of whether the associated cost and technical complications benefit other applications such as MR spectroscopy. Indeed, several comparisons were already reported on the signal-to-noise ratio (SNR) and spectral resolution improvements realized from raising B0 from the ubiquitous 1.5 T to 3, 4, and 7 T (2–5); however, common to all were that 1) few ( 10) healthy subjects were examined; therefore, 2) none evaluated capability to systematically identify or differentiate abnormal states; 3) none observed the theoretical gains ( B0) in either attribute; and 4) proton MR spectroscopy in the brain was used, which requires sequences of various echo times, fat, and water suppression. In this issue of the AJNR, Kantarci et al tackle, for the first time, the first two above-referenced objectives. This was done with single-voxel proton MR spectroscopy in the posterior cingulate gyri in 20 subjects with mild cognitive impairment, a condition thought to precede Alzheimer disease; 20 with symptoms consistent with Alzheimer disease; and 41 agematched control subjects. All 81 subjects underwent proton MR spectroscopy at 1.5 and 3 T. The metrics compared were ratios to creatine (Cr), a metabolite reflecting high-energy phosphate reserves in the cytosol of neurons and glial cells, of the neurotransmitters glutamine (Gln) and glutamate (Glu), and myoinositol (MI), a marker of gliosis, obtained at short (30-millisecond) echo times; and N-acetylaspartate (NAA), a neuronal marker, and choline (Cho), a membrane turnover indicator, acquired at intermediate (135-millisecond) echo time. NAA/Cr and Cho/ Cr, were also acquired and included in the analysis of the short–echo time experiment. The authors’ goal was to ascertain which metabolic characteristics, field strengths, and echo times were most appropriate to differentiate the three subgroups. Eighty-one quality-control acquisitions on a brainmetabolite phantom, the “best-case scenario” taken at both field strengths, provided an early indication of findings: the coefficients of variations (CV) of the metabolite ratios (to Cr) were lower and less variable at 1.5 T. It is no surprise, therefore, that this trend was repeated in the 41 elderly control subjects; only the Cho/Cr ratios were statistically different. Because the CV SD/mean, assuming the SD of the measurements is approximately the same at each field strength, the finding that CV at 3 T is not half that at 1.5 T indicates that the theoretical signal intensity gains were not approached. In fact, with an echo time of 30 milliseconds (ie, a sequence providing minimal T2 loss), the SNR of Cr at 3 T was only 23% better than that at 1.5 T, whereas the line width more than doubled. At 1.5 T, metabolite ratios of moderately cognitively impaired patients fell, as expected, between the elderly control subjects and those with Alzheimer disease, an additional indication of the intermediate nature of this condition. Specifically, MI/Cr and Cho/Cr acquired at echo times of 30 milliseconds progressively and significantly (P .05) higher in control subjects as compared with ratios acquired in subjects with moderate cognitive impairment and Alzheimer disease. The NAA/Cr acquired at an echo time of 135 milliseconds showed progressively lower NAA/Cr and NAA/MI differentiated moderately cognitively impaired subjects from those with Alzheimer disease. By contrast, at 3 T, no metabolite ratios differentiated control subjects from those with mild cognitive impairment, or the latter subjects from those with Alzheimer disease. Although a trend of decreasing (Glu Gln)/Cr was observed in moderately cognitively impaired subjects and those with Alzheimer disease, and was most pronounced in the latter group, the decrease did not reach statistical significance even in this large cohort. Consequently, Kantarci et al conclude that, in light of the current technology, 3-T proton MR spectroscopy offers no diagnostic performance advantage over the venerable 1.5-T field strength when applied to differentiating pathologic metabolism in the elderly. Two technical causes were identified for this shortfall: 1) shimming (only first order x, y, z was used at either field strength) and 2) the effective contraction of T2s with field strength. Indeed, the posterior cingulate gyri were probed instead of the hippocampus or entorhinal cortex because of the poorer B0 field homogeneity at the latter regions. Not discussed was the quantification by means of metabolite ratios rather than absolute metabolite concentrations. Although ratios benefit from cancellation of difficult to measure multiplicative factors, such as voxel partial CSF volume, instrumental gain, and interpersonal coil loading differences, their variation is the sum of individual components. Furthermore, ratios are also implicitly assumed to reflect the behavior of the numerator, because the denominator’s level (frequently Cr) is presumed to be constant. This assumption was recently criticized by Weiner et al (6), who argued that absolute Cr level variations, especially as a function of age, may have also contributed to the lack of differential correlation between ratios and clinical status. Editorials