Glutamate (Glu) and γ-aminobutyric acid (GABA) play a significant role in cognitive function, thus studying the levels of GABA and Glu in patients with cognitive impairment is important.¹ Magnetic resonance spectroscopy (MRS) provides a non-invasive tool for the biochemical characterization of pathophysiological processes in the brain. A technologically similar approach has been previously applied to the study of several brain biochemicals in cognitive impairment patients, including N-acetylaspartate (NAA), glutamate (Glu), myo-inositol (mI), choline (Cho) and creatine (Cr). However, there is not as much research about GABA with routine MRS techniques, because it is present in the normal brain with a distinctly lower concentration (2 mmol/l), and its spectral resonances show strong overlap with signals of other metabolites, making its quantification challenging. The purpose of our study was to use the MEGA-PRESS sequence, a spectral-editing MRS technique, to compare the Glu and GABA levels of the right hippocampus (rHP) and anterior cingulate cortex (ACC) in subjects with Alzheimer's disease (AD), mild cognitive impairment (MCI), and normal elderly controls (NEC). We also compared these metabolites between normal elderly controls (NEC) and normal young controls.14 normal young controls,15 NEC, 18 MCI and 7 probable AD patients were included. All data was acquired on a Siemens 3T MR scanner (MAGNETOM Trio Tim, Siemens Healthcare, Erlangen, Germany) using a 32-channel head coil. Prior to the MRS measurement, T1-weighted 3D anatomical reference data was collected for the spectroscopic voxel localization. The MRS voxels were set on ACC (V=40×40×25mm³) and the rHP (V=40×20×20mm³) in this study (Fig. 1). MRS data was acquired using the 1H MEGA-PRESS sequence. This sequence was composed of alternate acquisition of standard PRESS spectra with additional frequency-selective refocusing pulses affecting spins resonating at 1.9 ppm (edited spectrum) and 7.5 ppm (non-edited spectrum). GABA and Glu signals were derived from the difference between the edited and non-edited spectra (Fig. 2). The parameters of this sequence were as follows: TR=2000ms, TE=68ms, acquisition band width=1000Hz, acquisition time=4:44min. Throughout the process, the water suppression and automatic shimming were performed beforehand automatically. In this sequence, the spectral signal represents a mixture of Gln and Glu (Glx) which can only be acquired simultaneously but not Glu on its own, so Glx will be used instead of Glu in this study. The data was post-processed using Spectroscopy tools available in the syngo MR software (Siemens Healthcare, Erlangen, Germany). The levels of Glx and GABA were normalized to the commonly used internal reference Cr. The group differences were tested using SPSS.There was no significant difference in gender and age between the three elderly groups, and no significant difference in gender between normal young and elderly controls. In ACC, the levels of Glx/Cr and GABA/Cr were lower and there were significant differences between the NEC group and normal young controls group in Glx/Cr(P=0.000) and GABA/Cr(P=0.041). For Glx/Cr, there was a significant difference in level between AD group and NEC group(P=0.011). Between AD group and MCI group, there was also a significant difference (P=0.010), but for NEC group and MCI group, there was no difference. For GABA/Cr, the level did not decrease, and there was no significant difference between the three groups. In the rHP, the levels of Glx/Cr and GABA/Cr were lower, but there was no significant difference in both levels between NEC group and young normal controls group. Between the three elderly groups, the level of Glx/Cr was lower, but had no significant difference. GABA/Cr did not follow this trend, and there was no significant difference.For the location of measuring the Glx and GABA levels, ACC has an advantage over the rHP. Levels of Glx and GABA decrease with age. In the AD process, level of Glx decrease, and the balance between excitatory -inhibitory neurotransmission systems was broken, so Glx might be a more sensitive index for diagnosing AD.The GABA and Glx levels measured using the MEGA-PRESS MRS technique seem to reflect different cognitive impairment stages related to aging.