NIMG-12. 18F-FLUORETHYLTYROSINE POSITRON EMISSION TOMOGRAPHY VERSUS MAGNETIC RESONANCE SPECTROSCOPY IN PREOPERATIVE DETECTION OF NEWLY DIAGNOSED GLIOMAS BY USING MACHINE LEARNING

Abstract

Abstract BACKGROUND Advanced imaging techniques entered the field of neurooncology. In this analysis we compare the diagnostic potential of 18F-fluorethyltyrosine (FET) positron emission tomography (PET) and magnetic resonance spectroscopy (MRS) in their potential to preoperatively predict certain glioma subtypes. AIMS Goal of this analysis ist the evaluation of FET PET and MRS regarding the preoperative prediction of glioma subtypes. METHODS We analyzed 33 patients with histopathologically confirmed newly diagnosed glioma. The patients received FET PET and MRS during one single preoperative diagnostic session. According to the molecular portfolio patients were subdivided in IDH wildtype glioblastoma patients (GBM), IDH wildtype WHO grade II/III glioma patients (Astro_IDHwt), IDH mutant WHO grade II/III glioma patients without 1p/19q codeletion (Astro_IDHmut) and with 1p/19q codeletion (ODG). Mean and maximum tumor-to-brain ratio (TBRmean and TBRmax), N-acetylaspartate, choline and creatine peaks were correlated with postoperative tumor diagnosis. To gain generalizable implications we subdivided the study cohort into a development and validation subcohort. A support vector machine model was fitted to the development subcohort and evaluated on the validation subcohort. Receiver operating characteristic curve served to assess model performance. RESULTS GBM patients had highest TBRmax and TBRmean values (mean: 3.5 and 3.8) and the ODG patients showed the second highest TBRmax and TBRmean values (mean: 2.6 and 3). The distribution of MRS markers exhibited to clear trend. The performance of glioma subtyping was comparatively low for the TBR values (AUC: 0.68) and even lower for the MRS markers (AUC: 0.60). These results are in line with preliminary investigations performed by our institute for the comparison of 11C-methionine PET with MRS in preoperative glioma subtyping. CONCLUSIONS FET PET and MRS bear limited potential in glioma subgrouping. However, FET PET appears to be slightly superior. Investigation in a larger cohort is required to draw definite conclusions.