Identifying Multiple Invasive Intratumor Habitats in Glioblastoma Using Multi-Parametric Magnetic Resonance Imaging and Copula Transform

Purpose/Objective(s)
Glioblastoma (GBM) is characterized by remarkable intratumor heterogeneity. A non-invasive method to identify the intratumor invasive habitats is crucial. Perfusion and diffusion imaging show potential in characterizing tumor properties by describing tumor microstructure and vasculature. The purpose of this study was to investigate whether multi-parametric MRI could identify intratumor habitats for prognosis.
Materials/Methods
Patients We prospectively recruited 115 patients (mean age 59.3 yrs, range 22 - 76 yrs, 87 males) with de novo GBM. All subjects received maximal safe resection and confirmed by pathology. After surgery, temozolomide chemoradiotherapy was performed following the Stupp protocol. Imaging processing Dynamic susceptibility contrast (DSC) and diffusion tensor imaging (DTI) were pre-operatively acquired. All images were co-registered to T2-weighted images. The relative cerebral blood volume (rCBV) maps were generated from DSC images. DTI images were decomposed into isotropic (p) and anisotropic components (q). Contrast-enhancing (CE) regions were manually delineated on post-contrast T1-weighted images. Habitat partitioning DTI-p, DTI-q and rCBV voxels from all subjects were pooled and respectively normalized using the empirical copula transform. Based on the normalized DTI-p, DTI-q and rCBV values, each tumor was partitioned into multiple subregions using k-means clustering. The optimal cluster number was determined using average silhouette method. Statistical analysis The volumes of the tumor habitats were compared using Kruskal-Wallis rank sum test. Kaplan-Meier analysis using Log-rank test was performed to evaluate the prognostic value of the habitats in patient overall survival (OS) and progression-free survival (PFS). The hypothesis of no effect was rejected at a two-sided level of 0.05.
Results
Four habitats were identified from each subject. The proportions of the four habitats within the enhancing tumor were 24.7 ± 3.3 %, 26.0 ± 4.1 %, 25.1 ± 4.5 % and 24.8 ± 4.8 %, respectively. These habitats showed no significant differences in volumes. However, Kaplan-Meier analysis showed four habitats had distinct prognostic values. Specifically, the habitat characterized by highest rCBV value was associated with worsened OS (P = 0.017) and PFS (P = 0.002); the habitat characterized by highest DTI-q value was associated with better OS (P = 0.032) and PFS (P < 0.001).
Conclusion
Diffusion and perfusion imaging can describe key characteristics that are associated with tumor pathogenesis. Our results suggested that integrating perfusion and diffusion imaging may potentially identify clinically relevant intratumor habitats for precise treatment.