ESTRO 37 Abstract book

S1170

ESTRO 37

3 University of Minnesota of Public Health- Minneapolis- Minnesota- USA., Biostatistics, Minneapolis, USA 4 University of Iowa- Iowa City- IA- USA, Electrical & Computer Engineering, Iowa City, USA Purpose or Objective The aim of this work is to investigate the applicability of combining serially derived radiomic features with standard clinical variables into a xerostomia probability computational tool. Material and Methods Patients undergoing RT for oropharyngeal cancer with daily CT-on-rails imaging were reviewed. Xerostomia status at 6-months post-treatment was retrieved as well as chief clinical variables, (e.g. age, smoking status, T- category, chemotherapy, and dose received). A total of 437 scans were analyzed. Ipsilateral parotid glands were contoured on baseline CT and propagated to daily CT images using deformable vector fields generated for IGRT until mid-treatment time point. A total of 145 radiomic features were selected from the categories: intensity direct, neighborhood intensity difference, grey-level co- occurrence matrix, grey-level run length and shape. Spearman correlation was used to reduce the 145 features to 5 features based on a cutoff of 0.7. These features included: ‘LocalEntropyStd’, ‘LocalStdStd’, ‘Compactness2’, ‘Volume’ and ‘Contrast’. These were then used to build and evaluate three distinct types of models (1) using only the baseline (BL) value of the radiomic feature, (2) the ratio between the mid-RT value of the radiomics feature to its initial value at BL, and (3) a functional principal component analysis (FPCA) model that leverages the structure of the temporal trajectory in the evolution of the radiomic features from BL to mid-RT. Afterwards, we ran logistic regression was to explore the capacity of relevant clinical variables to predict xerostomia at 6-month post-RT; either alone or in combination with one of the radiomic-derived models. Results 28 patients were included. At 6 months, xerostomia was reported as follows: minimal or mild (70.4%) vs moderate-severe (29.6%). The corresponding Receiver operating characteristics area under the curve (ROC AUCs) and confidence intervals (CIs) for various predictive models plotted and depicted in Figure & Table 1 . Clinical only model showed worse AUC when compared with any composite clinical/radiomics models. Combination radiomics model included input from all 3 radiomics models: baseline, delta & FPCA. Noteworthy, each of the 3 individual radiomics models performed worse than the clinical model with corresponding AUCs of: 0.59, 0.64, and 0.72, respectively. This suggests that the functional approach yields a superior ROC compared to using either the baseline radiomics feature or using the delta ratio between the mid and initial time points. Table 1. Color legend, AUC and confidence intervals of various models ROC curves

Figure 1. ROC curve for radiomics features kinetics till mid-RT paired with key clinical variables

Conclusion Textural kinetic trajectories from consequential intra- treatment CT scans can predict for subsequent radiation- induced toxicities. Combining clinical and radiomics input can synergistically add up to the predictive capacity of post-RT xerostomia probability computation. EP-2122 Influence of DIR algorithms on correspondence model-based 4D dose accumulation for lung/liver SBRT T. Sothmann 1,2 , N. Mogadas 1 , T. Gauer 2 , R. Werner 1 1 University Medical Center Hamburg - Eppendorf UKE, Department of Computational Neuroscience, Hamburg, Germany 2 University Medical Center Hamburg - Eppendorf UKE, Department of Radiotherapy and Radio-Oncology, Hamburg, Germany Purpose or Objective 4D dose accumulation approaches rely on accurate DIR- based internal motion field estimation. This study, therefore, investigates and compares open source deformable image registration (DIR) frameworks regarding their influence on correspondence model-based 4D dose simulation in stereotactic radiotherapy of liver and lung lesions. Material and Methods Five wide-spread open source DIR frameworks (ANTS, VarReg, DIRART, NiftyReg, Elastix, Plastimatch) were considered, motivated by high ranks in the EMPIRE10 registration challenge. Registration accuracy was evaluated by means of DIRLAB 4D CT data sets and is quantified by landmark-based target registration errors (TRE). Further, regression-based correspondence models that correlate patient-specific internal motion and external breathing signals were built on the basis of each DIR framework (internal motion field estimation) and mentioned 4D CT DIRLAB data. This allows for model accuracy evaluation. Eventually, correspondence model- based 4D dose simulation was conducted employing 4D CT treatment planning data of 10 patients with 6 lung and 9 liver lesions with known 2-years local control of each lesion. For each DIR and patient, an individual correspondence model was built to simulate the delivered 4D dose distribution of the corresponding VMAT treatment plans. During treatment acquired breathing curves were used as external motion information to account for patient-specific motion effects. Deviations between planned dose and retrospectively simulated dose distributions were analyzed by ΔD 95% , the difference of D 95%,4D-Sim and D 95%,Plan with D 95% as the dose received by 95% of the GTV. A potential linkage between ΔD 95% and clinical local recurrence was investigated.