ESTRO 38 Abstract book

S158 ESTRO 38

PV-0309 Pretreatment ADC does not predict local recurrences in head and neck squamous cell carcinoma B. Peltenburg 1 , J. Driessen 2 , J. Vasmel 1 , R. De Bree 3 , C. Terhaard 1 , M. Philippens 1 1 UMC Utrecht, Radiotherapy, Utrecht, The Netherlands ; 2 UMC Utrecht, Otorhinolaryngology, Utrecht, The Netherlands; 3 UMC Utrecht, Head and Neck Surgical Oncology, Utrecht, The Netherlands Purpose or Objective In head and neck squamous cell carcinoma, pretreatment identification of radio-insensitive tumors would affect treatment planning. ADC has been reported to be a predictor of local recurrence. However, correction for known clinical parameters such as tumor volume has rarely been performed. The aim of this study is to find the added value of ADC to tumor volume in predicting local recurrence. Material and Methods This retrospective cohort study included 217 patients with T2-T4 oral cavity, oropharyngeal, laryngeal or hypopharyngeal squamous cell carcinoma. All patients were treated with (chemo)radiotherapy, prior to treatment an MRI examination was performed. The tumor delineation procedure was semi-automatic. First, a seed point was placed in the tumor on the axial DW-MRI with the highest available b value (800 or b1000 s/mm 2 ) and with a maximum intensity threshold of 50% the tumor was segmented. The delineation was transferred to the ADC map and high intensity areas at the edges of the segmentation were manually removed. The variables obtained from this segmentation were median ADC and total volume. The predictive effect of the variables on local recurrence was analyzed in univariable and multivariable regression. Results Univariable analysis showed no significant correlation between tumor ADC and local control within three years after (chemo)radiotherapy. However, tumor volume was predictive for local recurrence. Multivariable cox regression including ADC an volume showed that tumor volume was an independent predictor of local recurrence with a hazard ratio of 1.032 (CI95% 1.020 – 1.044). ADC was not an independent predictor of local recurrence. Conclusion ADC has no added value in predicting local control in patients with HNSCC. Tumor volume, however, is predictive of recurrence. Figure I: Example of the delineation of a T3N2cM0 oropharyngeal carcinoma. Left: Diffusion weighted image (b800 s/mm2). Right: Corresponding ADC map.

PV-0308 MRI based radiomics improves prognostic assessment in soft tissue sarcoma patients J. Peeken 1 , A. Ott 2 , M.B. Spraker 3 , D. Münzel 4 , M. Devecka 1 , A. Thamer 1 , M.A. Shouman 1 , F. Nüsslin 1 , N.A. Mayr 3 , M.J. Nyflot 3 , S.E. Combs 1 1 Klinikum rechts der Isar- Technical University of Munich, Department of Radiation Oncology, Munich, Germany ; 2 Technical University of Munich, Institut für Medizinische Statistik und Epidemiologie, Munich, Germany; 3 University of Washington, Department of Radiation Oncology, Seattle, USA; 4 Klinikum rechts der Isar- Technical University of Munich, Department of Radiology, Munich, Germany Purpose or Objective Multimodal therapy involving surgical resection and radiotherapy (RT) is regularly performed in patients with high-grade soft tissue sarcomas (STS) of the extremities and other anatomic sites. While local progression-free survival (LPFS) is high, distant progression-free survival (DPFS) and overall survival (OS) remain comparably low. In this work, we sought to determine whether radiomic analysis of multiparametric MRI carries a prognostic benefit for pre-therapeutic individual risk assessment. Material and Methods Fat-saturated T2-weighted sequences (T2FS) and contrast- enhanced T1-weighted fat-saturated (T1wFSGd) sequences were collected from two independent retrospective patients cohorts from the Technical University of Munich (TUM: 73 patients) and the University of Washington (UW: 136 patients). Patient records were assessed for demographics, staging and therapy information. After preprocessing, 2052 radiomic features were extracted. Features with a low correlation between a set of independent segmentations were excluded. For feature reduction and model building to predict OS, DPFS and LPFS the least absolute shrinkage and selection operator (LASSO) method was applied to the TUM cohort. External validation was performed on the UW cohort. Results The LASSO algorithm selected 10, 2, and, 5 features to predict OS, DPFS and LPFS using T1wFSGd-derived radiomic features. Prediction of OS and LPFS achieved better performances (OS: C-index: 0.80 (95% confidence interval: 0.68-0.92), LPFS: C-index: 0.81 (0.66-0.95)) than of DPFS (C-index: 0.65 (0.55-75)). All three models showed lower but significant performance in the validation set (OS: C-index: 0.66 (0.56-0.75), LPFS: C-index: 0.65 (0.54- 0.77), DPFS: C-index 0.58 (0.50-0.67)). A clinical model performed better for OS (C-index: 0.74 (0.65-0.83)) and DPFS (C-index: 0.70 (0.62-0.79) but similar for LPFS (C- index: 0.65 (0.53-0.77)) in the validation set. In multivariate cox-regression models accounting for age, grading and TNM staging, the radiomic scores of OS (HR=2.3, p=0.003) and LPFS (HR=3.7, p=0.007) were significantly associated and improved total model performance up to C-indices of 0.78 (0.68-0.87) and 0.69 (0.57-0.80), respectively. A model combining age and AJCC TNM staging groups with the radiomics scores showed a similar performance. T2FS-based radiomic phenotypes showed overall lower prognostic capabilities (validation set: OS: C-index: 0.61 (0.50-0.72), LPFS: C-index: 0.58 (0.45-0.72=, DPFS: C-index: 0.57 (0.44-0.66)). Combining both MRI sequences did not effect an incremental benefit. Conclusion We first show that a T1-based radiomic phenotype is able to improve prognostic assessment above clinical staging and independent of the T2-based radiomic phenotype for OS and LPFS. As consequence, the radiomics analysis can be simplified by focusing on the T1wFSGd sequence. Such a radiomic phenotype may help to personalize sarcoma therapy above the current clinical staging system.

PV-0310 A field strength independent MR radiomics model for pathological complete response in rectal cancer D. Cusumano 1 , G. Meijer 2 , J. Lenkowicz 3 , G. Chiloiro 3 , L. Boldrini 3 , C. Masciocchi 3 , N. Dinapoli 4 , R. Gatta 3 , C. Casà 3 , A. Damiani 3 , B. Barbaro 3 , G. Maria Antonietta 4 , L. Azario 1 , M. De Spirito 4 , M. Intven 2 , V. Vincenzo 3 1 Fondazione Policlinico Universitario A.Gemelli IRCCS, U.O.C. Fisica Sanitaria- Dipartimento di Diagnostica per immagini- Radioterapia Oncologica ed Ematologia, Roma, Italy ; 2 University Medical Center Utrecht,