ESTRO 2022 - Abstract Book

S43

Abstract book

ESTRO 2022

Conclusion Use of generative intensity augmentation to enhance training datasets aids deep learning contouring by improving generalisation to unseen data. The results are promising and suggest that similar modelling may be useful for other contrast- enhanced applications in RT. However, network generalisation may be at the cost of a minor decrease in performance on specific subsets of the data. This trade-off between specialisation and generalisation needs further investigation.

PD-0069 Automatic detection of facial landmarks in paediatric CT scans using a convolutional neural network

A. Rankin 1 , E. Henderson 2 , O. Umney 2 , A. Bryce-Atkinson 2 , A. Green 3 , E. Vásquez Osorio 2

1 University of Manchester, Division of Cancer Sciences, Manchester, United Kingdom; 2 University of Manchester, Division of Cancer Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Manchester, United Kingdom; 3 University of Manchester, Division of Cancer Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Derry, United Kingdom Purpose or Objective In children receiving radiotherapy, damage to healthy tissue and bone structures can lead to facial asymmetries. Tools for automatic evaluation of bone structures on 3D images would be useful to monitor and quantify the development of facial asymmetry in these patients. Due to the lack of available data, there has been little research conducted to assess the performance of machine learning models using paediatric data. Novel approaches such as transfer learning can facilitate the translation of adult-trained models to paediatric data. Here, we investigate the performance of a convolutional neural network (CNN) using transfer learning to detect facial anatomical landmarks in 3D CT scans. Materials and Methods Facial landmarks were selected through an extensive literature review, to contain both left and right components with minimal anatomical differences between adult and paediatric patients (fig 1A). A training dataset of 104 CT scans from adult head and neck cancer patients was annotated by two observers. A random point between the two annotations was used as a form of data augmentation and to reduce the impact of interobserver (IOV) bias. The UNet-based CNN model was trained for 100 epochs, using several commonly used data augmentation steps (e.g., translational shifts and left-right mirroring). A heatmap was produced by the network based on the confidence of the predicted landmark location. The model was evaluated on a test set of 10 scans annotated by an expert. We investigated the IOV between the locations selected by the two observers and the expert (Table 1) to determine if this had a considerable effect on the model’s performance. The best performing adult model was fine-tuned using 10 paediatric scans for a further 100 epochs and then evaluated with 6 unseen paediatric scans. All paediatric scans (age 4-17y) were annotated by an expert. We quantified model performance measuring the 3D Euclidean point-to-point (p2p) distance between the predicted and manual annotations for the best adult CNN model and the final paediatric model.