ESTRO 36 Abstract Book

S12 ESTRO 36 2017 _______________________________________________________________________________________________

Netherlands 2 Princess Margaret Cancer Centre, Cancer Clinical Research Unit, Toronto, Canada 3 Radboud University Medical Center Nijmegen, Radiation Oncology, Nijmegen, The Netherlands 4 VU University Medical Center, Otolaryngology/Head and Neck Surgery, Amsterdam, The Netherlands Purpose or Objective In order to facilitate a more personalized oncology, research into non-invasive tumor related biomarkers is essential. Radiomics, the comprehensive quantification of tumour phenotypes by extracting large numbers of quantitative image features, can capture intra-tumour heterogeneity and gene-expression patterns [1]. It has previously been shown that a radiomic signature may have high value for survival prediction in lung and head and neck cancer patients [1,2]. However, extensive analysis on the quality of this signature has yet to be done. In this project, we validate the existing radiomics signature in 5 validation sets. We hypothesize that the signature performs above the chance level, in terms of discrimination, on each validation set. Furthermore, we expect that high, medium and low risk patients can be identified using the radiomics signature. Material and Methods Five independent Lung and Head & Neck (H&N) cancer cohorts (in total 1418 patients) treated with (chemo- )radiation were used in this study. Radiomic features were extracted from the pre-treatment computed tomography (CT) images. The model was trained on the Institute 1 lung cohort (N=422) and validated on the other datasets (N=996). The outcome is two-year survival following treatment. An exponential curve was fitted to the radiomics signature predictions plotted against overall survival. Risk group allocation for the Kaplan-Meier curves was done by partitioning patients in 3 groups according to radiomics score. Results As can be observed in figure 1, an upward trend of radiomics signature response versus overall survival can be observed for each validation set. Risk groups can be identified using the radiomics signature in every dataset, as is shown in figure 2.

procedures. In mpUS different ultrasound modalities are combined, including contrast-enhanced ultrasound (CEUS), Doppler ultrasound, computerised transrectal ultrasound and elastography. Transrectal greyscale ultrasound is currently the standard imaging tool for the prostate and is e.g. used for guiding seed placement in brachytherapy. The performance reported in the literature varies widely. Several systems for computerised analysis of ultrasound images have been developed. The far best results published are from the artificial neural network/C-TRUS (ANNA/C-TRUS) system. The initial results for a different computerised analysis technique, histoscanning, were favourable, however, recent studies state that histoscanning is not recommended to reliably identify and characterise prostate cancer. Cancer requires angiogenesis to develop into clinically significant disease. The increased perfusion in malignant tissue can be visualised by Doppler ultrasound imaging. Various authors reported additional value of the Doppler techniques. However, the hypervascularity detected by Doppler ultrasound is not based on true angiogenic perfusion but on flow in larger feeding vessels. In contrast-enhanced ultrasound, gas filled microbubbles are administered intravenously during ultrasound imaging. The microbubbles were first used as additional reflectors in combination with Doppler techniques, increasing sensitivity. In the past years, contrast-enhanced ultrasound has emerged, and the technique now exploits the microbubbles nonlinear oscillations to extract a contrast specific image, sensitive enough to detect single microbubbles. Recently, quantification techniques are being developed that extract objective parameters from CEUS data to further improve interpretation. The latest developments focus on the assessment of the dispersion kinetics of the contrast agent passing through the microvasculature to image changes in the microvascular architecture resulting from cancer neoangiogenesis. Most prostate cancers are stiffer than normal prostatic tissue. Two variants of ultrasound elastography exploit this difference in stiffness: quasi-static or strain elastography and the novel shear wave elastography. The latter assesses stiffness by measuring the velocity at which a shear wave travels through the tissues. Shear wave elastography does not require manual cyclic compression of the prostate and quantification is possible because shear wave velocity is an absolute value. Conclusion The ultrasound modalities discussed above exploit different physical characteristics of (malignant) tissue. Combining the modalities has the potential to detect and localise accurately tumours and dominant lesions. Until now only limited data on the performance of combinations of ultrasound modalities have been published. Therefore, it is difficult to determine the exact value of mpUS in e.g. brachytherapy. Due to the advantages of ultrasound over MRI (i.e. more cost-effective, wider available, less time- consuming, more practical, more suitable for perioperative use and more easily combined with therapeutic devices), the frequent use of US modalities in therapy procedures, its enhanced performance in multiparametric fashion, it is expected that mpUS will become an increasingly interesting modality in also brachytherapy.

Proffered Papers: Radiobiological modeling

OC-0035 Characterization and validation of a radiomics signature for NSCLC and head and neck cancer patients A. Jochems 1 , F. Hoebers 1 , D. De Ruysscher 1 , R. Leijenaar 1 , F. Walsh 1 , B. O´Sullivan 2 , J. Bussink 3 , R. Monshouwer 3 , R. Leemans 4 , P. Lambin 1 1 MAASTRO Clinic, Radiotherapy, Maastricht, The