Glucose has multiple functions in the brain, and there is interest in estimating in vivo concentrations rather than merely the uptake determined by nuclear medicine. Glucose can be estimated using magnetic resonance spectroscopy, but measurement is difficult due to its multiple J-coupled proton signals overlapping with other metabolite signals. To minimize the effect of interfering signals, echo time (TE) values between 60 and 95 ms were averaged, and the loss in signal due to the T2 effect was corrected in both the estimation of glucose concentration and in creation of the basis files for fitting. The effectiveness of the TE-averaging method was evaluated by measuring the glucose concentration in fasted rats before and after feeding. The brain glucose in all rats increased after feeding with fasted and fed glucose-to-creatine ratios of 0.15 ± 0.03 and 0.24 ± 0.04, respectively. Data at a short TE of 13 ms measured ratios of 0.30 ± 0.16 and 0.36 ± 0.24 for the fasted and fed rats, respectively, demonstrating the difficulty in obtaining reliable glucose measurements at short TE. Overall, TE averaging minimizes the influence of macromolecular signals and nearby peaks to give precise, consistent estimates of glucose. Magn Reson Med 70:301–308, 2013. © 2012 Wiley Periodicals, Inc.