ESTRO 2022 - Abstract Book

S405

Abstract book

ESTRO 2022

target-volume (D 99 ) and at one-year follow-up were fitted using the linear-quadratic (LQ) and linear-quadratic-linear (LQL) model-based dose conversions to correct for the dose per fraction effect. LQ and LQL models were fitted for non- retreatment lesions. For the retreatment dataset, two different methods were used: (A) Two separate LQ and LQL TCP models for initial (D txt1 ) and second treatment (D txt2 ) (4 models total); (B) one set of LQ and LQL models for retreatment dose with a correction factor of 50%, i.e., D retx =D txt1 x0.5 + D tx2 (2 models total). Model A is based on the common clinical practice for retreatment of brain metastases. The quality of fit was assessed using Chi-2. Results The LQ and LQL models fitted well with the non-retreatment and retreatment lesion data (LQL Chi-2= 0.015-0.388, p=1.0). For our non-retreatment patient data, the fitted EQD2_50, a/b and gamma 50 values were close to HyTEC multi-institution modeling results and the 1-year TCP was estimated at 90% for a D 99 of 17.6Gy. Keeping the same 90% TCP, relatively higher D 99 (24.5Gy) was predicted for the retreatment model B. Interestingly, in model A, the predicted 90% TCP at D 99 for retreatment (D txt2 ) was lower than the initial (D txt1 ) 18.5Gy vs 23.3Gy. The mean fitted LQL EQD2_50 for non- retreatment and re-treatment model A and B, respectively were 15Gy, 39.3Gy (D tx1 ), 19.9Gy (D tx2 ) and 40.3Gy (D retx ).

Conclusion For our patient cohort treated with SRS for brain metastases, the non-retreatment TCP model fitting parameters were in a close approximation to the recent HyTEC brain SRS publication based on pooled data from published studies. This is an effort to understand the TCP in SRS for retreatment of brain metastases. For the 90% TCP in retreatment lesions, using model A, we observed lower D 99 for the second treatment. These unanticipated results highlight possible tumor sensitivity to radiation during re-treatment. Further validation needs to be done on a larger dataset.

OC-0460 Deep learning based time to event analysis with PET, CT and joint PET/CT for H&N cancer prognosis

Y. Wang 1,2 , E. Lombardo 1 , S. Zschaek 3 , J. Weingärtner 3 , A. Holzgreve 4 , N. Albert 4 , S. Marschner 1 , M. Avanzo 5 , G. Fanetti 6 , G. Franchin 6 , J. Stancanello 7 , F. Walter 8 , S. Corradini 1 , M. Niyazi 1 , C. Belka 1 , M. Riboldi 9 , C. Kurz 1 , G. Landry 1 1 University Hospital, LMU Munich, Radiation Oncology, Munich, Germany; 2 Sichuan Cancer Hospital, School of Medicine, University of Electronic Science and Technology of China, Radiation Oncology, Chengdu, China; 3 Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Radiation Oncology, Berlin, Germany; 4 University Hospital, LMU Munich, Nuclear Medicine, Munich, Germany; 5 Centro di Riferimento Oncologico di Aviano (CRO) IRCCS, Medical Physics, Aviano, Italy; 6 Centro di Riferimento Oncologico di Aviano (CRO) IRCCS, Radiation Oncology, Aviano, Italy; 7 ELEKTA SAS, Clinical Applications Development, Boulogne-Billancourt, France; 8 University Hospital, LMU Munich, Radiation Oncology, Munich, Germany; 9 Faculty of Physics, Ludwig-Maximilians-Universität München, Medical Physics, Garching, Germany Purpose or Objective Recent studies have shown that deep learning (DL) is promising for distant metastasis (DM) and overall survival (OS) prognosis in H&N cancer with segmented PET or CT. However, the predictive power could be diminished by the variation in primary and lymph node gross tumor volume (GTV) segmentation. Moreover, the potential of joint modality prognosis remains to be investigated. This study aimed to explore prognosis without GTV segmentation, to extend the single modality prognosis to joint PET/CT, and to investigate the predictive performance with different modality inputs. Materials and Methods We implemented 3D-Resnet and extended it to time-to-event analysis by using an existing survival model to incorporate all censoring and survival information. Publicly available CTs and PETs from 4 different Canadian hospitals (293 patients) and MAASTRO clinic (74 patients) were used for training by 3-fold cross-validation (CV). For independent testing, we used 110 patients from a collaborating institution. All the 477 patients received radiotherapy (RT) or chemo-RT as primary treatment. Different modality inputs were trained and tested, including PET or CT with (by masking the images) or without primary and lymph node GTV contours. For the joint PET/CT prognosis, we set each modality as a separate input channel. The predictive performances were evaluated by Harrell’s Concordance Index (HCI) and Kaplan-Meier curves.