Application of Radiomics Based on 18F-fluorodeoxyglucose Positron Emission Tomography for Predicting of Allograft Rejection in a Rat Lung Transplantation Model

Purpose The 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) has the potential of a noninvasive method for lung allograft rejection (AR) detection, but the accuracy is unsatisfied. Radiomics has been shown to improve the accuracy of diagnosis and treatment response assessment in other fields. Using a rat model of lung transplantation (LTx), we aim to assess the correlation between 18F-FDG PET-based radiomics and lung AR, and compare the predictive power to standardized uptake values (SUVs). Methods Twenty-eight rats were included in three groups at week 3 (Syngeneic, Immunosuppression and AR) and four groups (Syngeneic, Immunosuppression, Hybrid and AR) at week 6. The PyRadiomics package was used to extract the radiomic features from the PET images. The least absolute shrinkage and selection operator (LASSO) algorithm was used to establish the AR-related radiomics signature. A radiomics signature-based nomogram was created to visualize the outcome prediction. We compared the predictive performance of the radiomics signature and SUVs using the area under the receiver operator characteristic curve (AUC). Histopathology results served as a reference. Results From the PET images, a total of 110 radiomic features were extracted, and most of the radiomic features had an acceptable correlation with the histopathological results (P Conclusion The 18F-FDG PET-based radiomics is a potential noninvasive and accurate method to predict lung AR in a rat model of an orthotopic LTx. The radiomics signature showed a better predictive power than SUVs.