Magnetic resonance spectroscopy (MRS) as a noninvasive technique has been widely used in medical imaging. Currently, metabolite concentrations, which could be detected by lots of softwares, are represented mainly in ratios (i.e., relative to creatine) or with absolute concentrations. In this contribution, we investigate the reliability and accuracy of absolute concentration quantified by Linear Combination of Model spectra (LCModel), and data were collected by single voxel technique (SVT) and multi-voxel technique (MVT). Ten amnestic mild cognitive impairment (aMCI) patients and one phantom were studied. In order to improve the clinic practicality of MRS, the factors affecting the accurate quantification of absolute concentration were investigated using LCModel when we used SVT. By changing the number of scans, the tendency of metabolite concentrations, standard deviation (SD) and Signal to Noise Ratio (SNR) was obtained. It was found that if the number of scans was set to be 64, the concentration of creatine, inositol, glutamate, and choline was reached to its maximum, and if the number of scans was increased, the concentration of each metabolite was nearly same as 64 times. On the other hand, with the increase of scanning times, the standard deviations of metabolite concentration decreased and the signal-to-noise ratios increased.