ESTRO 2021 Abstract Book

S295

ESTRO 2021

study evaluates the stability of mpMRI-based dose prescriptions based on DP with respect to the variations of the underlying imaging. Materials and Methods mpMRI data from a test-retest study (including diffusion weighted imaging at b-values of 0 and 700 s/mm 2 , and dynamic susceptibility contrast perfusion imaging) was used to evaluate the stability of DP prescription maps for GBM in a MR-only treatment planning workflow. Publicly available data from 11 newly diagnosed GBM patients (Data from QIN GBM Treatment Response, The Cancer Imaging Archive) acquired on two separate sessions post-surgery and prior to commencing adjuvant chemo-radiation was used. The voxel-wise TP of IT within the clinical target volume (CTV) margins was derived from image features with a logistic regression model, involving the linear combination of diffusion and perfusion MRI-derived parameters. A linear dose mapping function was applied to obtain a DP prescription map. The perfusion and diffusion MRI-derived parametric maps, TP and DP prescription maps were compared between the two MRI sessions ( Figure 1 ) and the intraclass correlation coefficient (ICC) within the CTV margins was calculated to quantify repeatability of the planning workflow. Repeatability was considered excellent for ICC values > 0.9, good between 0.75 and 0.9, moderate between 0.5 and 0.75, and poor < 0.5.

Results Median ICC values were 0.97, 0.69, 0.77, 0.78, 0.81, 0.95 and 0.95 for ADC, rBV, rBF, ADC-rBV TP, ADC-rBF TP, ADC-rBV DP and ADC-rBV DP maps, respectively ( Figure 2 ). The variability between voxels relative to measurement error was lower for diffusion-derived parameters (ADC) than for perfusion-derived parameters (rBV and rBF). However, this variability was reduced in the TP maps obtained from the combination of the individual parametric maps, and even further in the DP maps derived from the respective TP maps.