ESTRO 38 Abstract book

S204 ESTRO 38

high risk to persist or recur could be identified before treatment by using the nomogram constructed for this model to guide decisions on the preferred treatment strategies such as neck dissection or treatment intensification.

compared to CT-based planning, despite different clinical experiences. Finally, the superior soft tissue contrast offered by TRUS seems to insure a RO independent prostate planning, which might lead to more homogeneous clinical outcomes.

Proffered Papers: PH 7: Proffered paper: Outcome modelling

OC-0401 Pre-treatment radiomic features predict individual nodal failure in head and neck cancer T. Zhai 1,2 , R.J.H.M. Steenbakkers 1 , L.V. Van Dijk 1 , J.G.M. Vemer-van den Hoek 1 , H.P. Bijl 1 , M. Dieters 1 , W. Noordzij 3 , A. Van der Schaaf 1 , N.M. Sijtsema 1 , J.A. Langendijk 1 1 University Medical Center Groningen, Department of Radiation Oncology, Groningen, The Netherlands ; 2 Cancer Hospital of Shantou University Medical College, Department of Radiation Oncology, Shantou, China ; 3 University Medical Center Groningen, Department of Nuclear Medicine and Molecular Imaging, Groningen, The Netherlands Purpose or Objective Optimal management of neck nodes after definitive radiotherapy for head and neck cancer patients is still under debate. The main concerns are how to identify patients that need neck dissection and lymph nodes that need super-selected neck node resection after treatment. The main objective of this study was to test whether the addition of pre-treatment radiomic features of pathological lymph nodes to clinical prediction models improves the prediction of nodal failure for individual lymph nodes. Material and Methods This was a retrospective analysis in a prospective cohort study, which was composed of 277 node-positive head and neck squamous cell carcinoma patients with 1025 pathological neck nodes treated with definitive radiotherapy with or without systemic treatment. A total of 165 patients with 558 pathological lymph nodes treated before January 2013 were enrolled in the training cohort and 112 patients with 467 pathological lymph nodes treated between January 2013 and June 2016 were enrolled in the validation cohort. Overall 82 pre- treatment CT radiomic features and 8 clinical features from each positive lymph node were analyzed. The endpoint was the cumulative incidence of nodal failure. Clinical, radiomic and combined models were created from the multivariable Cox proportional hazard analysis based on clinical features, radiomic features, and both clinical and radiomic features, respectively. The performances of the models were assessed in the validation cohort. A nomogram was constructed for individualized nodal failure estimation. Results There were 71 and 28 lymph node failures in the training and validation cohorts with 31.1 and 29.0 months median follow up, respectively. Multivariable analysis revealed two radiomic features (Least axis length (LALLN, p<0.001) and Correlation of Grey level co-occurrence matrix (Corre-GLCM, p=0.039)) and three clinical risk factors (WHO performance score (PS,p<0.001), T stage(p=0.009) and gender(p=0.005)). LALLN means the shortest diameter and Corre-GLCM represents the heterogeneity of the lymph node. The combined model showed good discrimination with a c-index of 0.87 (95% CI: 0.80 to 0.95) in the training cohort and 0.79 (95% CI: 0.78 to 0.80) in the validation cohort and was significantly better than models based on clinical features (p<0.001) or radiomics (p=0.002) only (Fig.1). The relation between the risk of nodal failure for male patients with WHO PS>0 and the radiomic features is shown in Fig.2. Lymph nodes with a

Conclusion A pre-treatment prediction model was developed to predict the risk of individual lymph node failure based on non-invasive radiomic and clinical features. This model may support personalized lymph node treatment adaptations in clinical practice after further multi-center validation. OC-0402 Tumour blood perfusion from baseline contrast-based MRI predicts radiation outcome in rectal cancer K. Bakke 1,2 , S. Meltzer 1 , E. Grøvik 3,4 , A. Negård 5,6 , H. Stein Harald 5 , A. Hansen Ree 1,6 , K. Gjesdal 1 , K. Røe Redalen 1,7 1 Akershus University Hospital, Department of Oncology, Lørenskog, Norway ; 2 University of Oslo, Department of Physics, Oslo, Norway ; 3 Oslo University Hospital, Department of Diagnostic Physics- Division of Radiology and Nuclear Medicine, Oslo, Norway; 4 University of South-Eastern Norway, Department of Optometry- Radiography and Lighting Design, Drammen, Norway; 5 Akershus University Hospital, Department of Radiology, Lørenskog, Norway; 6 University of Oslo, Insitute of Clinical Medicine, Oslo, Norway; 7 Norwegian University of Science and Technology, Department of Physics, Trondheim, Norway Purpose or Objective To evaluate the prognostic and predictive potential of parameters obtained from baseline intravoxel incoherent motion (IVIM) diffusion-weighted MRI and multi-echo contrast-based dynamic MRI in locally advanced rectal cancer. Material and Methods 192 patients with suspected rectal cancer were enrolled onto a prospective biomarker study. 45 of these patients scheduled for neoadjuvant radiation consisting of either short-course radiotherapy (5 Gy x 5) (n = 10) or long-course chemoradiotherapy (2 Gy x 25 with concomitant capecitabine) (n = 35) were selected for analysis. At baseline, patients underwent routine MRI followed by an extended diffusion weighted sequence for IVIM analysis and a multi-echo contrast-based dynamic sequence with