We directly compare long TE semi-LASER acquisition schemes at 3T and 7T in glioma patients to assess the advantages of ultra-high field (UHF) for in vivo 2-hydroxyglutarate detection.2-hydroxyglutarate (2-HG) accumulates in the majority (~80%) of low grade gliomas as a product of somatic mutation of isocitrate dehydrogenase (IDH)^1,2. Robust detection of the ‘tumour metabolite’ 2-HG, therefore, presents a unique clinical opportunity to non-invasively stratify glioma patients and target therapy³. Several 2-HG-specific MRS acquisition methods at 3T have been suggested, such as J-editing⁴ and PRESS^5,6, yet detection poses a significant challenge because of overlaps with glutamate (Glu), glutamine (Gln) and γ-aminobutyric acid (GABA) signals. Spectral patterns are also modulated by compartmental artifacts, chemical shift and B₁ inhomogeneity. Semi-LASER localisation, with low chemical shift displacement and insensitivity to B₁ inhomogeneity⁷, has recently been proposed for 2-HG detection at 3T (TE = 110ms)⁸ and shows reduced signal overlap from Gln and Glu as well as resolved 2-HG peak at 1.9 ppm (Fig. 1: top), which leads to lower fitting cross-correlations. Additionally, 2-HG detection with semi-LASER has been demonstrated at UHF (≥7T) using the same echo time⁹ (Fig. 1: bottom). UHF benefits from increased SNR and spectral resolution, but experiences greater B₁ inhomogeneity and chemical shift. Hence, compartmental J-evolution artifacts may lessen UHF gains for 2-HG. As clinical 7T systems begin to emerge, there is a need to assess the advantage they provide relative to current 3T systems. We therefore performed a direct comparison using long TE (110ms) semi-LASER, with broadband adiabatic refocussing, on six glioma patients at 3T and 7T.Six glioma patients, with histopathologically confirmed IDH status following surgical resection (5 IDH-mutant, 1 wild-type IDH), were scanned over two sessions on 3T and 7T whole body MR systems (Siemens, Erlangen) with 32-channel receive array head-coils. Our [20x20x20]mm³ voxel was positioned, with reference to a 1mm isotropic resolution T1-MPRAGE image, in tumour regions of maximum homogeneity. Care was taken to position voxels in similar positions at both field strengths. We use 2-HG-optimised semi-LASER sequences at 3T⁸ and 7T⁹ (TE = 110ms, TR = 3 / 5-6 s, sw = 6kHz, np = 2048, NT = 71-128) with outer volume suppression (OVS) and VAPOR water suppression¹⁰. First- and second-order shimming was performed with GRESHIM¹¹ and first-order shims were adjusted at 7T using FASTMAP¹². An unsuppressed water spectrum was acquired for eddy-current correction. LCModel¹³ was used to fit metabolite concentrations using fully-localised basis sets of 20 metabolites, generated from density matrix simulations with real refocussing pulses in Matlab (Mathworks, Inc). Fitted concentrations are reported relative to total creatine (tCr). Only metabolite concentrations with Cramér-Rao lower bound of fitting (CRLB) ≤ 40% are reported. High quality spectra were achieved across field strengths with similar voxel positioning (Fig. 2). Using semi-LASER (110ms) at 3T and 7T resulted in clear 2-HG peaks in the spectra at 1.9 ppm (and additionally 2.25 ppm at 7T) in all IDH-mutant subjects (Fig. 3). As expected, 2-HG could not be quantified at either field strength in the wild-type IDH (i.e. non-mutated) patient (7T: 55%; 3T: 47%) (Fig. 3, right). The mean error of fitting (CRLB) for 2-HG in IDH-mutant tumours was much lower at 7T (5.8 ± 2.6%) than at 3T (14.8 ± 6.5%). We find no significant difference between the relative concentration values of 2-HG at 3T or 7T (Fig. 4). Similar concentrations of Glu, Gln, total Cho (tCho) and inositol (Ins) were found at both field strengths, with slightly reduced CRLBs at 7T (Fig. 5). The metabolites GABA and glycine (Gly) were undetected in all IDH-mutant tumours at 3T, yet detected in small quantities at 7T in three patients. We confirmed that semi-LASER at 3T and 7T was robustly able to identify increased 2-HG levels arising from IDH-mutants in a sample of six patients. The increased spectral resolution at 7T, in addition to the unambiguous inverted 2-HG peak at 2.25 ppm, results in reduced overlap with Glu/Gln/GABA thus improving the fitting. The absence of GABA at 3T may be explained by peak overlap with 2-HG at 1.9 ppm, which also resulted in larger CRLBs. The additional detection of Gly at 7T may allow further metabolic profiling of tumours. We conclude that semi-LASER at TE=110ms may provide a unique clinical opportunity to simultaneously obtain a robust in vivo measurement of 2-HG as well as detect other cancer-specific metabolic markers, particularly at 7T. Further work, on a larger cohort of patients, is required to fully establish these findings.