Abstract Book

S1093

ESTRO 37

Material and Methods Eight GC patients, previously treated with photon-beam VMAT, were included in this study. The prescribed dose was 45 Gy, given in 25 fractions of 1.8 Gy. For each patient, a proton plan was generated with the single-field uniform-dose (SFUD) method, using field-specific planning target volumes (PTV) derived from the clinical target volumes (CTV). A two-field configuration with one frontal and one lateral field (incident from the left) was used. The dose calculation and robust optimization to create the SFUD opt plans were done on CT sets in which the Hounsfield units (HUs) of the air cavities along the beam path and within the target volume were replaced with tissue-equivalent HUs. These plans were prepared accounting for 1.0 cm of setup uncertainty and a proton range uncertainty of 3.5 %. To evaluate the plan robustness against possible density changes, the SFUD opt plans were recalculated on the original CT sets which contained gas cavities, to produce the SFUD ver plans. The dose-volume values obtained for relevant organs at risk (OARs) with the VMAT and SFUD plans were then compared. Thereafter, the LKB-model was used to estimate the normal-tissue complication probability (NTCP) for different toxicity endpoints for the kidneys, liver, spinal cord, heart and bowel. A generic RBE of 1.1 was assumed for the proton beams. Results For all the patients, the SFUD plans fulfilled the robustness objectives set for the CTV dose coverage (V 95 > 98% for all scenarios) and for the normal-tissue constraints. The dose-volume values determined for all the OARs, in the SFUD opt - and SFUD ver -plans, were of comparable sizes, showing that robustness against density changes was achieved. For the SFUD plans, significant reductions of the dose-volume values were determined for the kidneys, spinal cord and liver, compared to the VMAT plans ( p < 0.05). For the heart and bowel, the dosimetric values obtained in the VMAT and SFUD plans were comparable ( p > 0.05). In terms of the estimated NTCPs, risks of 0 % were obtained with both the VMAT- and SFUD-plans for the right kidney, liver, spinal cord, heart and bowel ( p > 0.05). However, for the left kidney, a median NTCP value of 9.6 % (range 0 – 86.3 %) was estimated for the VMAT plans and a median NTCP value of 0 % (range 0 – 20.7 %) was estimated for the SFUD ver plans ( p < 0.05). Conclusion The risks for treatment-related side effects were found to be negligible for the two treatment techniques studied, except for the left kidney. The risk reductions obtained for the left kidney with proton-beam therapy could be of clinical relevance for preserving the renal function after radiotherapy of GC. EP-2007 Radiomics based method to predict overall survival of inoperable NSCLC patients L. HUANG 1 , M. Fan 1 , J. Chen 1 , J. Wang 1 , X. Xu 1 , J. Lu 1 , G. Qin 1 , J. Wen 1 1 Fudan University Shanghai Cancer Center, Radiation Oncology, Shanghai, China Purpose or Objective It could be difficult to predict prognostics of inoperable ALK+ mutated patients as they are in large part complicatedly affected by diverse factors in treatment process. The lack of a high-performance and pragmatic indicator for prediction make it imperative to investigate any further. So that we focused on a cohort of inoperable ALK+ mutated patients received targeted or non-targeted therapy and probed into the radiomics based methods for prognostics. Material and Methods Two cohort of clinic patients in Netherlands (n=317) and Chinese (n=54) cancer center was enrolled respectively. The former for model training, received radiotherapy or

chemo-radiotherapy, was downloaded from NSCLC- Radiomics collection in The Cancer Imaging Archive (TCIA). The latter for test, received Targeted treatment (35) or Non-Targeted treatment (19) for ALK+ mutation, were all reported as ALK+ in Fudan University Shanghai Cancer Center. An in-house code-set based on Matlab 2015b (Mathworks, Natick, MA, USA) was used for feature extraction from the pre-treatment computed tomography (CT) images. Totally 203 features were extracted. A test- retest was conducted on RIDER NSCLC dataset and a single-measurement, absolute-agreement, 2-way mixed- effects model was selected for ICC calculation. Features with ICC value larger than 0.9 was considered as stable, which were then incorporated in the least absolute shrinkage and selection operator (LASSO) Cox regression and a leave-one-out cross-validation was practiced. C- index was calculated also in each therapy group. And log- rank test was performed for stratified analysis. Rad_index was calculated as the sum of feature multiplied by its coefficient. (Figure 1).

Figure 1 workflow

Results Three-feature signature was developed and C-index was calculated in test set (CI = 0.65). In each treatment group, C-index of Non-Targeted therapy was 0.85. Nonetheless, poor result was observed in Targeted therapy group (CI = 0.54). Rad_index = 0.04435912* Compactness - 0.01378887* Wavlet_GLCM_Std + 0.01888021* Wavlet_GLRMS_RLN. And Patients were divided into high or low risk according to Rad_index. log- rank p-value was 0.036 in Non-Targeted therapy group (Figure 2).

Table 1 Patient characteristics of two cohort of patients