Early variation of FDG-PET radiomics features in NSCLC is related to overall survival - the “delta radiomics” concept

Glioblastomas (GBM) may receive inadequate radiation dose coverage, specially the nonenhanced hypercellular component, using post-Gd T1 weighted and T2 FLAIR images for radiation target definition.A mixture of edema, micronecrosis, tumor cells and normal tissue in a single image voxel result in an unpredictable value on diffusion images using a conventional b-value of 1000 s/mm 2 or less.This study aimed to develop a technique to identify the hypercellular components of GBM by using high b-value diffusion weighted imaging (DWI) to suppress fluid, edema, micro-necrosis and even normal tissue to an extent.The hypercellular volume (HCV) of the tumor was evaluated for radiation dose coverage and prediction of progression-free survival (PFS).Forty patients with newly diagnosed GBM underwent chemoradiotherapy post-resection/biopsy.The target definition was based upon conventional MRI and RT treatment planning followed standard guidelines.Pre-RT DWI was acquired with b-values of 0, 1000, and 3000 s/mm 2 .A HCV was defined on the DWI with b=3000 s/mm 2 by a threshold obtained from normal tissue.The nonenhanced HCV was delineated by comparing to the Gd-enhanced gross tumor volume (GTV-Gd) on T1-weighted images.Radiation coverage of the HCV was evaluated by the 95% prescribed dose-volume (95%PDV) of the planned dose distribution.Association between HCV and PFS or other clinical covariates were assessed using univariate proportional hazards regression models.For the first 21 patients with a minimum follow-up of 18 months, the HCVs varied 0.58-67 cc (median: 9.8cc), 7 times smaller than the FLAIR-defined CTV.The non-enhanced HCV was 0.15-60 cc (median: 2.5cc).However, incomplete dose coverage of the HCV was seen in 14 patients, in whom 6 had at least 1-cc HCV missed by the 95%PDV (range: 1.01-25.4cc).Of the 21 patients, 15 had progressed, 5 patients earlier within 6 months post-RT, and 10 patients >6 months post-RT.HCV and nonenhanced HCV were significant negative predictors for PFS (p<0.002 and p<0.01, respectively).The component of the HCV that was not covered by the 95%PDV was a significant negative predictor for PFS (p<0.05).The proportion of pre-RT HCVs that overlapped with recurrent Gd-enhanced tumor volume was 78% (range: 65-89%) for the 5 earlier progressors, and 53% (range: 0-85%) for the later progressors.The HCV identified by high b-value DWI most likely represents an aggressive component of the tumor, and analysis of the data from the whole cohort is in progress.Pathological validation is on-going.Post-Gd T1 weighted images (a qualitative measure of vascular leakage) and T2 FLAIR images (influenced by edema) are not adequate for radiation target definition of GBM.Selectively boosting the HCV component of GBM will be tested in a prospective clinical trial.