ESTRO 2022 - Abstract Book

S774

Abstract book

ESTRO 2022

modelling. Combinations of 8 feature reduction methods and 10 ML classification algorithms were compared, producing a risk-stratification model for predicting recurrence OS at 2 years from the first fraction of radiotherapy (Figure 1). Models were compared with 10-fold cross validation and an external test set and benchmarked against TNM-stage and performance status. Youden Index was derived from validation set ROC curves to distinguish high and low risk groups and Kaplan-Meier analyses performed. Results Median follow-up time was 852 days. Parameters were well matched across train-validation and external test sets: Mean age was 73 and 71 respectively. OS rate at 2 years was 54% vs 47% across train-validation and external test sets respectively. The best feature reduction and ML combination was Kendall’s rank correlation followed by an ensemble of Mixture Discriminant Analysis, XG Boost and a single hidden layer neural net. The respective validation and test set AUCs are shown in Table 1. Our model AUC values were superior to TNM stage and performance status in predicting 2-year OS. Kaplan-Meier curves show good separation with significant log-rank test in the external test set.

Conclusion We demonstrate that our model is superior to TNM or Performance status and believe that our methodology can be replicated across health systems using local clinical datasets, without complex imaging and computational requirements, to benefit surveillance stratification for patients following radical radiotherapy for NSCLC globally. Our models are built on routinely available clinical data and set the stage for future clinical trials entailing quantitative personalised risk-stratification and surveillance following curative-intent radiotherapy for NSCLC.

MO-0885 Automatic segmentation of brain structures in longitudinal MR images of growing children

M. Aznar 1 , A. Bryce Atkinson 1 , G. Whitfield 2 , M. van Herk 1 , E. Vasquez Osorio 1

1 University of Manchester, Division of Cancer Sciences, Manchester, United Kingdom; 2 The Christie NHS foundation Trust, The Christie Proton Beam Therapy Centre and University of Manchester, Manchester, United Kingdom Purpose or Objective Children treated for brain tumours may suffer long-term cognitive damage and other sequelae after treatment. Neuroscientists have developed open-access sophisticated software packages, such as FreeSurfer, for automatic segmentation of brain sub-regions structures on MR images. Those tools are potentially of great interest for large multicentric studies to retrospectively estimate the dose received by paediatric patients with brain tumours and evaluate structure growth or atrophy. Here, we evaluate 1) the agreement between FreeSurfer structures and radiotherapy-specific segmentation atlases, and 2) the sensitivity of FreeSurfer to capture volume changes due to aging. Materials and Methods 20 healthy children, each imaged at approximately 5, 7 and 9 years, were selected from OpenNeuro data. All 60 T1- weighted MR images (non-contrast, 1mm slice) were corrected for image inhomogeneity before automatic segmentation of 47 structures using FreeSurfer v7.1.1. The list of FreeSurfer structures was compared to the European Particle Therapy Network atlas (EPTN, Eekers 2021) for concordance in definition. Segmentation quality was visually assessed by a single observer for each image. Volumes of all substructures were calculated for each time point. Results 15/47 FreeSurfer structures could be matched with the EPTN atlas: brainstem (divided into pons, midbrain, medulla oblongata), optic chiasm, cerebellum, corpus callosum, ventricles (enabling the definition of the periventricular space), and left/right hippocampi, caudate nuclei, thalami, and amygdalas. Segmentation quality was judged satisfactory for retrospective dose estimation. Small volume changes were captured for the brainstem, bilateral hippocampi, bilateral