ESTRO 2023 - Abstract Book

S1899

Digital Posters

ESTRO 2023

Fig. 2: Screen capture of the second module for late urinary frequency: NTCP as a function of the OAR dose. Each curve corresponds to a patient. The upper table allows for patient selection. Conclusion AQUILAB OncoPlace® was very useful in project RADprecise to efficiently merge distinct cohorts in a secured environment and to fulfil all steps, from data collection to validation, for faster clinical adoption of new predictive models. OncoPlace® was developed as a modular open platform to ease model testing on additional cohorts while being already ready for predictive clinical routine (Fig. 1). Hence, we will now integrate additional cohorts to test our global methodology and the possibility of continuous model updating. RADprecise was funded by the ERA PerMed Network, Reference Number ERAPERMED2018-244. B.N. Huynh 1 , A.R. Groendahl 1 , O. Tomic 1 , I.S. Knudtsen 2,3 , F. Hoebers 4 , W. van Elmpt 4 , E. Malinen 3,5 , E. Dale 6 , C.M. Futsaether 1 1 Norwegian University of Life Sciences, Faculty of Science and Technology, Ås, Norway; 2 Norwegian University of Science and Technology, Department of Circulation and Medical Imaging, Trondheim, Norway; 3 Oslo University Hospital, Department of Medical Physics, Oslo, Norway; 4 Maastricht University Medical Center, Department of Radiation Oncology (MAASTRO), Maastricht, The Netherlands; 5 University of Oslo, Department of Physics, Oslo, Norway; 6 Oslo University Hospital, Department of Oncology, Oslo, Norway Purpose or Objective Different machine learning (ML) methods, including deep learning (DL) with image-oriented methods such as convolutional neural networks (CNN), were used to predict disease free survival (DFS) and overall survival (OS) in two cohorts of head and neck cancer (HNC) patients. Materials and Methods HNC patients from two centers, Oslo University Hospital (OUS, N=139) and Maastricht University Medical Center (MAASTRO, N=99) with 18F-FDG PET/CT images acquired before radiotherapy, were included. Two types of input data were analyzed: (D1) clinical factors and (D2) PET/CT images with delineated primary tumors (GTVp) and affected lymph nodes (GTVn). The prediction targets DFS and OS were treated as binary responses, in which class 1 indicated an event. Seven models (M1-M7) with increasing complexity levels were trained and validated on the OUS dataset using nested 5-fold cross-validation. The external MAASTRO dataset was used for testing the models on previously unseen data. Five performance metrics were computed: (I) Accuracy, (II) Area Under the Receiver Operating Characteristic Curve (AUC), (III) Matthews correlation coefficient (MCC), F1 score on class 1 (IV) and class 0 (V) separately. As the event ratios were different between the two datasets (DFS: 49% (OUS), 41% (MAASTRO); OS: 60% (OUS), 54% (MAASTRO)), all metrics were calculated from 1000 bootstrap samples from each dataset, using a 1:1 ratio between the two classes. Prediction models based on clinical factors only (D1) were constructed using the conventional ML methods logistic regression (M1) and random forest (M2), as well as two DL approaches: one using a simple neural network (M3) and the other using a neural network with interactions between network nodes (M4) to learn possible feature interactions within the data. PO-2114 Machine learning and image-oriented methods for head and neck cancer treatment outcome prediction

A downscaled 3D version of the EfficientNet CNN was used to derive patterns or possibly radiomics features from 3D image