P04.16 Multiparametric magnetic resonance imaging revealed tumor niches in glioblastoma patients

Abstract

Abstract Heterogeneity poses challenge to personalised treatment of glioblastoma (GBM). Different habitats within GBM may have diverse treatment response. Magnetic resonance imaging (MRI) is crucial for targeted surgery. However, conventional imaging methods fail to target intratumor habitats. Finding validated imaging method to address GBM regional heterogeneity is crucial. Methods: Patients 112 patients (mean age 59.41 years, range 22–76, 84 males) with newly diagnosed supratentorial GBM were recruited. All patients received surgery and diagnosis was confirmed by pathology. Survival data of 79 patients were available for study. Exclusion criteria include history of cranial surgery/radiotherapy/chemotherapy. The study was approved by local Institutional Review Board. Anatomical, spectroscopy (MRS), dynamic susceptibility contrast (DSC) and diffusion tensor imaging (DTI) were acquired preoperatively with a 3T MRI scanner. All images were co-registered to T2W images. DTI, DSC and MRS data were processed as previously described. Imaging markers, i.e., apparent diffusion coefficient (ADC), relative cerebral blood volume (rCBV) and a series of metabolites were calculated. Region of interest (ROI) was manually delineated on the contrast-enhanced (CE) T1W and FLAIR images. ADC and rCBV values were obtained from each pixel within CE. Thresholds of 0.25 and 0.75 were used for both rCBV and ADC maps to generate four overlapping ADC-rCBV ROIs, which were low-ADC-high-rCBV (LH), low-ADC-low-rCBV (LL), high-ADC-high-rCBV (HH) and high-ADC-low-rCBV (HL) ROIs. The volumetric and metabolic analysis of all ROIs were performed. Results of volume and metabolites were correlated with patient survival. LH, HH, LL, and HL all displayed higher Choline/N-acetylaspartate ratio than normal appearing white matter (<0.001, <0.001, <0.001, <0.01). HL, mainly locating in the necrosis on T1 contrast images, showed higher lactate level than LH, HH, LL (<0.001, <0.001, <0.001) and higher macromolecule/lipid level than LH and HH (<0.05, <0.05). LL showed similar metabolic pattern with HL but may display enhancement on T1 contrast images. The volumes of LL and LH were inversely correlated (<0.001). The proportional volume of LH in CE were positively correlated with non-enhancing tumor volume in FLAIR (<0.01). Kaplan-Meier and multivariate survival analysis showed an increased volume ratio of LH/LL in CE was associated with worse progress free survival (p<0.05). Difference in cellularity and vascularity can create diverse tumor microenvironment. The LH, LL, HL and HH regions may represent the proliferative, hypoxic, necrotic and invasive tumor niches respectively. This approach shows potential in targeting intratumor niches. In next step we will take multiple biopsies and validate this approach at histological and genomic levels.