Predicting outcomes in glioblastoma patients using computerized analysis of tumor shape - Preliminary data

Published

Conference Paper

© 2016 SPIE. Glioblastoma (GBM) is the most common primary brain tumor characterized by very poor survival. However, while some patients survive only a few months, some might live for multiple years. Accurate prognosis of survival and stratification of patients allows for making more personalized treatment decisions and moves treatment of GBM one step closer toward the paradigm of precision medicine. While some molecular biomarkers are being investigated, medical imaging remains significantly underutilized for prognostication in GBM. In this study, we investigated whether computer analysis of tumor shape can contribute toward accurate prognosis of outcomes. Specifically, we implemented applied computer algorithms to extract 5 shape features from magnetic resonance imaging (MRI) for 22 GBM patients. Then, we determined whether each one of the features can accurately distinguish between patients with good and poor outcomes. We found that that one of the 5 analyzed features showed prognostic value of survival. The prognostic feature describes how well the 3D tumor shape fills its minimum bounding ellipsoid. Specifically, for low values (less or equal than the median) the proportion of patients that survived more than a year was 27% while for high values (higher than median) the proportion of patients with survival of more than 1 year was 82%. The difference was statistically significant (p < 0.05) even though the number of patients analyzed in this pilot study was low. We concluded that computerized, 3D analysis of tumor shape in MRI may strongly contribute to accurate prognostication and stratification of patients for therapy in GBM.

Full Text

Duke Authors

Cited Authors

  • Mazurowski, MA; Czarnek, NM; Collins, LM; Peters, KB; Clark, K

Published Date

  • January 1, 2016

Published In

Volume / Issue

  • 9785 /

International Standard Serial Number (ISSN)

  • 1605-7422

International Standard Book Number 13 (ISBN-13)

  • 9781510600201

Digital Object Identifier (DOI)

  • 10.1117/12.2217098

Citation Source

  • Scopus