Abstract:
|
Tumor heterogeneity is a crucial area of cancer research wherein inter- and intra-tumor differences are investigated for assessing disease development and progression. In this work, we examine Magnetic Resonance Images (MRIs) from Glioblastoma patients to assess image-based tumor heterogeneity. Standard approaches based on scalar summary measures (e.g. intensity histogram statistics) do not adequately capture complete information in the voxel-level data. In this paper, we explore the entire tumor heterogeneity density profiles (THDPs) obtained from the full tumor voxel-space. THDPs are smoothed, probability density function-based representations of MRIs, and we develop tools for statistical analysis of such objects under Fisher-Rao Riemannian framework. The Fisher-Rao metric is used to generate THDP comparisons across patients, which can be used in conjunction with standard clustering approaches. Our analyses reveal two clusters of patients, with marked differences in tumor morphology, genomic characteristics and clinical outcomes. We see enrichment of image-based clusters which give further validity to our tumor heterogeneity representation and subsequent clustering techniques.
|