ESTRO 2022 - Abstract Book

S950

Abstract book

ESTRO 2022

Validation 0.67 (95% CI: 0.54-0.71) 0.56 (95% CI: 0.51-0.68) 0.63 (95% CI: 0.56-0.73)

Materials and Methods The study is a fully federated replication of a previous distant metastasis classifier derived from 300 head-and-neck cancer patients. Two hospital cohorts were selected for training, and 2 others for validation. A “privacy-by-architecture” federated DL prototype was developed by extending the open-source distributed learning system VANTAGE6. An authenticated researcher submitted a task to a master server, that distributed a DL network to be trained by the data hosts. The DL model consisted of 3 fully-connected convolutional blocks. In each iteration of training, synchronous averaging and update of model weights were managed by a separate aggregation server connected to each data host via a key and encrypted web interfaces. The aggregation server prevented direct connections between hosts, researcher and master server during training. Once the model training was finished, final globally-averaged weights were stored on the master server for retrieval by the researcher. To find best combination of epoch and weights aggregation frequency, 3 experiments were conducted(1x200, 2x100, 4x50). Schematic diagram of the FL infrastructure is shown in Figure1. Results We demonstrated functional equivalence of a federated DL classifier to its centralized version, without transferring any subject-level data from data host machines. Furthermore, by preventing access to model weights, we added an additional layer of privacy protection. AUCs of 3 hyperparameter search experiments on training and validation data set are shown in Table 1. The centralized learning and federated learning models delivered statistically equivalent performance in terms of accuracy, loss and AUC. 5-fold cross validation mean AUC of centralized and federated DL models are 0.88 vs 0.85 and 0.88 vs 0.86 respectively. Conclusion We demonstrated a privacy-preserving federated DL methodology that enhances privacy protection by preventing visibility of model weights during training; globally-averaged model weights reduce the risk of privacy violation by " reconstruction" attacks on the data hosts. Our architecture may be further enhanced by adding differential privacy and encryption in the future. Besides, experiments with more centers should be conducted. J.A. Domínguez Rullán 1 , A. Hervás Morón 1 , V. Duque Santana 1 , C. Vallejo Ocaña 1 , M. Martín Martín 1 , M. Valero 1 , D. Sevillano Martínez 2 , R. Morís Pablos 1 , S. Sancho García 1 1 Hospital Universitario Ramón y Cajal, Radiation Oncology, Madrid, Spain; 2 Hospital Universitario Ramón y Cajal, Medical Physics, Madrid, Spain Purpose or Objective The objective of this study is to make a synthesis of the prognostic factors in patients with locally advanced head and neck cancer treated with surgery and postoperative chemoradiation with weekly Cisplatin 40mg/m2. Materials and Methods From 2004 to 2020, 133 consecutive patients diagnosed with locally advanced head and neck carcinoma were treated with surgery and postoperative chemoradiotherapy. Cox-regression analysis was used to identify prognostic factors related to recurrence (local, nodal or distant), overall survival (OS) and cancer-specific survival (CSS). Results Mean age at diagnosis was 59 years (range 36-80). 106 (79.1%) were male and 27 (20.9%) were female. 3% of patients were stage II, 9.7% stage III and 87.3% stage IV. All patients received adjuvant radiotherapy up to 66 Gy to the surgical bed concurrently with weekly cisplatin 40mg/m2. Median follow-up was 60 months (4-187). Cancer-specific survival (CSS) was lower in patients with extranodal extension (ENE) (p=0.003), higher pN stage (p=0.017) and if ≥ 2 involved lymph nodes (p=0.02). OS was also lower in patients with ≥ 2 involved nodes (p=0.007), higher pN stage (p=0.004) and if radiotherapy treatment was ≥ 8 weeks (p=0.013). Local relapse was at increased risk in patients with higher pT stage (p=0.001), positive surgical margins (p=0.023) or if less than 5 cycles of weekly-cisplatin were administered (p=0.037). Lymph node relpase-free survival was lower in patients with positive ENE (p = 0.042). Risk of metastasis was higher in patients with extranodal lymph node involvement (p=0.025) and higher pN stage (p=0.003). Conclusion Taking into account clinical and pathological prognostic factors, head and neck cancer patients with higher pN stage and/or ECE are at higher risk of relapse and death so treatment intensification strategies should be explore in this subgroup of patients. PO-1117 Clinical and pathological prognostic factors in postoperative locally advanced head and neck cancer

PO-1118 PTV changes analysis throughout treatment in H&N patients

S. Pena Vaquero 1 , L. Cardoso Rubio 2 , A. del Castillo Belmonte 1 , L. Gómez heras 2 , I. Conles Picos 1 , A. Hurtado Romero 1 , D. Miguel Pérez 1 , D. Alonso Hernández 1 , R. Torres Cabrera 1