ESTRO 2022 - Abstract Book

S1406

Abstract book

ESTRO 2022

1 Lady Davis Institute for Medical Research , Jewish General Hospital, Radiation Oncology, Montreal, Quebec, H3T 1E2, Canada; 2 McGill University ; Medical Physics Unit, Department of Oncology, Faculty of Medicine, and Lady Davis Institute for Medical Research , Jewish General Hospital, Montreal, Quebec, Canada; 3 McGill University ;Medical Physics Unit, Department of Oncology,Faculty of Medicine ; and Research Institute of the McGill University Health Centre, and Lady Davis Institute for Medical Research , Jewish General Hospital, Montreal, Quebec, Canada Purpose or Objective Dynamic Positron Emission Tomography (dPET) is increasingly used for diagnosis and treatment outcome prediction of many diseases including cancer. By injecting radiotracers in the patient’s body and analyzing their kinetics over time, dPET can provide an accurate assessment of the patient’s metabolic response to a treatment. To perform kinetic modelling of dPET data sets, the time-course activity concentration in the patient’s arterial plasma, called the arterial input function (AIF) is required. The gold-standard to measure the AIF is through arterial blood sampling from the patient throughout dPET scans. In our group, a non-invasive radiation detector is under development. The AIF is obtained non-invasively by placing the detector on a patient’s wrist during a dPET scan measuring the number of positrons and photons escaping the radial artery. To accurately measure the AIF with the developed detector, it is essential to know the distance between the radial artery and the skin surface, the surface area of the radial artery, and the blood volume flow (VF) in the radial artery. The aim of this study was to map the human wrist to obtain these parameters. Materials and Methods A 2D ultrasound and a musculoskeletal wide linear array transducer operating at 12MHz was used to scan the human wrist. 23 participants with different height and weight were recruited. 5 wrist scans per participant were performed. Participants were asked to rest their left wrist on a table with palms up, where 3 translational scans at 2, 4 and 6 cm from the wrist crease, 1 longitudinal scan along the radial artery and 1 doppler scan at the 2 cm mark were acquired. Using the first 3 scans, the distance between the artery and the skin as well as the radial artery’s cross-sectional area were measured. The longitudinal scan was used to measure the depth variation of the radial artery along the wrist while the doppler scan was used to measure the VF. Results The average depth of the radial artery at 2, 4 and 6 cm was 2.8 ± 1.2 mm, 3.5 ± 1.5 mm, 4.5 ± 2.3 mm respectively. The average surface of the radial artery was 3.0 ± 1.2 mm 2 . The longitudinal scans showed that the radial artery could have different depths along the wrist, it generally goes deep between 2 and 6 cm from the wrist crease but can change its direction, going upwards towards the skin surface. Using the doppler scan, we were able to measure the VF, which varied between 0.080 mL/min and 4.9 mL/min with an average of 1.6 mL/min. It should be noted that VF depends on the patient’s heart rate during the scan. Measurement of the VF in real-time during a dPET scan, along with the radiation counts detected by the developed non-invasive detector will help us calculate the AIF. Conclusion Mapping the human wrist is an important step in development of the non-invasive radiation detector for application in dPET enabling accurate measurements of the AIF during a dPET scan without drawing blood samples from the patient. A. haidar 1,2,3 , M. Field 1,2,3 , V. Batumalai 1,4 , K. Cloak 1,2,3 , D. Al Mouiee 1,2,3 , P. Chlap 1,2,3 , X. Huang 5,2,3 , V. Chin 1,2,3 , M. Carolan 6 , J. Sykes 7,8 , S. Vinod 1,2,3 , G. Delaney 1,2,3 , L. Holloway 1,2,3 1 University of New South Wales, South Western Sydney Clinical School, Sydney, Australia; 2 South Western Sydney Local Health District, Liverpool and Macarthur Cancer Therapy Centres, Sydney, Australia; 3 Ingham Institute for Applied Medical Research, Medical Physics Research Group, Sydney, Australia; 4 GenesisCare, Radiation Oncology, Sydney, Australia; 5 University of Sydney, ImageX, Sydney, Australia; 6 Wollongong Hospital, Illawarra Cancer Care Centre, Wollongong, Australia; 7 Sydney West Radiation Oncology Network, Radiation Onology, Sydney, Australia; 8 University of Sydney, Institute of Medical Physics, Sydney, Australia Purpose or Objective Data mining and analyses using retrospective radiotherapy imaging datasets sourced from single/multiple centres requires translation of local ontologies for structure names to a standardised ontology. Our aim was to investigate machine learning (ML) based tools for standardising target and organ-at-risk (OAR) volume definition in breast cancer radiotherapy plans. Materials and Methods Radiotherapy imaging data for 1613 breast cancer patients treated between 2014 and 2018 were collected from a single centre. The volumes were initially classified based on discussions with clinicians. 1440 patients were selected for ML model development, and 173 patients were used for testing (hold-out). To represent each target and OAR volume, four characteristics were generated: textual features, geometric features, dosimetry features, and central slices representing the slice with the highest number of contoured pixels in a volume. Five datasets were created from the original cohort, the first four represented different subsets of volumes and the last represented the whole list of volumes (Table1). For each dataset, 15 sets of feature combinations were created to see how the use of different attributes affected the standardisation performance. Three types of artificial neural networks were used to model different combinations of features: feed forward neural networks (FFNN), convolutional neural networks (CNNs), and multi-input neural networks (MINN). FFNN were used for training tabular data combinations (e.g. text and dosimetry features), a CNN was used for training imaging data (central PO-1618 Standardising Nomenclatures in Breast Radiotherapy Imaging Data using Machine Learning Algorithms