The reproducibility of the MEGA‐PRESS (MEshcher‐GArwood Point RESolved Spectroscopy) MR spectroscopy sequence for the measurement of gamma‐ aminobutyric acid (GABA) is addressed, focusing on optimizing the number of repetitions at two voxel locations in the human brain and associated possibilities in analysis tools.

Two 20‐min MEGA‐PRESS acquisitions were run (echo time = 68 ms, repetition time = 1800 ms, repetitions = 328): one from a 21 mL volume in the anterior cingulate cortex (ACC) and one from a 22 mL volume in the left Broca's area in 21 healthy male volunteers (age 32 years ± 6[SD]). Subjects were scanned twice with identical protocols, 1 week apart. Data were acquired on a 3 Tesla GE Discovery 750 scanner using a 32‐channel head coil. Spectroscopy data were partitioned into shorter epochs, numerically equivalent to scans of progressively increasing duration, and compared both within and between sessions. Three different analysis schemes were applied: (1) Vendor prototype preprocessor, with quantification by LCModel. (2) Pure Gannet pipeline. (3) Preprocessing with Gannet, and quantification with LCModel. The coefficient of variation (CV) were calculated as a measure of reproducibility.

Increasing the number of repetitions showed improvements for within‐ and between‐session reproducibility up to around 218 repetitions. (CV ranging from 4 to 14%). Gannet combined with LCModel approach proved the best method. (CV = 4–5%). Measurements from the ACC area had higher CVs than the Broca area. (CV = 6–14% versus 4–7%).

Measurement in the Broca area yields better reproducibility than the ACC. With appropriate acquisition times and preprocessing tools, measurements from the ACC area are also reliable.

Level of Evidence: 1

Technical Efficacy: Stage 1

J. MAGN. RESON. IMAGING 2017;46:421–430