NIMG-14. MACHINE LEARNING-BASED EVALUATION OF STATIC AND DYNAMIC FET-PET FOR THE DETECTION OF PSEUDOPROGRESSION IN PATIENTS WITH IDH-WILDTYPE GLIOBLASTOMA

Abstract

Abstract BACKGROUND Pseudoprogression (PSP) detection in glioblastoma has important clinical implications and remains a challenging task. With the significant advances provided by machine learning (ML) in health care, we investigated the potential of ML in improving the performance of PET using O-(2-[18F]-fluoroethyl)-L-tyrosine (FET) for differentiation of tumor progression from PSP in IDH-wildtype glioblastoma. METHODS We retrospectively evaluated the PET data of patients with newly diagnosed IDH-wildtype glioblastoma following chemoradiation. All patients presented imaging findings suspected of PSP/TP on contrast-enhanced MRI. For further diagnostic evaluation, patients underwent subsequently an additional dynamic FET-PET scan. The modified Response Assessment in Neuro-Oncology (RANO) criteria served to diagnose PSP. To develop a robust ML model, we trained a Linear Discriminant Analysis (LDA)-based classifier using FET-PET derived features on a training cohort and validated the results on a separate test cohort. The results of the ML model were compared with a conventional FET-PET analysis using the receiver-operating-characteristic (ROC) curve. RESULTS Of the 44 patients included in this study, 14 patients were diagnosed with PSP. The mean (TBRmean) and maximum tumor-to-brain ratios (TBRmax) were significantly higher in the TP group as compared to the PSP group (p=0.010 and p=0.047, respectively). The area under the ROC curve (AUC) for TBRmax and TBRmean was 0.68 and 0.74, respectively. Using the LDA-based algorithm, the AUC (0.93) was significantly higher than the AUC for TBRmax. CONCLUSIONS This study shows that in IDH-wildtype glioblastoma, ML-based PSP detection leads to better diagnostic performance compared to conventional ROC analysis.