ESTRO 37 Abstract book

S277

ESTRO 37

was to develop a harmonization process to allow radiomic analysis of multicenter cohorts and strengthen statistical relevance of predictive models by external validation. Then, we assessed the robustness of radiomic features computed on multicenter images after post-processing step. Material and Methods 190 GB patients were retrospectively included: G1, N=62 for methodology elaboration (5 MR scanners); G2, N=128 for clinical validation (11 MR scanners). Contrast- enhanced T1-weighted series were collected. Two homogeneous volumes of interest were delineated in the white matter (WM) for each patient from G1 images. Tumor volume (GTV) was manually segmented for G2 patients. Postprocessing was performed using an in-home Python code on G2 images (called G2’ after this step). Intensities were rescaled first to [0−32767] to erase inter- patient differences in intensity range. G2’ images were then resampled on a 0.5x0.5x0.5mm3 grid size. 8 histogram-based and 31 textural features were extracted using LIFEx freeware (Fig.1). Univariate analysis was performed in WM to assess the robustness of features between different datasets (Wilcoxon’s test: p<0.05 means feature value depends on MR scanner). Different methods of discretization were investigated for the calculation of textural features in GTV: relative (RD) or absolute (AD) with 32, 64, 128, 256 levels. Spearman correlation coefficients ρ were computed in GTV between all features, and between each feature and tumor volume.

Conclusion Development of harmonization processes is crucial for radiomic analysis on larger cohorts. We proposed a methodology showing promising results on a large multicenter cohort. These methods will co ntribute to suppress the intrinsic limits of large scale, int er- center radiomics studies. PV-0529 Reproducibility of radiomic feature es in apparent-diffusion coefficient images of rectal cancer A. Traverso 1 , M. Kazmierski 1 , P. Kalendralis 1 , Z. Shi 1 , M. Welch 2 , H. Dahl Nissen 3 , A. Dekker 1 , L. Wee 1 1 Maastricht Radiation Oncology MAASTRO clinic, Radiotherapy, Maastricht, The Netherlands 2 University of Toronto, Department of Medical Biophysics, Toronto, Canada 3 Vejle Hospital, Department of Medical Physics, Vejle, Denmark Purpose or Objective Changes in Apparent Diffusion Coefficient (ADC) values derived from magnetic resonance imaging (MRI) have been shown to potentially predict therapy response in rectal cancer (Monguzzi 2013). Mentioned models could be improved by adding radiomic features. However, to avoid the risk of false positive rates, the stability of radiomic features in ADC maps must be examined. This preliminary study of MRI-radiomic features addresses the stability of ADC features with respect to inter-observer variability, image pre-processing and conversion to binary mask. Material and Methods We used a data set of ADC images of 23 patients taken from the THUNDER rectal cancer trial. Three independent observers manually delineated tumours. Radiomic features were computed with an open source software (pyradiomics) library for first-order intensity statistics (FO), shape metrics (SM) and textural analyses (TA) specifically grey-level co-occurrence matrix (GLCM) and grey level size-zone matrix (GLSZM). Reproducibility was assessed using concordance and intra-class correlation coefficients (CCC and ICC, respectively). The difference between drawn contours was quantified with Dice similarity. Sensitivity to image pre-processing was assessed by (i) in-slice spatial resolution down-sampling by up to half, (ii) histogram bin width quantization from 20 to 100 in steps of 20 and (iii) effect of different interpolation during down-sampling. The effect of filtration (Gaussian, curvature flow smoothing filters, Laplacian edge detection and Gaussian noise) and different conversions to binary mask were tested. Features computed from each combination of parameters were compared to features computed in the raw image Results FO features were the most reproducible, independent of resampling (78% reproducible at highest value) and quantization (89%), but stability was adversely affected by filtering and inter-observer variability. SM were highly sensitive to inter-observer variability, but insensitive to image pre-processing. TA features exhibited high

Results Basic intensity characteristics (min, max, mean) extracted from WM on G1 highlighted a large dependence on MR scanner (p<0.05). RD with 32 or 64 bins limits this dependence for textural features (G1). Only two 1st- order features were robust among all devices, whereas 7 histogram and 7 textural features were robust between 1.5T and 3T magnets (p>0.05, G1). Basic characteristics showed less variability on G2’ in GTV. RD highlighted correlations between features and tumor volume (|ρ|>0.5, 8 features). AD (256 bins in [0−32767]) decreases these correlations and reduces the dependence of features between devices (Fig.2). For 25 pairs of devices, there were at least 25 robust features, and 11 textural features were robust among magnetic field strength (p>0.05, G2’).