ESTRO 2020 Abstract book

S880 ESTRO 2020

Conclusion The performed analysis focused on investigating the usability and usefulness of the proposed phantom, and it showed that Radiomik can be an useful tool to study Radiomic Feature reproducibility and repeatability. [1] Mackin D. et al. Measuring Computed Tomography Scanner Variability of Radiomics Features. Invest Radiol 2015. [2] Dinapoli N. et al, Moddicom: a complete and easily accessible library for prognostic evaluations relying on image features. Proc of the 37th Annual International Conference of the IEEE (2015). PO-1537 Robustness analysis of standardized radiomic features extracted from T2-weighted MR images C. Piazzese 1 , P. Whybra 1 , C. Bernori 1 , B. Omar 1 , E. Spezi 1 1 Cardiff University, School of Engineering, Cardiff, United Kingdom Purpose or Objective Many studies have demonstrated the potential value of radiomics features as biomarkers to derive prognostic/predictive information. However, imaging features are strongly related to the modality used to computed them. While CT or PET are measured in absolute units with physical meaning, magnetic resonance (MR) images are expressed in arbitrary units and require several pre-processing steps to perform quantitative analyses. In this work, we assessed the robustness of MR-based standardized radiomics features when various pre- processing methods (normalization, quantization and T2-weighted MR images and radiotherapy volumes of 47 patients with soft-tissue sarcomas 1,2 were collected. Four methods were used to normalize the images: original grey levels, same maximum or same mean for all images and grey levels dynamics limited to μ±3σ. Grey levels range was then quantized to 8, 16, 32 and 64 bins and the follow interpolation schemes were tested: original resolution, trilinear isotopic resampling to 0.5, 0.8 and 1 mm. Each MR image was pre-processed with 64 combinations of normalization, quantization and interpolation. Radiomic features, in compliance with the IBSI protocol 3 , were automatically extracted using SPAARC 4 (SPAARC Pipeline for Automated Analysis and Radiomics Computing), a software developed in-house. Features reproducibility was assessed with intraclass correlation coefficient (ICC>0.90). Reproducible features were further investigated with Spearman’s rank correlation coefficient to assess patient ranking consistency. Results Out of the 302 standardized radiomic features computed, 121 showed to be reproducible when using various pre- interpolation) are used. Material and Methods

Tests- 9 CT studies were acquired with a Somatom Definition Flash (Siemens, Erlangen, Germany) in helical mode, using the SAFIRE iterative reconstruction algorithm with strength 3, kernel i50f, 2mm slice thickness, 0,25mm pixel size, 120kV and 300, 250, 200, 175, 150, 125, 100, 75, 50 mAs. MIM-Maestro (MIM Software, Cleveland, OH) was used to manually segment 4 ROIs and the Moddicom package (2) to extract Radiomic features. Results A CT cross section of Radiomik (fig 1b) shows the different materials, textures and fill levels (20%, 60%, 100%) of the 23 elements; their mean HU values ranged between -630 HU to 1420HU and 90 HU for epoxy resin. Contrast between inserts and background is good but the boundary of inserts is not sufficiently sharp to enable automatic segmentation with threshold or region growing algorithms. The 4 ROIs considered for Radiomic Features extraction are highlighted in fig 1b). For each radiomic feature (n=74), we calculated the Spearman’s correlation coefficient between feature value and acquisition mAs (fig.2). Radiomic features showing the highest dependency from mAs were mostly not shared across the different ROIs, suggesting a not-negligible interplay between ROI characteristics and CT protocol setting in feature value estimation.