P10.17 Predicting isocitrate dehydrogenase genotype in malignant glioma using FET PET radiomics

Abstract

INTRODUCTION: Accurately determining isocitrate dehydrogenase (IDH) genotype pre-operatively may have both prognostic and diagnostic value that contributes to treatment decision-making. Here, we investigated the potential of O-(2-[18F]fluoroethyl)-L-tyrosine (FET) PET textural features in comparison with standard static and dynamic FET PET parameters for differentiation of IDH-mutated anaplastic astrocytoma (mut-AA) from IDH-wildtype glioblastoma (wt-GBM). MATERIALS AND METHODS: Thirty-six patients with histologically confirmed malignant glioma (26 wt-GBM, 10 mut-AA) based on the updated WHO classification 2016 underwent a dynamic FET PET. Volumes-of-interest were defined on summed images from 20-40 min post-injection (p.i.) by a 3-dimensional auto-contouring process using a tumor-to-brain ratio (TBR) of 1.6 or more. For each lesion, TBRs and the time-activity curves (TACs) of the FET uptake were determined. The TACs were used to evaluate the dynamic FET PET parameters time-to-peak (TTP), slope (slope of linear regression line 20-50 min p.i.) and intercept (intercept of linear regression line with y-axis). Additionally, 39 textural parameters were calculated using the software LifeX (lifexsoft.org). The diagnostic accuracy of TBRs, TTP, slope, intercept, and textural parameters for prediction of IDH genotype was evaluated using ROC analyses. In order to further increase the diagnostic accuracy, parameters were combined using linear logistic regression. RESULTS: The standard static FET PET parameters TBRmean and TBRmax showed a diagnostic accuracy to discriminate between IDH genotypes of 78% and 75%, respectively. The dynamic FET PET parameters TTP, slope, and intercept yielded an accuracy of 69%, 81% and 72%, respectively. The best five textural parameters showed a diagnostic accuracy of 81%. Parameter combinations could not further increase the diagnostic accuracy. CONCLUSIONS: Textural feature analysis may have the potential to predict the IDH genotype of malignant gliomas without the need for acquiring a more time consuming dynamic FET PET acquisition. Textural parameters might yield additional valuable information about tumor heterogeneity and should be further evaluated prospectively in larger cohort of patients.