We've Moved!
Visit SDSU’s new digital collections website at https://digitalcollections.sdsu.edu
Description
Introduction: Prognostic and predictive models based on pre-operative magnetic resonance (MR)imaging features can be a valuable tool in a clinical setting. The goal of this project is to determine the correlation between a highly prognostic genomic signature and a clinically applicable classification based on a combination of clinical (age, Karnofsky Performance Status) and qualitative imaging features (e.g., tumor volume, edema, noncontrast enhancing tumor (nCET), multifocality). Methods: Radiogenomic correlation methods were carried out using the caIntegrator tool. Publicly available data known as The Cancer Genomic Atlas (TCGA)-glioblastoma multiformae (GBM) and expert derived qualitative imaging features from The Cancer Imaging Archive (TCIA) website were used. The caIntegrator tool that allows users to access and search across databases was used to perform radiogenomic correlations. The imaging features were categorized by experts reading pre-operative MRI scans and gene signatures were based on gene expression levels. The correlation of gene expression levels to three imaging features (edema, non-contrast enhancing tumor (nCET), satellites) was explored. Further, the correlation of two clinical features (age and KPS) to gene expression levels was also investigated. Cohorts segregated by median value of the top up/down regulated gene were then analyzed for median survival time. The mean value of the top upregulated gene was determined for each GBM subtype to test for significant differences in GBM sub-types. Results: All three imaging features identified the same top regulated gene Periostin (POSTN). Kaplan-Meier (KM) analysis using POSTN median value showed that subjects in the cohort with greater than median value had a lower mean survival time. The Classical GBM subtype also had significantly elevated values of the POSTN compared to the other sub types. Discussion: The top upregulated gene identified in this study was the same as that identified by an earlier study using the quantitative index of edema volume. This shows that qualitative index of edema volume (this study) was as effective as the quantitative index of edema volume. The other two imaging features (nCET and satellite) were shown to be associated with edema volume; this may partially explain that the same set of genes were upregulated for all imaging features. Based on the findings in this pilot study, a more comprehensive study of more subjects, qualitative and quantitative imaging features is warranted.