OS9.4 The added value of radiomics to a clinical prognostic model in patients with low-grade glioma

Abstract

Abstract BACKGROUND In recent years, radiomics has become of increasing interest in studies of patients with low-grade glioma (LGG). Imaging features extracted from Magnetic Resonance Imaging (MRI) have shown to have the potential to predict genotype and overall survival in patients with LGG. However, until now, the potential of radiomics to add prognostic value to clinical prognostic models has hardly been investigated.The purpose of this study is to investigate the added prognostic value of quantitative magnetic resonance imaging features to a clinical prognostic model for patients with LGG. MATERIAL AND METHODS This retrospective cohort study included adult patients with newly diagnosed LGG who underwent tumor resection or biopsy between October 2002 and March 2017 at the Erasmus MC, University Medical Center Rotterdam (EMC). A set of 77 imaging features were extracted from preoperative T1w and T2w magnetic resonance sequences. These included first order histogram, texture, shape and location features of the volume of interest. The primary outcome of our study was overall survival (OS) and the secondary outcomewas progression-free survival (PFS). RESULTS A total of 259 patients were included. The median follow-up was 5.3 years (interquartile range, 3.4–8.2). The median OS was 9.08 years (95% CI: 6.70–11.5). Four imaging features increased the fit of our clinical prognostic model for OS significantly: T1 local binary pattern peak (hazard ratio (HR) 1.028, p=0.041), T2 histogram energy (HR 1.000, p<0.001), T2 histogram peak (HR 1.000, p=0.002) and standard deviation of radial distance (SDRD) (HR 1.412, p<0.001). The median PFS was 4.14 years (95% CI: 3.49–4.79). Only one imaging feature increased the fit of our clinical prognostic model for PFS significantly: standard deviation of radial distance (HR 1.278, p=0.006). CONCLUSION This study demonstrates that several radiomics features have the potential to help predict OS and PFS in patients with LGG more accurately than models constituted only by clinical features. In our study, four quantitative MRI features were found to be of added prognostic value. However, future studies should validate our results.