ESTRO 2021 Abstract Book

S1540

ESTRO 2021

Conclusion Based on our model, single-field transmission plans achieved highest FLASH effects, while clinical multi-field plans only showed marginal FLASH. The clinical benefits of FLASH plans compared with clinical IMPT were highly patient specific. Within the investigated parameter space, the FLASH effect outweighed fractionation effects for hypofractionated treatments. PO-1811 Multi-modality fusion dosiomics for prediction of survival in nasopharyngeal carcinoma patients Z. Dai 1 , C. Cai 1 , W. Lv 2 , L. Lu 2 , X. Wang 1 , L. Zhu 1 , B. Zhang 1 , G. Yang 1 1 The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Department of Radiation Therapy, Guangzhou, China; 2 Southern Medical University, School of Biomedical Engineering, Guangzhou, China Purpose or Objective To evaluate the prognostic performance of multi-level fusion dosiomics that combined CT and 3D dose based on intratumoral and peritumoral region to predict recurrence-free survival (RFS) and metastasis-free survival (MFS) in nasopharyngeal carcinoma patients treated with intensity modulated radiotherapy. Materials and Methods 219 patients with nasopharyngeal carcinoma receiving radiotherapy from multi-institutions were enrolled. We proposed a multi-level fusion strategy to generate the prediction model by combining dose and CT information at the image-, matrix- and feature-levels based on the datasets from our institution. Models were constructed by 127 radiomics features extracted from(1)dose images alone; (2–7) dose and CT images fused via wavelet-based fusion (WF) using CT-weights of 0.2, 0.4, 0.6 and 0.8, gradient transfer fusion (GTF), and guided filtering-based fusion (GFF); (8) fused matrices (sumMat); (9–10)fused features constructed via feature averaging (avgFea) and feature concatenation (conFea); (11) CT images alone in (1) gross tumor volume of nasopharynx (GTVnx); (2) gross tumor volume of lymph node (GTVnd); (3) RING_3mm_nx; (4) RING_3mm_nd; (5) RING_5mm_nx; (6) RING_5mm_nd respectively. Different combinations of regions of interest (ROI) were considered. The models were also validated in the multi-institutional datasets to evaluate the robustness. The CI and the Kaplan–Meier curves with log-rank analysis were used to assess and compare the performance of these models.