Abstract
BACKGROUND AND PURPOSE: Despite the prominent peak of N-acetylaspartate (NAA) in proton MR spectroscopy ((1)H-MR spectroscopy) of the adult brain and its almost exclusive presence in neuronal cells, the total amount of NAA, regarded as their marker, is difficult to obtain due to signal contamination from the skull lipids. This article compares the performance of 2 methods that overcome this difficulty to yield the whole-brain NAA signal, important for the assessment of the total disease load in diffuse neurologic disorders. MATERIALS AND METHODS: The heads of 12 healthy volunteers, 3 women and 9 men, 31.0 +/- 7.1 years of age, were scanned at 3T by using 2 nonlocalizing (1)H-MR spectroscopy sequences: One nulls the NAA (TI = 940 ms) every second acquisition by inversion-recovery to cancel the signals of the lipids (T1 << TI) in an add-subtract scheme. The other nulls the signal of the lipids (TI = 155 ms) directly after each acquisition, requiring half as many averages for the same signal-to-noise ratio. Each sequence was repeated 3 times back-to-back on 3 occasions, and the comparison criteria were intrasubject precision (reproducibility) and total measurement duration. RESULTS: NAA nulling is nearly twice as precise in its intrinsic back-to-back (5.8% versus 8.6%) as well as longitudinal (10.6% versus 19.7%) coefficients of variation compared with lipid nulling, but at the cost of double the acquisition time. CONCLUSION: When speed is a more stringent requirement than precision, the new lipid-nulling sequence is a viable alternative. For precision in cross-sectional or longitudinal global NAA quantification, however, NAA nulling is still the approach of choice despite its x2 ( approximately 5 minutes) time penalty compared with the lipid-nulling approach.