Mutations in isocitrate-dehyrogenase (IDH1 and IDH2) are present in the majority of low grade gliomas and secondary glioblastomas (~60-90%)¹ and are associated with longer patient survival when compared to IDH wild-type tumors². The mutations impart a functional gain and convert α-ketoglutarate to 2HG, resulting in substantial elevation in 2HG levels. Precise measurement of 2HG is challenging largely due to the interferences from the overlapping multiplets of glutamate (Glu), glutamine (Gln), γ-aminobutyric acid (GABA) and macromolecule signals. The purpose of the study was to develop a new 1H-MRS method that overcomes the spectral challenges at 7T in vivo.Numerical simulations of triple-refocusing difference/summing editing for separation between the strongly-coupled 2.25-ppm resonance of 2HG and the weakly-coupled 2.29-ppm resonance of GABA at 7T were performed incorporating the volume-localizing RF and gradient pulses. A non-slice selective 180° pulse (E180) was implemented between the two 180° pulses of PRESS (Fig. 1). A pair of subecho time-sets was searched for, with numerical analyses, using the following criteria: 1) large 2HG 2.25-ppm signal via subtraction, 2) small GABA, Glu, and Gln signals via subtraction, and 3) small 2HG signal via summing. Two subecho time sets were obtained at a total TE of 90 ms; (t₁,t₂,t₃,t₄)=(20,17,25,28) and (12,36,33,09)ms, both with a 12 ms long E180 tuned to 2.5 ppm (Fig. 1). Following phantom validation, the editing method was tested in 5 patients with gliomas and 5 healthy volunteers. Data were acquired with 32-channel head coil in a 7T whole-body scanner (Philips Medical Systems). Voxel size was 4 – 8 mL depending on the tumor volume identified with T2w-FLAIR. In healthy brain the voxel size was 14mL in gray-matter and 9mL in white-matter region. Data acquisition parameters included TR=2.5s, sweep-width=5kHz, number of sampling points=4096, and signal-averages = 64–256 for each subecho time set. Unsuppressed water data was acquired for eddy current compensation and multi-channel combination. Spectral fitting was performed with LCModel³, using basis spectra calculated incorporating the volume localizing RF and gradient pulses with published chemical shift and J-coupling constants^4,5. Metabolite quantification was performed using water as a reference at 42 M.Figure-2 shows the performance of the triple-refocusing difference editing. The 2HG signal at 2.25 ppm exhibited a positive signal at subspectrum-A and an inverted signal in subspectrum-B, thereby leading to a positive edited signal following subtraction and null signal following summing. In contrast, the weakly-coupled GABA 2.29 ppm resonance was a positive signal with similar amplitude in the subspectra, giving null and edited signal following subtraction and summing, respectively. The Glu 2.35 pm resonance was substantially attenuated in the difference spectrum, enhancing the 2HG detection. Also at the optimized subecho times, the 2HG, GABA and Glu signals were much narrower than at short TE, offering improved separation of the metabolite signals. Figure-3 presents the 2HG signals from simulation and phantom (2HG 10mM and Gly 20mM), and in-vivo edited data from a glioma patient. In consistent with simulation and phantom data, a signal was clearly discernible at 2.25 ppm in the difference spectra, while the sum spectra showed null 2HG signal. Figure-4(a) shows an in-vivo editing result from a glioma patient. 2HG was measured at 4.4 mM, while Glu and GABA where not measurable. Gln was estimated at 1.8 mM. Figure-4(b,c) shows in vivo difference- and summing-edited spectra from five glioma patients. 2HG was measurable in 4 tumors, with CRLBs < 15%, including a low 2HG concentration case (0.6 mM, 13% CRLB). The performance of the summing editing, which was designed for GABA detection, was test in healthy volunteers (Figure-5). In the summing spectra from gray-matter dominant regions the GABA signal at 2.29 ppm was clearly discernible. GABA was estimated to be 0.9 mM (CRLB 9%) and 0.4 mM (CRLB 11%) for gray- and white-matter regions, respectively, in good agreement with prior studies⁶.We report a new triple-refocusing difference-editing method for in-vivo detection of 2HG in brain tumors at 7T, which was designed for edited 2HG signal at 2.25 ppm and suppressed GABA signal at 2.29 ppm, thereby leading to 2HG measurement with minimal GABA interference. With additional benefits from macromolecule attenuation at the long TEs, the editing method offers excellent detectability of 2HG when 2HG concentrations are low. Further study will be required to evaluate the capability of the editing for detection of both 2HG and GABA, which may be important for evaluating the tumor malignancy and neuronal disruption such as seizures in brain tumors.