ESTRO 2024 - Abstract Book

S5114

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

2292

Digital Poster

A comparative analysis of preclinical CT-based radiomics using cone beam- and micro-CT scanners

Bríanna Kerr 1 , Kathryn Brown 1 , Mihaela Ghita 1 , Kate Connor 2 , Ian Miller 2,3 , Liam Shiels 2 , Colum Connolly 2 , Conor McGarry 1,4 , Annette Byrne 2 , Karl Butterworth 1 1 Queen's University Belfast, Patrick G. Johnston Centre for Cancer Research, Belfast, United Kingdom. 2 Royal College of Surgeons in Ireland, Department of Physiology and Medical Physics and Centre for Systems Medicine, Dublin, Ireland. 3 Royal College of Surgeons in Ireland, National Preclinical Imaging Centre, Dublin, Ireland. 4 Belfast Health & Social Care Trust, Northern Ireland Cancer Centre, Belfast, United Kingdom

Purpose/Objective:

Radiomics is a novel field that aims to extract quantitative data (features) from medical images that reflect the underlying biological characteristics of tissues. Radiomics features have potential as biomarkers to improve patient diagnosis and inform treatment decision-making in the clinic. Despite recent advances in radiomics workflows and analysis, there is currently a lack of comparative studies validating radiomics outputs across different preclinical computed tomography (CT) scanners. In this study, we conducted a cross-centre study to compare reliable radiomics features derived from different CT scanners to develop comprehensive radiomics signatures. Inter-scanner variability was assessed in both phantom and preclinical mouse scans to identify a subset of radiomics features which were standardised across both scanners. Reliability was measured using the intra-class correlation coefficient (ICC) for the phantom measurements and in lung, heart, and bone tissues. This study is the first cross-centre comparison of radiomics outputs in preclinical scanners and a comparison of the transferability of cone beam CT (CBCT)- and µCT-based radiomics features. CBCT and µCT scans of an anatomically correct phantom and two mouse models were acquired using onboard imaging from a small animal radiotherapy research platform (SARRP, Xstrahl) and Quantum GX2 (Revvity), respectively. Different segmentation volumes (44, 92, and 238 mm3) and tissue densities (lung, heart, and bone) were segmented in phantom and mouse, respectively, using ITK-SNAP software. Radiomics features were extracted from each region of interest (ROI) using Pyradiomics software. ICC analysis was performed using RStudio and features that had an ICC score of > 0.8 were deemed reliable. The reliability of radiomics outputs was directly compared between both CT-based scanners for changing segmentation volumes and tissue densities in phantom and mice, respectively. To further optimise the radiomics workflow, the impact of using different imaging protocols and normalising pre-processing parameters on feature reliability was also assessed. Material/Methods:

Results: