ESTRO 38 Abstract book

S519 ESTRO 38

Montpellier, France ; 12 Institut de Cancérologie de la Loire Lucien Neuwirth, Department of Radiation Oncology, Saint-Priest-en-Jarez, France ; 13 Clinique Claude Bernard, Radiation oncology, Albi, France ; 14 INSERM U1037- Centre de Recherches contre le Cancer de Toulouse, Radiation oncology, Toulouse, France ; 15 Université Toulouse III Paul Sabatier, Radiation oncology, Toulouse, France Purpose or Objective Gliomics is a research project with the goal of extracting radiomics from multimodal imaging data of multicenter prospective phase III trial for newly diagnosed glioblastoma (NCT01507506) [1]. This study compares quantitative radiomics features extracted from the boost, ie. the high-dose volume (HDV) with the low-dose volume (LDV) on planning CT scan. The aim is to quantify increased aggressiveness of the tumor in area defined by MRI (the high dose). Material and Methods Eighty Three patients included in the SPECTRO-GLIO trial presenting with and without local recurrence at the time of the radiomics analysis were analyzed in this study, Forty-one patients (49.4 %) were in Arm A, with Stupp protocol (3DCRT or IMRT 60 Gy/2Gy on contrast enhancement (GTV1)+ 2 cm margin with concomitant temozolomide (TMZ) and six months of TMZ maintenance); Forty-two patients (50.6 %) were in Arm B with the standard treatment with an additional simultaneous integrated boost of intensity-modulated radiotherapy (IMRT) of 72Gy/2.4Gy delivered on the MR spectroscopic imaging metabolic volumes of CHO/NAA>2 and contrast- enhancing lesions or resection cavity(GTV2)+ 1cm margin [2]. For each patient, the GTV1 and GTV2 were delineated in the participating centers following guidelines for arm A, delineated in the coordinating center for arm B and were all validated by central review. The CT and structure data in DICOM RT format were imported into IBEX (Imaging Biomarker Explorer software) [3] and automatically processed using in-house developed data analytics. As the majority of relapses were in the CTV, 116 features (GrayLevelCooccurenceMatrix, GrayLevelRunLengthMatrix,Intensity, NeighborIntensityDifference, Shape) [3] were extracted from CTV1 and CTV2 volumes. The Wilcoxon signed-rank test was used for paired comparison and Bonferonni’s correction was applied. Corrected p-value <0.05 was considered statistically significant. Statistical analyses were conducted using R version 3.5.1. Results The statistics analysis showed that of the 116 features extracted, 39 of them proved to be significantly different between CTV1 and CTV2. As reported in table 1(below), 18 of significant features were RayLevelRunLengthMatrix, Intensity and NeighborIntensityDifference with a p value <0.0001.

Table 1: The 18 most significant radiomics features between CTV1 and CTV2 Conclusion There was a significant difference of features between the two doses volumes which may be a reflection of tumor heterogeneity. To consolidate the findings of this study we will extend the radiomics analysis to multimodal MR imaging data [4] that was acquired longitudinally during this trial and correlate the features of interest to survival. [1] Laprie et al, BMC Cancer, under review after minor revision [2] Ken et al, Radiat Oncol 2013 PO-0958 Mortality Risk Stratification Model based on Radiomics Only: Analysis of Public Open Access HNC Data Z. Shi 1 , C. Zhang 2 , T. Zhai 3 , M. Welch 4 , L. Wee 1 , A. Dekker 1 1 Maastricht University Medical Centre, Department of Radiation Oncology Maastro Clinic- GROW – School for Oncology and Development Biology, Maastricht, The Netherlands ; 2 Maastricht University, Department of Data Science and Knowledge Engineering, Maastricht, The Netherlands ; 3 University of Groningen, Department of Radiation Oncology- University Medical Center Groningen, Groningen, The Netherlands ; 4 University of Toronto, Department of Medical Biophysics, Toronto, Canada Purpose or Objective The primary aim of this study was to investigate whether CT image-derived radiomics are able to predict overall survival (OS) of patients diagnosed with primary Head and Neck Squamous Cell Carcinomas (HNSCC). Material and Methods 5 independent cohorts, 411 patients in total were collected in this study, in which patients were treated with radiation only or chemo-radiation therapy as part of their treatment. CT scans with visible artifacts (e.g., metallic dental fillings) within the GTV were excluded from further analysis. The dataset was split into training set (cohort 1, 2 and 3, n=308) and validation set (cohort 4 and 5, n=103) with the ratio of 3:1. CT images were resampled to isotropic voxels of 2 mm via linear interpolation. A total of 1105 features, consisting of [3] Zhang et al, Medical Physics 2015 [4] Khalifa J et al. Eur Radiol. 2016