Valine and lactate have been recognized as key metabolic markers to distinguish brain abscess from tumor ¹. However, macromolecular contamination constitutes a major impediment for signal quantification of brain metabolites resonating in the spectral region between 1 and 2 ppm when conventional PRESS or STEAM protocols are used. MEGA-PRESS difference editing has recently been proposed for filtering out the resonances of J-coupled metabolites such as GABA ², glutathione ³, ascorbate ⁴, N-acetylaspartateglutamate ⁵ and lactate ⁶ from a background of other metabolite signals. Furthermore, simultaneous MEGA-PRESS editing of two metabolites (glutathione and ascorbate) has been proposed ⁷. In this work, simultaneous MEGA-PRESS editing of valine and lactate from a large background of lipid signal is validated in vitro and demonstrated in a patient with a brain abscess.All experiments were performed on 3T MR systems (Trio for phantom studies and Skyra for brain abscess case, Siemens Healthcare, Germany). The original MEGA-PRESS sequence used for GABA editing was modified to enable complete refocusing of weak J-coupling evolution over variable echo time (TE) intervals by using two frequency-selective refocusing pulses (bandwidth = 30 Hz) with an inter-pulse delay of TE/2. For editing of both valine and lactate with J ≈ 7 Hz, an echo time of TE = 142 ms was used for complete inversion of the target resonances at 1.0 ppm and 1.33 ppm, respectively, in the unedited spectra. The spectrally selective editing pulses were irradiated at ν = 2.25 ppm for valine editing and at ν = 4.11 ppm for lactate editing (shot 1). For the acquisition of the unedited spectra (shot 2), the frequency-selective refocusing pulses were irradiated in the down-field region at 7.15 ppm and 5.29 ppm for valine and lactate editing, respectively, to ensure optimal water suppression in the difference editing spectrum. For combined valine and lactate editing the frequency-selective pulses were irradiated at 2.25 ppm (shot 1) and 4.11 ppm (shot 2) in the two interleaved acquisitions. In vitro experiments were performed in a phantom solution containing valine, lactate and a mixture of soja and olive oil. In vivo data were acquired from a patient with brain abscess. Figure 1 shows the spectra acquired for valine editing, lactate editing and combined valine/lactate editing. All shots contain contamination of lipid resonances in the region of interest. The lactate and valine resonances at 1.33 and 1.0 ppm are contaminated by lipid signal arising from methylene protons (1.3 ppm) and methyl protons (0.9 ppm) respectively. However, the difference spectra exhibit uncontaminated valine triplets and lactate doublets. Combined valine and lactate editing yields a positive valine triplet and a negative lactate doublet in the difference spectrum. PRESS and MEGA-PRESS editing spectra acquired from the brain abscess patient are presented in Fig. 2. Combined valine/lactate editing successfully filters lactate and valine resonances from a strong lipid background. While large lipid peaks (at 1.3 ppm) are visible in the PRESS spectrum as well as in the valine- and lactate-edited MEGA-PRESS shots, the difference spectrum shows only minor lipid signal at 1.3, suggesting some lipid co-editing. However, compaired to the lipid signal in the unedited PRESS spectrum, contamination is negligible in the dual-edited MEGA-PRESS spectrum. MEGA-PRESS was optimized for valine and lactate editing and successfully applied in vivo. Both metabolites can be detected with an editing efficiency of 100% since their most prominent resonances exhibit full inversion at TE = 142 ms. This is in contrast to GABA detection where only the two outer lines of the triplet resonances are inverted, resulting in an editing efficieny of only 50%. The almost identical J-coupling constants of valine and lactate benefit simultaneous MEGA-PRESS editing of these two metabolites. The long echo time required for valine and lactate inversion enables the utilisation of frequency-selective editing pulses with a very small bandwidth (16 Hz). However, it has to be noted that such small-bandwidth pulses make the sequences susceptible to frequency drifts during the scan, e.g. through temperature changes or subject motion, and can strongly decrease the editing efficiency. On the other hand, editing pulses with larger bandwidths may increase co-editing of other resonances, e.g. arising from lipids. An editing pulse bandwidth of 30 Hz empirically turned out to be a good compromise. The proposed method may prove beneficial for the distinction of brain abscesses and tumors where lactate (in tumors and abscesses) and valine (in abscesses only) have been recognized as key metabolic markers.