Abstract PO-033: The impact of the variation of CT scanner on the prediction of HPV status in head & neck cancer patients

Abstract

Abstract Radiomics involves the use of high-dimensional quantitative imaging features for predictive purposes. However, studies showed that these radiomic features are sensitive to the variability of imaging parameters (e.g., scanner model). One of the major challenges in radiomics lies in improving the robustness of quantitative features against the variation in the imaging dataset in multi-center studies. Here, we assess the impact of scanner choice on the computed tomography (CT)-derived radiomic features to predict association of oropharyngeal squamous cell carcinoma with human papillomavirus (HPV), which has a well-established impact on CT-derived radiomic features. This experiment was performed on CT image datasets acquired with two different scanner types. We demonstrate strong scanner dependency by developing a machine learning model to classify HPV status from radiological images. These experiments revealed the effect of scanner type on the robustness of the radiomic features, and the extent of this dependency is reflected on the performance of HPV prediction models. The result of this study highlighted the importance of implementing an appropriate approach to reduce the impact of the imaging domain radiomic features and consequently on the machine learning models. Citation Format: Reza Reiazi, Collin Arrowsmith, Farnoosh Abbas-Aghababazadeh, Christopher Eles, Aria Rezaie, Scott V. Bratman, Andrew J. Hope, Benjamin Haibe-Kains. The impact of the variation of CT scanner on the prediction of HPV status in head & neck cancer patients [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-033.