Local curvature analysis for differentiating Glioblastoma multiforme from solitary metastasis

Abstract

Ambiguous imaging appearance of Glioblastoma multiforme (GBM) and solitary metastasis (MET) is a challenge to conventional Magnetic Resonance Imaging (MRI) based diagnosis. In this study, a local curvature analysis scheme is implemented to enable morphological differentiation between GBMs and METs. The first stage of the scheme takes advantage of a Diffusion Tensor Imaging (DTI) clustering segmentation technique, complemented by post-contrast T1 imaging for final tumor boundary definition. 3D tumor models are generated by morphological morphing interpolation to compensate for low z-axis resolution of a widely utilized MRI acquisition protocol, followed by triangulated surface mesh generation. Five 3D morphology descriptors, based on local curvature analysis, are tested in a pilot case of 12 lesions (8 GBMs and 4 METs) in terms of morphology differentiation capability, utilizing four first order statistics. Statistically significant differences are identified for all five descriptors tested, however for a varying first order statistics. Results demonstrate the potential of morphology analysis in pre-treatment brain MRI tumor differentiation. © 2016 IEEE.