Radiation therapy with adjuvant temozolomide (TMZ) is the standard of care for treating Glioblastoma Multiforme (GBM). These patients may exhibit enhancing lesions on magnetic resonance imaging (MRI) within six months after treatment. About 20-30% of these lesions may be pseudo-progression (PsP), which show improvement without further treatment beyond TMZ, rather than a true progressive tumor (TP), which does not (1). These lesions often present as an admixture of treatment effect and viable tumor, making metabolic characterization with single-slice or single-voxel spectroscopy techniques difficult. High resolution 3D echo-planar spectroscopic imaging (EPSI) may help to assess this tumor heterogeneity and reduce the need for histology with an invasive second biopsy to differentiate PsP from TP (2). The purpose of this study was to evaluate 3D EPSI in differentiating PSP from TP, using increases in choline (Cho) (indicative of tumor (3,4)) as a biomarker of TP.21 patients were scanned using 3D EPSI on a Siemens 3T whole-body scanner. Seven patients were histology-confirmed or clinically-suspected PsP (defined as a neoplasm containing <25% tumor), and another seven were classified as TP with > 25% recurrent tumor on histological sections. Four patients (with partly or mostly recurrent tumor) yielded spectra that could not be reliably quantified and three other patients had not yet undergone repeat surgery. These seven patients were excluded from the analysis. EPSI scan parameters included: TE/TR = 17.6ms/1550ms, 280x280mm FOV, 180 mm total thickness, 512 complex points with 616 Hz bandwidth and 15 min scan time. Data were post-processed using the Metabolite Imaging and Data Analysis System (MIDAS) (2) which zero-filled the acquired 50x50x18 resolution to 64x64x32 with a final effective voxel size of 1 ml. Contrast-enhanced T1 and fluid attenuated inversion recovery (FLAIR) T2-weighted MRI scans were acquired and co-registered to facilitate tumor segmentation. Using custom IDL and MATLAB based scripts, spectroscopic parametric maps were segmented into three regions: contrast enhancing voxels, voxels adjacent to enhancement (peritumoral) and voxels appearing hyperintense under FLAIR (distant peritumoral). Quantification of Cho, creatine (Cr) and N-acetylaspartate (NAA) was performed using prior-knowledge fitting, with particular focus on the comparison of Cho/Cr and Cho/NAA ratios between segmented regions in the neoplasm and normal contralateral tissue.Figure 1 shows registered T1 and FLAIR images, under-sampled to match the EPSI spatial resolution, along with single and multi-slice Cho/Cr maps from a representative EPSI scan of a TP patient. Figure 2 shows normalized histograms of the Cho/Cr ratio from enhancing lesions in PsP (C) and TP (D) compared to contralateral tissue from the same slice level in PsP (A) and TP (B). Table 1 shows the median factor by which Cho/NAA and Cho/Cr ratios were higher than the enhancing, peritumoral and distal peritumoral areas compared to contralateral regions. For example, median Cho/NAA ratio was 2.57 ± 0.63 times higher in the contrast-enhancing area of TP patients compared to the contralateral region. The elevation in Cho/NAA was less (1.70 ± 0.48) among PsP patients, which was significantly different than TP (using two-sided student’s t-test and a p-value of 0.013). A general trend of higher Cho/NAA ratios was observed among TP patients compared with the corresponding regions in PsP in all regions studied.These results indicate that 3D EPSI can adequately assess metabolic heterogeneity in treatment response and aid in differential diagnoses of PsP versus TP which can be used for patient management in that the PsP patients can be spared invasive and potentially morbid surgery, while TP patients can be offered earlier alternative treatments or repeat surgery, increasing their chances of survival.