Abstract

Info

Title:: Prediction of radiation pneumonitis based on radiomics and dosiomics

Author(s):: ZHOU Lu; WANG Linjing; ZHANG Guoqian; LI Huijun; LIAO Yuliang; WU Shuyu; Department of Radiation Oncology, Affiliated Cancer Hospital and Institute of Guangzhou Medical University, Guangzhou 510095, China

Keywords:: non-small-cell lung cancer radiomics dosiomics radiation pneumonitis machine learning

Abstract:: Abstract: Objective To establish and validate an effective CT-based radiation pneumonitis (RP) prediction model using the multi-omics approach combining radiomics with dosiomics. Methods A retrospective analysis was carried out on 91 non-small-cell lung cancer patients who were treated with radiotherapy from 2019 to 2021 in Affiliated Cancer Hospital and Institute of Guangzhou Medical University. The whole lung excluding clinical target volume (Lung-CTV) was taken as the region of interest (ROI), and the radiomics and dosiomics features were extracted from the CT image and dose distribution of Lung-CTV. The dose-volume histogram (DVH) features, radiomics combined with DVH (radio+DVH) features, and radiomics combined with dosiomics (radio+dose) features were imported into 11 different classifiers for constructing prediction models. The 5-fold cross-validation was used to complete the classification experiment. The area under the curve (AUC) of the receiver operating characteristics (ROC), accuracy, precision, recall and F1-score were calculated to assess the model performances. Results The AUC of radio+DVH and radio+dose was higher than that of DVH model (P<0.05) and radio+dose had higher accuracy and F1-score than DVH and radio+DVH (P<0.05). Conclusion The multi-omics approach using machine learning-based radiomics and dosiomics to predict RP exhibits better performance and is expected to provide guidance for clinical treatment.